Proceedings of the International Cancer Imaging Society (ICIS) 18th Annual Teaching Course

Whole body MRI is recognized as the most sensitive imaging test for the diagnosis of myeloma bone involvement and is recommended by the International Myeloma Working Group for work up of patients with asymptomatic myeloma or solitary plasmacytoma [1]. In some countries such as the United Kingdom, whole body MRI is now recommended as first line imaging in all patients with a suspected diagnosis of myeloma [2]. This is because abnormal signal on MRI has been shown to be associated with very high risk of asymptomatic myeloma progression with development of lytic bone lesions [3,4] and because treatment of patients with high risk asymptomatic disease confers a survival benefit [5]. Contemporary whole body MRI protocols, which incorporate diffusion weighted MRI, combine high sensitivity with quantitative response assessments, in addition to early detection of other features of myeloma such a vertebral collapse, fractures, spinal cord/nerve root progression and extramedullary disease [6]. The incremental net monetary benefit of whole body MRI and whole body CT are reportedly similar compared to a negative benefit reported for FDG PET/CT. Deterministic base case results reveal whole body MRI to have the largest rise in quality-adjusted life years of 0.0142 compared with 0.0119 for whole body CT and 0.0103 for FDG PET/CT [2]. The non-invasive nature of whole body MRI facilitates longitudinal assessments which are particularly important for monitoring patients with non-secretory disease, oligosecretory disease and extramedullary disease. It is important to recognise that patients with secretory disease may become oligosecretory at relapse [7] and therefore as patients with myeloma live longer and undergo multiple lines of treatment, follow up imaging will gain in importance. Whole body MRI can also be used to direct biopsy or to judge the potential of sampling error from posterior iliac crest trephine. The capability of whole body MRI to demonstrate both focal and diffuse marrow infiltration throughout the whole skeleton makes this extremely promising as a subjective tool for monitoring disease status and assessment of response. For focal lesions changes in size and number can be easily assessed and for diffuse infiltration signal changes are evident. However, diffusion weighted MRI offers the capability to quantify disease which is particularly relevant as changes in Apparent Diffusion


Staging
The utility of Positron Emission Tomography (PET) for staging in lymphomas relates to the pattern of marrow involvement, with high sensitivity where there is predominantly focal involvement of the marrow, including Hodgkin and Large B Cell Lymphomas 1 . In these subtypes, PET-CT is very sensitive and more sensitive than the bone marrow biopsy (BMB) which is taken from the iliac crest and will miss sites of focal disease elsewhere in the marrow. Approximately 1% of patients with Hodgkin lymphoma (HL) are upstaged by BMB compared to PET imaging 2 , but this has no impact on management 3 and PET has replaced the BMB at diagnosis 4

.
In diffuse large B cell lymphoma (DLBCL) PET has high sensitivity for large cell lymphoma in the marrow and it is uncommon to miss clinically significant marrow disease 5 . 'Discordant' small cell disease, with a mixture of large cells in the lymph nodes and small cells in the marrow may be missed on PET imaging 5 , however this does not affect prognosis. There is no evidence that altering treatment or closer follow up for patients with discordant disease in the marrow affects outcome, however practice varies as to whether routine BMB is still performed as part of routine staging. Marrow involvement with indolent lymphomas including many follicular lymphomas where there is diffuse, often low grade involvement of the marrow is not well imaged using PET-CT and bone marrow biopsy remains a better test to stage the bone marrow 6 . Response assessment Diffuse uptake frequently occurs in the marrow related to chemotherapy and following administration of myeloid colony stimulating factors, which reporters of PET should recognise and not misinterpret as lymphomatous involvement 7 . Ablation of diseased marrow can occur after successful treatment with photopaenia at these sites. Stimulation of normal adjacent marrow may lead to increased uptake, whereby the uptake in the marrow at response looks like a 'mirror image' of the baseline scan. Correct interpretation of serial scans in this situation relies on having a baseline scan for comparison. Complete metabolic response may be inferred if uptake at sites of initial involvement in the marrow is no greater than surrounding normal marrow even if higher than normal liver uptake using the Deauville criteria 8 . Response in the marrow should be viewed in the context of response in the nodes. If there are persistent focal changes in the marrow in the context of a nodal response, consideration should be given to further evaluation with MRI, biopsy or an interval scan 8 , as increased marrow uptake may persist where there is active healing/ inflammation. Myelofibrosis (MF) is a histological term associated to the development of high density collagen and reticulin fiber depositions in the hematopoietic bone marrow. This histological feature can be observed in numerous disorders from tuberculosis to hematologic disorders such as myelodysplastic syndromes, leukemias and lymphomas. We usually call MF two entities: Primitive Myelofibrosis (PMF) and secondary myelofibrosis (post Essential thrombocytemia (PostET-MF) and polycythemia vera (Post-PV-MF)) that are myeloproliferative neoplasms (MPN). The other pathologies are associated to myelofibrosis but the fibrosis development is usually milder than what is observed in PMF and PET-MF or PPV-MF. Myelofibrosis: pathophysiology, Clinical findings and therapeutics PMF is classified broadly as a chronic myeloproliferative disorder. The main cause of the disorder is the acquisition in the hematopoietic stem cell compartment of mutations in proliferative-associated transduction pathways such as JAK2, CALR or Thrombopoietin-receptor (MPL) genes. These mutations are at least in part responsible for the disease initiation. Secondary molecular events are also frequent in PMF. Whatever the mutational status, the driver mutations are responsible for megakaryocytic hyperplasia, ineffective erythropoiesis, dysplastic megakaryocytes, and an increase in the ratio of immature granulocytes to total granulocytes. The fibrosis development is considered to be reactive to this hematopoietic disorder and is associated to abnormal cytokine releases (TGFb, FGF, VEGF..) by the hematopoietic cells. These releases are also considered as responsible for the constitutional symptoms observed in this disorder (fever, weight loss, night sweats, pruritus…). The impaired bone marrow hematopoiesis is associated to a shift of blood production to spleen and liver, and to spleen and liver enlargement. This disorder is the most severe myeloproliferative disorder with a median life expectancy reduced to 4 years after diagnosis and varying from 6 months to 10 years). Complications associate thrombo-hemorrhagic events and transformations in acute myeloid leukemias. Current treatments are palliative and bone marrow transplantation (BMT) is the only curative approach. Unfortunately, BMT can be proposed only to young patients (median age at diagnosis of PMF is 60 years) without comorbidities. Prognosis classifications have been proposed in PMF and PPV-MF and PET-MF to delineate indications of such therapy and only patients with life expectancy lower than 3 years are considered as BMT-eligible. Other palliative treatments consist in anti-JAK2, Hydroxyurea or Interferon and supportive therapies just like Erythropoietin. Despite palliative, these treatments can increase the life expectancy. Myelofibrosis: The Unsolved Questions concerning imaging Imaging in MF could be relevant in three main aspects i: diagnosis, complications and response to therapies. 1-MPN classification has been recently modified to include the misdiagnosed essential thrombocytemia (ET) patients who present mild myelofibrosis. Because differentiating ET and PMF impact survival, histological analysis is now recommended in ET. Could imaging differentiate ET and smoldering PMF in order to limit the biopsy for ET patients? The second aspect concerning diagnosis is about the quantification of fibrosis and its homogeneity in the bones. It is not yet clear if quantification of collagen and reticulin amounts in the bone marrow are correlated to prognosis. Could imaging answer this question? 2-About the diagnosis of PMF complications and more particularly extramedullary hematopoiesis (EMH) that is a specific complication of this disorder. It is well known that such extramedullary hematopoiesis can involve spleen, liver, lymph nodes, lung, bones, pleural localization are also observed… What is the best way to diagnose such complications? 3-Since new therapies emerge and seem to be more efficient than the ones previously prescribed, fibrosis quantification could be relevant as a marker of response to therapies. What is the kinetic of response of fibrosis and hematopoietic pathogenic clone? appears promising in this clinical context. Two PET tracers have especially been studied to explore MF, 18F-Fluorodeoxyglucose (18F-FDG) and 18F-Fluoro-L-thymidine (18F-FLT).

18F-FDG PET in myelofibrosis
In 30 patients assessed with FDG PET before stem cell transplantation (SCT), Derlin et al. analyzed the distribution of FDG uptake in bone marrow and the spleen, resulting from the inflammation present in MF [1]. They described four patterns of FDG uptake: from grade 1, with a mildly increased uptake in the central skeleton and the proximal extremities, to grade 4 with an increased uptake in the central skeleton and the extremities extending into the small bones of the feet. They showed that there was an inverse correlation between bone marrow uptake extent and histolopathological grade of fibrosis, and an inverse correlation between extent of metabolically active osseous disease and time since diagnosis of MF, suggesting that grade 4 corresponded to early and highly active state of the disease, while grade 1 was observed when a burn-out state of the disease was reached, with established fibrosis resulting from the longstanding BM inflammation. They also retrospectively reviewed pre and post-SCT FDG PET in 12 patients [2]. In most patients with complete response, FDG uptake in bone marrow and spleen decreased, whereas it did not significantly change in patients with partial response or not responding disease. Despite these promising results, they emphasized that the interpretation of FDG uptake in MF could be difficult, because FDG uptake caused by inflammation and by myelopoiesis could not be discriminated. Besides, FDG uptake in BM can be influenced by other factors, such as infection, or injection of granulocyte-colony stimulating factor. 18F-FLT PET in myelofibrosis 18F-FLT is a tracer of cell proliferation, mostly used in the oncological setting for staging and therapeutic evaluation. It has also been proved useful to identify the toxicity of radiotherapy and chemotherapy on BM, illustrating the ability of the tracer to evaluate BM function [3]. It could provide some complementary information in the context of MF. Indeed, 18F-FLT PET is expected to become more and more pathological while the condition worsens [4]. In a pilot study on 15 patients, three patterns of FLT distribution in MF were described, from a slight extension of hematopoiesis with normal uptake in axial skeleton, to a major decrease of uptake in axial skeleton, with or without extension in long bones, and increased uptake in the spleen [5]. Though the patterns did not correlate with histological grade of fibrosis, SUVmax in axial BM did, suggesting that severity of disease could be assessed non-invasively with FLT PET. Besides, splenic myeloid metaplasia could be easily assessed, when compared with standard bone marrow scintigraphy, with a high FLT uptake. This talk reviews the critical concepts of artificial intelligence (AI) and machine learning, describes emerging applications in pelvic tumours, and outlines the limitations and future directions in this field. The aim is to help the audience to integrate machine learning into clinical workflow. AI and machine learning refer to computer algorithms that improve with the exposure of more data, and deep learning is the subtype that attracts particular interest in radiology. Deep learning has the benefit of not requiring feature identification and calculation as a first step; instead, features are identified as part of the learning process. For computer vision tasks such as radiology, convolutional neural networks (CNN) have proven to be effective. The weightings of connections between nodes or neurons are iteratively adjusted based on the inputs and target outputs by back-propagating a corrective error signal through the network.
Recently, several clinical applications of machine learning have been proposed and studied in pelvic tumours. Most prominent applications are found in prostate cancer, including tumour detection, segmentation and computer-aided diagnosis. Machine learning also helps in tumour segmentation in rectal and cervical cancers. Radiologists who become familiar with the principles and potential applications of machine learning would have the opportunity to leverage AI to become a centre of diagnostic information to improve patient outcomes. Imaging requirements for Gastrointestinal Stromal Tumors (GIST) are quite challenging due to the spare occurrence of this tumor entity and its biology, which differs from most epithelial cancers. Also, patterns of appearance and response to treatment are somewhat unique and therefore require distinct knowledge for the treating multidisciplinary team including the radiologist. GIST arise from the interstitial cells of Cajal and are characterized by expression of c-KIT and PDGFR-A; mutations in these genes drive tumor biology and define the medical treatment that can be effective only if certain prerequisites exist. GIST can occur anywhere in the GI tract, yet the most common location is the stomach with over 60% of primary GIST diagnosed there. The role of radiology is detection/specification of tumors, securing diagnosis by performing biopsies if adequate, response assessment and follow up. For diagnosis, multidetector CT (MDCT) is the modality that should be used in the primary workup, with specific technical requirements that need to be followed when performing the scans. MRI is used mainly to assess liver involvement (especially in situations where liver-directed therapies are considered), in rectal GIST and in young patients requiring multiple follow-up examinations. Biopsies should always be performed after consultation with the primary treating physician since the method of choice remains to be endoscopic ultra-sonography-guided biopsy. Percutaneous biopsy approaches should only be performed in situations where no endoscopic biopsy is possible since seeding has been reported quite frequently and thus should be omitted.
Response assessment to treatment with tyrosine kinase inhibitors is one of the major tasks for the radiologistsno other clinical tool enables to monitor treatment effect in this disease. Specific response criteria have been developed, which, quite in opposite to RECIST criteria, take also qualitative tumor appearance into account. Knowledge of those "Choi" criteria is crucial when determining treatment response and examples of pit-falls will be presented. PET/CT and MRI is used in this setting as adjunct modalities, since the Choi criteria are only valid for CT interpretation. An outlook will be given about novel strategies to better assess response in the future. Follow-up is also a crucial topic in GIST since those patients might require many (20 or more) follow-up scans after initial tumor clearance. Individual recurrence risk, which can be predicted and calculated, should be taken into account when recommending the next follow up scan. Some patients will require tight follow-up and some will require no follow-up at all, based on their individual recurrence risk.
In conclusion, radiology plays a central role in the management of GIST and helps the clinician to reach the correct management decisions in the primary diagnosis, during response assessment and in the follow-up of this challenging tumor entity. Since the turn of the millennium biomedical research has undergone a fundamental transition to large-scale experiments and trials that produce massive volumes of highly complex data. "Big data" has been characterized in a four-dimensional space defined by: volume, variety, velocity and veracity. Dealing with this data deluge requires new tools and new approaches to data: capture, curation, storage, search, sharing, transfer, analysis, and visualization. Big data analysis methods enable researchers to identify new directions for research and provide critical support for the development of precision methods for healthcare.
Cancer is a complex, multifactorial disease state. Medical Imaging, particularly Radiology and Pathology imaging, provide valuable diagnostic and prognostic indicators. Imaging based cancer research is undergoing a transition from analyses based on human observers to the use of advanced computing platforms and software to automatically extract quantitative features relevant to prognosis or treatment response. Identifying quantitative imaging phenotypes across scales through the use of robust image analysis and deep learning methods is an evolving approach to improving our understanding of cancer biology. Such methodologies require large collections of well-curated data for development, validation and to ensure research reproducibility. Thanks to advances in imaging technologies and computational platforms, we are able to capture high resolution images in a variety of modalities and compute large volumes of imaging features. This information explosion mandates the need to build a systematic, integrated, reproducible way to evaluate cancer diagnosis, prognosis and treatment. Prediction of response to treatment and development of optimized therapies require integration of Radiology and Pathology imaging information, molecular and genomics data, and clinical information. High quality imaging feature sets from Radiology, (radiomics) and Pathology images, (pathomics) play a critical role in the development of robust and predictive models for characterization of cancer onset and progression. Clinical and research studies now have greater opportunities to incorporate high throughput generation, mining, and integration of imaging features in their research protocols to aid optimization of treatments and better prediction of response to treatment.
As basic and translational investigators dig deeper into the complexity of underlying disease processes, larger sample populations are required for research studies and increasing quantities of data are being generated. In addition, well-curated research and trial data are being pooled in public data repositories to support large-scale analyses. Our ability to extract new knowledge from huge volumes of research data is the key to advances in cancer diagnosis and treatment.

Primary Soft Tissue Sarcoma
Limb-sparing surgery is potentially curative for primary nonmetastatic extremity soft tissue sarcomas. However, the addition of radiotherapy results in higher local control rates [1]. MRI is standard imaging before and after radiotherapy and is critical for surveillance of often complex post treatment appearances. However, histopathologic changes that occur after radiotherapy confound dimensionbased assessments of response. Except for myxoid liposarcoma, significant dimensional radiologic responses are rare. Dimension-based assessments have been shown to have no correlation with outcome or histopathologic response, and alternative assessments are needed [2]. Current consensus from the EORTC-STBSG (European Organization for Research and Treatment of Cancer Soft Tissue and Bone Sarcoma Group) suggests that the addition of functional diffusion-weighted MRI to standard anatomic and post contrast MRI greatly facilitates interpretation because the combination of reduced enhancement and increasing apparent diffusion coefficient (ADC) may increase confidence for response. After radiotherapy, new enhancement alone should be interpreted with caution because this may be a result of vascular disruption rather than histologic response [3]. Machine learning is likely to play a crucial role in allowing us to fully exploit the quantitative capabilities for use in routine, research and adaptive treatment regimens. Increasing conversations in the sarcoma community regarding the role for neoadjuvant chemotherapy in primary soft tissue sarcoma has highlighted the need for an early marker of response. Preliminary data on FDG PET/CT has been promising, indicating that FDG PET/CT predicts histopathologic responses after one cycle of neoadjuvant chemotherapy as well as predicting survival [4].

Metastatic Soft Tissue sarcoma
CT is the mainstay of surveillance and response assessment for metastatic disease. However, patients with the best radiological responses by RECIST 1.1 often have the worst survival outcomes. This is because patients with high grade sarcomas respond very well to chemotherapy but responses are short lived followed by rapid deterioration and short overall survival [5]. This highlights the need to appreciate the significance of imaging responses with regards to outcome. A novel alternative approach to interpretation of RECIST 1.1 response assessments is to shift focus from response to absence of progression. A large EORTC database study has shown that absence of progression in patients with metastatic soft tissue sarcoma does indicate survival benefit [6]. One of the most promising features of Choi criteria, which incorporate size and density for response assessment in Gastrointestinal Stromal Tumours (GIST), is the correlation between response by Choi criteria with improved survival [7]. There is also evidence to suggest that Choi criteria may also be useful in non-GIST sarcoma response assessment [8]. Imaging plays a central role in the diagnosis, staging and management of bone and soft-tissue sarcomas in children. When taken together with clinical findings, imaging can narrow the differential diagnosis and, thus, guide the diagnostic work-up. The most common bone and soft-tissue sarcomas affecting children include osteosarcoma, Ewing sarcoma family of tumors and rhabdomyosarcoma [1,2]. These tumors present unique imaging challenges that are not encountered with other solid malignancies, especially regarding assessment of response to therapy. Most notably, bone tumors do not significantly decrease in size, even when they respond well to therapy. Therefore, the conventional method of assessing solid tumor response (i.e. the RECIST criteria) may not reflect true therapeutic response in bone tumors [2]. Furthermore, response to neoadjuvant therapy, prior to local control of rhabdomyosarcoma has not accurately predicted patient outcome [3]. In this lecture I will review the salient imaging and clinical features that are useful in guiding patient management and will briefly discuss challenges and future directions in assessing pediatric sarcoma response to therapy. In the era of precision oncology, we aim to give the right treatment, to the right patient, for the right time and the right duration. Next generation diagnostic, including imaging and genomics, as well as blood, serum and tissue samples, will be needed to drive next generation clinical trials based on molecular pathology, and will enable to have a measure of tumour activity. One of the main challenges for the next generation diagnostics will be the management of bone metastases, as bone is the first site of metastases for the most common cancers in men and women (prostate and breast). Current imaging tools for bone assessment are limited for detecting presence, volume, viability and response of metastatic bone disease, and may not be adequate for promoting precision oncology. BS has limited sensitivity compared to PET/CT and WB-MRI 1 (78% vs 91% vs 97%, respectively) and does not directly evaluate malignant bone disease. Due to the risk of flare responses, any progressive disease (PD) has to be confirmed on a subsequent BS examination. CT shows osteoblastic and osteolytic activity, but it still shows limitations: RECIST criteria are not applicable for bone metastases (they are considered not measurable 2 ), and osteosclerotic progression and response are difficult to be distinguished. Such limitations for current imaging tools may result in a delayed detection of PD. WB-MRI has demonstrated the potential to detect PD earlier than BS or CT 3 , thus promoting timely interventions and therapeutic changes. Another big challenge for the next generation diagnostics will be capturing tumour heterogeneity. Cancer evolves in time and space, in a kind of Darwinian process 4 . Various factors drive this selection, including anti-cancer therapies, and this evolutionary process eventually leads to the formation of metastases. MET-RADS guidelines 5 for reporting of WB-MRI enable recording of the presence, location and extent of discordant responses in bone metastases, that could be linked to a different gene expression or to the emergence of clonal resistance, and therefore to tumour heterogeneity.
Immunotherapy has become the latest addition to oncological therapeutics. A number of different agents are entering clinical trials but anti-CTLA-4 and anti-PD1/PDL1 agents have been the first to gain regulatory approval and reimbursement in many countries. In all but a few malignancies that have either a high mutational burden or incorporation of viral DNA, overall response rates are low. Nevertheless, immunotherapy has captured the public imagination because of apparent "miracle" cures in patients who have failed all other available therapies. Despite its promise, immunotherapy has significant direct financial costs and, secondarily, those related to management of a wide array of immune-related toxicities. Accordingly, early identification of non-responders could ameliorate the negative impact of ineffective or toxic therapy. In particular, early recognition of immunerelated toxicity could prevent life-threatening complications. Conventional imaging with CT and MRI has been recognized to be suboptimal for early immunotherapy response assessment and has led to recent adaptions of response criteria for such treatment. The use of FDG PET/CT has also come into focus for this purpose but is similarly challenged by the unique mechanisms of response to immunotherapy. A significant issue for both anatomical and molecular imaging is the phenomenon of pseudoprogression. For cancer imaging specialists, it is important to understand that the two broad classes of immunotherapy, anti-CTLA4 and Anti-PD-1/PDL-1 agents, have different mechanisms of action. The former acts primarily on the afferent limb of immune response to expand and recruit immune cells into tumour deposits, whereas the latter enhance immune-cell killing by diminishing immune checkpoints. Accordingly, the anti-CTLA-4, in particular, can lead to a transient increase in metabolic activity and lesion enlargement due to increased infiltrating lymphocytes that precede cell killing as reflected by progression followed by regression, which is the process known as pseudoprogression. The anti-PD1/PDL1 agent generally leads to tumour cells being cleared and a consequent reduction in FDG-uptake paralleling morphological regression. In this regard, these agents lead to responses that are more similar to response to more conventional cytotoxic therapies. When these agents are combined, the degree and temporal profile of these conflicting tendencies to increase or decrease global lesional FDG accumulation can make interpretation more difficult. There can, however, be additional clues that can help to differentiate between true and pseudoprogression. These include development of new lymph node abnormalities in the drainage basin of otherwise stable or responding lesions and activation of lymphoid organs, especially the spleen. Similarly, immune-related complications of therapy are manifold and more common with anti-CTLA4 agents than with anti-PD-1/PDL-1 agents. It is important to be aware of these complications and to conscientiously evaluate scans for them. Most important of these, since they can have fatal consequences are colitis and pneumonitis but other important complications include endocrinopathies, hepatitis and rheumatological disorders. Novel agents are emerging that may further enhance the role of molecular imaging not only in monitoring response but also potentially in selecting patients for treatment with these agents and to identify sites that might benefit from combination therapies. There is, in particular, growing interest in the possibility of enhancing abscopal effects through increased neo-antigenic challenge following radiation of lesions, which could be external or internal.
Introduction: Over the last 10 years, the management of cancer patients has been revolutionized by the advances in molecular targeted therapy and immunotherapy with significant benefits for patient outcomes and comfort. These therapies however are associated with new toxicities and complications that: can be mild, moderate or lifethreatening; may require alteration or cessation of therapy; or simulate disease progression. In this presentation the various classes of immunotherapy associated with pulmonary complications are reviewed and the drug-associated injuries and their differential diagnosis are presented.
Pneumonitis: Drug induced pneumonitis develops in up to 10% of patients on immunotherapy and remains a diagnosis of exclusion that must be differentiated from infection and malignant lung infiltration. Five different patterns have been described on CT: ground glass opacities with preserved bronchovascular markings; increased interstitial markings, interlobular septal thickening, peribronchovascular infiltration, subpleural reticulation, and honeycomb pattern in severe cases ; cryptogenic organizing pneumonia-like, with discrete patchy or confluent consolidation with or without air bronchograms, predominantly peripheral or subpleural in location; non-specific, with a mixture of nodular and other subtypes, not clearly fitting into other subtype classifications. Bronchiolitis Obliterans: There is myxoid fibrous tissue filling the distal bronchioles and extending into alveolar ducts and associated with inflammatory cells. On CT imaging findings include: bilateral regions of patchy consolidation or small irregular nodular opacities, bronchial wall thickening and dilation, and small pleural effusions. Radiation Recall Pneumonitis: This is an inflammatory reaction in previously irradiated areas of lung producing well defined areas of alveolar consolidation, ground glass opacities or infiltrates corresponding to the radiation portals. This pneumonitis usually presents 3-4 months following radiotherapy and the patient presents with cough and dyspnea. Pulmonary Veno-Occlusive Disease: Progressive occlusion of postcapillary pulmonary venules leads to increased pulmonary resistance, pulmonary hypertension, and right ventricular failure. CT findings include diffuse ground-glass opacification, septal thickening, peribronchial thickening, soft tissue oedema around the hila and mediastinum, small pleural effusions, and dilatation of the central pulmonary arteries.
Sarcoid-Like Granulomatous Reactions: Intrathoracic lymphadenopathy simulating sarcoidosis develops in up to 10% of patients following ipilimuab and nivolumab therapy. The adenopathy may manifest and newly enlarged lymph nodes or enlargement of pre-existing lymph nodes that occur in isolation or associated with bilateral upper lobe and middle lobe predominant ground glass opacities, parenchymal consolidations and/or irregular nodules. Most patients are asymptomatic and biopsy show non-caseating granulomas with elevated CD4:CD8 levels.
Extrathoracic diffuse adenopathy and cutaneous non-caseating granulomas have also been described.
Pseudoprogression: Immunotherapy often may initially provoke infiltration of cytotoxic T lymphocytes and other immune cells into the tumor bed. This may cause an increase in tumor size or the development of new lesions as an early response. Pseudoprogression is defined as ≥ 25% increase in tumor burden that is not seen on repeat imaging performed 4 weeks or more after the initial study. Mixed immune-related responses or pseudoprogression are quite problematic in assessing treatment response using RECIST criteria. This abstract has been previously published [4].
Bringing novel cancer drugs from discovery to regulatory approval is a lengthy, expensive and high-risk undertaking that takes about 10 to 15 years to complete with risk of failure at each step. Oncology leads the number of products in clinical development with more than 4,000 projects as of August 2016 [1]. However, only approximately 12% of the drugs entering the clinical trial phase make it to regulatory submission [2], and the average cost to yield a single FDA-approved drug has been reported to range from 648 million USD [3] to as high as 2.6 billion USD [4]. New strategies that can help identify promising new drug candidates early, while eliminating those that are unlikely to be successful could significantly improve the overall drug development process [5]. Cancer imaging is well positioned to make a significant impact on the different phases of drug development, and shorten the timeline between discovery, pivotal preclinical/clinical trials and regulatory approval [6,7]. Current multimodality anatomic, functional, and molecular imaging techniques provide longitudinal whole-body evaluations of the entire tumor burden in one setting and noninvasive in vivo characterization and quantification of biologic, metabolic and molecular processes in the tumor and its microenvironment equivalent to a "virtual biopsy". Cancer imaging technologies can, therefore, serve as a biomarker with potential prognostic value, help enrich trials with proper patient selection and stratification, confirm target hit (or lack thereof) within hours or days after initiation of treatment with potential predictive value, and detect primary and secondary resistance [2][3][4][5][6][7][8].
One of the most exciting recent breakthrough in cancer therapy is the introduction of cancer immunotherapy with immune checkpoint modulators (anti-CTLA4, anti-PD1 and anti-PD-L1), bispecific antibodies, T cell therapy and personalized cancer vaccines [9][10][11][12]. This novel therapeutic strategy can be a "game-changer" for patients with a variety of solid and liquid tumors with some patients achieving long term survival. This paradigm shift has also instigated a wide range of challenges for cancer imaging to minimize translational failures, assess response (or lack thereof), and evaluate toxicity [13,14]. Adding specificity to the tumor microenvironment by developing robust and predictive biomarkers that can assess the presence and phenotypes of immune cells and their activation status prior to and after treatment would greatly help stratify patients and differentiate inflammatory response from tumor progression. These immune imaging biomarkers could be validated as companion diagnostics, and serve as a pharmacodynamic marker, or "virtual biopsy" in integrated prospective preclinical/clinical studies, such as small phase I coclinical trials that can inform the conduct of each trial bi-directionally, and where pre-and post-biopsies are taken and appropriate clinical trial endpoints are chosen. This approach would help assess who to treat, and identify promising drug candidates and combinations with the greatest likelihood of success, efficacy and safety. The systemic inclusion of cancer imaging techniques and probes in prospective pre-clinical/clinical trials or co-clinical trials would greatly contribute to, and increase the successful translation of the development of novel immunotherapeutic drugs for cancer patients in a safe, timely and cost-effective manner. Diagnosis of prostate cancer by systematic TRUS-guided biopsies has led to a substantial raise of the detection of clinically insignificant tumors [1]. Which has resulted in overtreatment and overdiagnosis of prostate cancer. Furthermore, TRUS-guided biopsies misses a substantial proportion of significant tumors because of sampling error, especially in the anterior part and apical region of the prostate [2]. This is related to the fact that TRUS, when performed as a first line imaging modality, lacks both sensitivity and specificity in the detection of cancer foci. Conversely, the high accuracy of multiparametric MRI for localizing significant prostate cancers has been established and MR findings can thus be used to target any suspicious prostatic area to improve the detection of aggressive tumors and to avoid the detection of non-significant tumors. In genitourinary (GU) cancers, like prostate, bladder and kidney many non-vascular interventions are established. The most frequent used non-vascular interventions are diagnostic punctures (in kidney and prostate cancer) and thermal (radiofrequency (RFA), microwave (MWA) or cryotherapy (CRYO)) or non-thermal (irreversible electroporation (IRE)) ablations of kidney and prostate cancer. In kidney cancer the main indication for local ablative therapy are T1a (≤ 4 cm in diameter) tumors, especially in patients with reduced renal function or single kidney. The main competitor of local ablation in the kidney is nephron-sparing surgery or partial nephrectomy (PN). In a retrospective analysis of 1424 T1aN0M0 renal cell cancers treated by PN, RFA or CRYO in the Mayo Clinic the local recurrencefree survival was similar between all three techniques, whereas the metastasis-free survival was significantly better with PN and CRYO compared to RFA. But RFA and CRYO patients were in mean more than 10 years older than PN patients due to the retrospective nature of the study and probably the clinical decision was made to treat the unfitter patients with local ablation and the fitter with PN. Without an prospective randomized trial this clinical selection bias will largely influence the analysis of outcome between surgical and ablative techniques. Many other studies on renal cell tumor ablation could show very similar results of ablation to surgical resection. In prostate cancer due to the higher availability of multiparametric prostate MRI the demand of image-guided prostate biopsies is increasing. Two techniques are well established: MRI-guided biopsy and fusion biopsy of transrectal ultrasound and MRI (MRI-TRUS fusion). Additionally these new technologies are increasing the demand for local treatment of localizes prostate cancer using thermal ablation (high-focused ultrasound (HIFUS)), radioablation (brachytherapy, cyberknife) or biochemical ablation (irreversible electroporation (IRE)).
local expertise. Ultrasound provides real-time guidance, and is widely available and biologically safe. Previously published papers have reported high sensitivity and accuracy values for percutaneous US-FNA, with low complication rates. Percutaneous biopsy or FNA can be also performed under CT guidance, but this technique has some drawbacks, such as the lack of real time visualization and the radiation dose. EUS guidance can be also used for FNA; however, it is not available in all institutions, is more expensive than percutaneous FNA, and must be performed at least under deep sedation. No significant differences in terms of diagnostic accuracy have been reported between the endoscopic and percutaneous approaches for the diagnosis of solid pancreatic malignant masses. Percutaneous FNA and EUS-FNA have similar relatively low complication rates, ranging between 0 % and 5 %. Guidelines from EFSUMB society will be analyzed. This abstract was taken from a previous publication [1]. There is no arterial contact or abutment of the celiac axis (CA), SMA or common (PV)hepatic artery (CHA), nor contact with the SMV or portal vein) or < 180 degrees contact without vein contour deformity/irregularity.

II. BORDERLINE RESECTABLE
A tumor is deemed borderline resectable depending on its location and presence of vascular involvement as follows: 1. Tumor located in pancreatic head/uncinated process: -Tumor contact with CHA without extension to CA or hepatic artery bifurcation allowing for resection and arterial reconstruction -Tumor contact with < 180 degree of SMA -Tumor contact with variant arterial anatomy (accessory right hepatic or replaced right hepatic artery, replaced CHA, and origin or replaced or accessory artery). Presence and degree of tumor contact should be noted -Tumor contact with SMV or PV of > 180 degrees, contact of < 180 degrees with vein contour deformity/irregularity or vein thrombosis but suitable vessel proximal and distal to the site of involvement for resection and venous reconstruction -Tumor contact with IVC 2. Tumor located in body/tail of pancreas: -Tumor contact with CA of < 180 degrees -Tumor contact with CA of > 180 degrees without aortic involvement and uninvolved GDA (permitting performance of Appleby procedure-controversial-some put this in unresectable category) III. UNRESECTABLE -Presence of distant metastasis (incl. non regional nodal involvement) 1. Tumor located in head and uncinate Process -Tumor contact with SMA >180 degrees -Tumor contact with CA > 180 degrees -Unreconstructible SMV/PV due to tumor involvement or occlusion (can be due to tumor or bland thrombus) -Tumor contact with most proximal jejunal branch draining into SMV 2. Body and Tail -Tumor contact > 180 degrees with SMA or CA -Tumor contact with CA and aortic involvement -Unreconstructible SMV/PV due to tumor involvement or occlusion (can be due to tumor or bland thrombus) Limitations of NCCN: The NCCN guidelines are a general set of guidelines that are applicable to most cases of pancreatic ductal carcinomas but has its limitations. Not all scenarios are that may be seen in individual cases as there may be many variations of tumor contact with vessels that have not been included. This was addressed in the recent manuscript by Garces-Decovich A et al [Ref. 4). In this report they report that while the NCCN guidelines accurately classified 90% of the cases they reported on, that in 10% of cases, they encountered 6 scenarios of vascular involvement that were not covered in the NCCN guidelines. In addition, Noda and colleagues have recently reported that the NCCN guidelines are less accurate in predicting resectability when compared to the Japanese Pancreatic Surgery scoring system (Ref. 5).
In addition, they found that a combination of the NCCN and JPS scores that the highest AUC for distinguishing a R0 resection from a R1/R2 resection.

Conclusion:
Overall the designation of resectable, borderline and unresectable categories appears to be suitable for the majority of pancreatic ductal adenocarcinoma, management decisions. But in some scenarios, decisions by management teams at local institutions will have to be made depending on local expertise. The NCCN guidelines are flexible/changeable, and future modifications may be made depending on changing surgical and medical therapies. Pancreatic adenocarcinomas are associated with significant mortality because tumours often present late or with metastatic disease. Patients suitable for surgical resection of the primary disease are associated with better treatment outcomes. In patients with disease of borderline surgical resectability, the use of neoadjuvant chemotherapy can downsize and downstage the tumour to maximise the chance of surgical resection. In this regard, CT and MRI are increasingly used to assess the response of locally advanced pancreatic cancers to neoadjuvant treatment, and the potential for surgical resection. The diagnostic utility of the different imaging techniques in this scenario will be discussed, including the imaging criteria for disease resectability.
For patients with non-resectable or advanced metastatic disease, treatment is usually with chemotherapy and/or radiotherapy. In this setting, conventional CT and MR imaging are most frequently used to assess the effectiveness of therapy. However, with the increasing use of novel therapeutics, including therapies that modify the stromal component of pancreatic cancers and immunotherapies, functional imaging and PET imaging are increasingly used to assess treatment response. Techniques that have been applied in this setting include CT perfusion, diffusion-weighted MRI, dynamic contrastenhanced MRI, MR elastography and FDG-PET/CT imaging. These techniques are now being investigated as potential response, predictive and prognostic biomarkers. Such techniques are also being evaluated for the early detection of disease relapse. Pancreatic ductal adenocarcinoma is a highly aggressive tumor with an overall 5-year survival rate of less than 5%. Prognosis and treatment depend on whether the tumor is resectable or not, which mostly depends on how quickly the diagnosis is made. Chemotherapy and radiotherapy can be both used in cases of nonresectable pancreatic cancer. In cases of pancreatic neoplasm that is locally advanced, non-resectable, but non-metastatic, it is possible to apply percutaneous treatments that are able to induce tumor cytoreduction. To describe the multiple currently available treatment techniques (radiofrequency ablation, microwave ablation, cryoablation, and irreversible electroporation), their results, and their possible complications. This abstract was taken from a previous publication [1]. The PRETEXT classification system has been used as a means to assess paediatric liver tumours in cooperative group trials since its inception. In 2017, paediatric radiologists, oncologists, and surgeons from Europe, Japan, and the United States met to update the classification system via a consensus session. The purpose of this update was to bridge variations across the cooperative groups, standardize terminology, and simplify the methodology for determining vascular involvement. In this lecture, attendees will learn how to apply the 2017 PRETEXT classification system to paediatric liver tumours, how to define the different annotation factors, and how this classification system has changed from prior versions. The Liver Imaging Reporting and Data System (LI-RADS®) was created by the American College of Radiology (ACR) to standardize the reporting and data collection for hepatocellular carcinoma (HCC) [1]. LI-RADS® allows a characterization of liver findings in patients with risk for HCC. These imaging criteria aim to apply a consistent terminology, reduce imaging interpretation variability and errors, and enhance communication with referring clinicians. LI-RADS® was recently updated in 2017 including a modified diagnostic algorithm for CT/MRI, a recommendation for the use of hepatobiliary contrast agents (e.g. Gd-EOB-DTPA), a standardized reporting guidance, a new treatment response assessment algorithm, and a basic management guidance. The focus of this talk is the presentation of LI-RADS® v2017 for characterization and monitoring of hepatic lesions in CT and MRI. Moreover, structured reporting in the daily routine will be discussed. The assessment of response after localized tumor treatment differs from response assessment after systemic tumor treatment. First not all tumor manifestations in the body have been treated, second hemorrhages within the treated tumor could be misinterpreted as contrast enhancement and third tumor diameter and volume could even increase due to treatment of a safety margin of healthy tissue around the tumor.
In the liver in arterial enhancing tumors (e.g. hepatocellular cancer or metastasis from neuroendocrine tumors) the concept of absence of arterial enhancement after thermal ablation or chemoembolisation is an accepted concept of local response assessment (1,2,11). Under these circumstances the modified RECIST criteria are widely accepted for local tumor control. These criteria only taking into account the arterial enhancing portion of the tumor. In radioembolisation (RE, SIRT) of hepatic malignancies the assessment of response differs from these mRECIST criteria, because RE/SIRT is a radiation therapy and similar to external beam radiation RE/SIRT does not alter arterial tumor enhancement (9, 10, 15). For response assessment after RE/ SIRT no established criteria exist, but there is increasing interest in functional imaging, especially with MRI using DWI, perfusion imaging and hepatocyte-specific contrast (gadoxetic acid). After thermal ablation in renal cell cancer hemorrhagic ablation areal could be misinterpreted as a tumor recurrence-suspect arterial enhancement, so a native scan in MRI and CT is essential to quantify the extent of hemorrhagic tissue after ablation, in MRI additional subtraction images after intravenous contrast enhancement could simplify the detection of residual tumor or local tumor recurrence after thermal ablation.
In thermal ablation of lung tumor with CT the most frequent early imaging appearance is ground-glass opacity (GGO) of the ablated area. This early GGOs should extent more than 5 mm beyond the tumor margins (6) to ensure complete tumor ablation. After this early phase the ablated area appears as a well-demarcated homogeneous dense opacity in CT. In the further follow-up imaging this area will show a relatively slow involution with constant decrease of size. Additionally different CT shapes (atelectasis, cavitation, disappearance, fibrosis or nodular) of the treated area have been described and patter changes from one category into another have been described.
To detect tumor recurrences after pulmonal ablation in CT is often difficult, but relies mostly on morphological changes of the ablation area. PET-CT is an alternative to detect tumor recurrancies after thermal ablation, but should not be used within the first 3 months after ablation, because inflammation could mimic active tumor. The current classification introduces the concept of "integrated" or "layered" diagnoses based on a combination of both traditional histological criteria and molecular biomarkers, with the intended goals of improving diagnostic accuracy and patient management. Under the new WHO classification scheme, molecular and genetic data supplement rather than displace histologic classification, even though the rule "molecular beats histology" was introduced [1][2][3]. The result is a major restructuring in many of the brain tumour categories. Major changes include restructuring of diffuse gliomas, medulloblastomas and other embryonal tumours, incorporation of a genetically defined ependymoma variant, and recognition of newly defined entities, variants and patterns, including diffuse leptomeningeal glioneuronal tumour and multinodular and vacuolating tumour of the cerebrum. Several previously recognized brain tumour diagnoses, such as oligoastrocytoma, primitive neuroectodermal tumour, and gliomatosis cerebri, were redefined or eliminated altogether [4,5].
The glioma category has been significantly reorganized, with several infiltrating gliomas in children and adults now defined by genetic features for the first time.
Review of the major changes to the WHO brain tumour classification system that are most pertinent to radiologists is the focus of the present work with special emphasis on gliomas. Medulloblastoma is the most common malignant pediatric brain tumor, and is a rapidly growing, invasive neoplasm that accounts for 40% of childhood tumors in the posterior fossa. In the United States, the overall incidence of medulloblastoma is around 1.5 per million people for the past 40 years, with a marginal male to female preponderance of approximately 1.5:1.4. Although the treatment and mortality of medulloblastomas in children improved significantly in the last decades, this group of pediatric brain tumors remains an important cause of severe morbidity in children.
Not until the recent past, medulloblastomas were classified based primarily on histopathology which included types such as classic medulloblastoma, desmoplastic/nodular, medulloblastoma with extensive nodularity (MBEN), large cell, and anaplastic medulloblastoma.
Recently, a novel classification of medulloblastomas based on the underlying genetic mutation has been introduced and includes 4 different subgroups medulloblastomas: 1) wingless (WNT) group, 2) sonic hedgehog (SHH) group, 3) group 3, and 4) group 4. The differentiation between genetic subgroups of medulloblastomas has shown to be beneficial for therapy decision (different subgroups may benefit from different ,specific therapies) and influence long-term morbidity. More recent work aims to further refine subtype classification based on similarity network fusion and genome-wide DNA methylation and gene expression data, resulting in a total of 12 subtypes. Despite the molecular advances and use of immunohistochemistry markers, logistical challenges remain including costs, lack of widespread availability and practicality to screen all patients with medulloblastoma in order to implement a subtype-specific treatment approach . MR imaging (MRI) is universally used as a standard protocol in all patients with brain tumors and remains the primary method for diagnosis, assessment of therapy efficacy, and surveillance for these tumors. Conventional MRI is useful in identifying and evaluating the different histopathological types of medulloblastomas and may suggest specific molecular subtypes based on morphological pattern, location, extent and contrast enhancement.
Advanced MR imaging such as diffusion-weighted imaging (DWI) and diffusion tensor imaging (DTI) have been shown to provide important additional information about the biology of brain tumors which can be utilized for medulloblastoma imaging, for example the correlation of higher mean apparent diffusion coefficient (ADC) values which were found to be more characteristic of large-cell/anaplastic medulloblastoma than classic medulloblastoma. Preliminary results showed that apparent diffusion coefficient analysis of the enhancing solid component of the tumor may contribute to the pre-operative molecular classification of pediatric medulloblastomas. Common presentations of lung cancer Lung cancer most commonly presents as a pulmonary nodule or mass. The lesion is usually lobulated and/or spiculated due to lymphangitic spread, desmoplastic reaction or distal subsegmental collaps. If it occludes a bronchus a poststenotic atelectasis or pneumonia may be present.

Atelectasis
In rare cases endobronchial tumors may not present as nodule or mass even at CT but only with atelectasis or obstructive pneumonia. Histologically these centrally located endobronchial tumors most commonly represent squamous cell carcinoma. Atelectasis with no extraluminal mass is also commonly found in carcinoid / neuroendocrine carcinoma Lymphadenopathy with no lung lesion In aggressive lung cancer, particularly small cell lung cancer, imaging may show extensive hilar or mediastinal lymphadenopathy with no discernible lung lesionprobably either due to a purely endobronchial primary tumour or coalescence of a small pulmonary tumour with large hilar masses. Differential diagnosis in these cases is mainly Non-Hodgkin lymphoma, which may also present with extensive hilar and mediastinal lymphadenopathy.

Diffuse consolidation
Diffuse consolidation is typical for adenocarcinoma with lepidic growth.
In a single lesion this may mimick a focal inflammatory lesion such as focal pneumonia or granulomatosis with polyangiitis (Wegener´s disease). In more diffuse disease infiltration of larger parts of a segment or lobe it may mimick pneumonia or other infiltrative lung disease. The pattern may be multifocal with involvement of other ipsilateral or contralateral lobes. Interstitial lung disease Occasionally lung cancer may be associated with extensive lymphangitic spread. If the primary pulmonary is small the disease may appear as a diffuse interstitial lung disease mimicking sarcoidosis or other DILD. Lung cancer is the leading cause of cancer deaths and the second most common cancer in both men and women in the United States. Early detection is key to early intervention leading to improved survival. Low-dose CT screening has been shown to significantly reduce mortality but suffers from false positive rates as high as 58%. This not only increases the cost for unnecessary diagnostic tests, but also causes unnecessary anxiety for patients and their families. Quantitative image analysis coupled with deep learning techniques hold the potential to reduce this false positive rate. Numerous computer-aided approaches have been developed for chest image analysis in the past fifty years. Most computational approaches to date focus on detecting and characterizing lung nodules. This dependency on predefined objects of interest requires detection and segmentation steps that are difficult to automate and limit the applicability of many approaches for automated screening. A recent literature review concludes that deep learning is currently becoming the dominant approach in the field and has produced very promising results. Our group has chosen to employ a deep learning technique and to focus on risk prediction at the patient level, taking into account information from the whole lung. Our studies indicate that deep learning approaches can significantly reduce the false positive rate in lung cancer screening and have the potential to produce a fully automated screening assay. Depending on the type of malignancy, children with cancer are treated with different types of (chemo)therapy, all coming with a set of side effects and complication risks. The acute complications associated to therapy can be divided into those related to the effect they may have on the immune status of the patient and those directly related to therapeutic agent toxicity. Signs and symptoms of acute toxicity are myelosuppression, alopecia, nausea and emesis, mucositis, liver function disturbances, and allergic reactions. Immunosuppressive therapy and the disruption of the muco-cutaneous integrity by local tumor invasion, insertion of foreign bodies such as vascular catheters, surgery, radiotherapy and cytotoxic chemotherapy renders patients vulnerable to infection. During the course of their treatment and when a complication is clinically suspected, children with cancer are often referred for imaging studies. Awareness of these possible complications and their presentation allows radiologists to recognize them. In this lecture we will focus on abdominal complications of cancer therapy. The most common types in children will be discussed. As in other parts of the body, the oncological treatment of malignancies in the pelvis may be undertaken with surgery, radiotherapy, medical therapies, and other minimally invasive therapies or with a combination of interventions. The types of treatments depend on the cancer type, stage and patient related factors. In all these imaging plays a central role in assessment of cancer therapies. This is most often looking for response or in the detection of recurrent disease and less commonly for treatment related complications.

A27
In the interpretation of imaging following cancer therapies it is important for the radiologist to understand the malignancy, the treatment administered and to be aware of the spectrum of finding following the different types of treatments. This will allow the assessment of the therapy given to the patient but also to look for complication related to the treatment. For surgical procedures common complications apply to all surgical operations e.g. haemorrhage. However to appreciate other specific complication the radiologist needs to know when and the details of procedures that have been performed e.g. pelvic exenteration or bladder reconstructions. Similar with minimally invasive therapies, the radiologist needs to have knowledge of the procedure and its likely outcomes eg HIFU treatment for prostate cancer. Radiotherapy is commonly used in treating pelvic malignancies. The range and manifestation of radiotherapy on the pelvic viscera depends on the type and dose of radiotherapy and when it was administered. Changes in the target organ as well as other adjacent organs may be seen depending on the types of radiation treatment.
The combination of appropriate clinical information, imaging, including correlation with previous findings will allow for differentiation between therapy-induced complications and tumour recurrence or tumour related complication in the majority of cases. Therefore, imaging plays a central role in the multidisciplinary management of these patients. Pseudo-progression, that has been described in about 3-10% of patients treated using ICIs, corresponds to increase of tumor burden and/or appearance of new lesions due to Infiltration of the tumor by activated T-cells before the disease responds to treatment. To overcome the limitation of RECIST criteria to assess this specific changes in tumour burden, new criteria so-called irRC and then irRECIST were proposed. The major modification involved the inclusion of the measurements of new target lesions into disease assessments and the need of a 4-week CT re-assessment to confirm progression. More recently (2017) a consensus guideline iRECIST was developed by the RECIST working group.

A32
The interventional oncologistthe fourth musketeer of cancer care F Cornelis, I Thomassin, F Boudghene Sorbonne Université -ISCD / APHP -HUEP -Hôpital Tenon, Paris, FRANCE Cancer Imaging 2018, 18(Suppl 1):A32 Interventional Oncology (IO) is currently one of the most active medical specialties. Alongside medical oncology, surgery, and radiotherapy, IO turns out to be the fourth pillar of cancer care. The ongoing research may even expand its scope and importance worldwide. The interventional oncologist is already involved in all stages of the patient's disease, from diagnosis to treatment. Likewise, he or she participates in the management of all related problems due to cancer during the follow-up. Therefore, the interventional oncologist has to set up a perfect organization, workflow and access to appropriate guidance tools or devices. This allows for performing the procedures safely and efficiently while being in daily practice. To achieve these goals and gain recognition, the interventional oncologist must be well-versed with the methods of the "dream team" (i.e. oncologists and surgeons), not only for determining the best-fit treatments but also for the evaluation of treatments. Besides consultation and participation in multidisciplinary team meetings, the interventional oncologist must be in close contact with his colleagues. He or she will thus have a complete appreciation of oncology, identifying each patient's needs and finally cementing his or her place in the team effort involved in cancer care.
IO is rarely proposed as first-line treatment but rather as an alternative in fragile patients. The current recommendations exist based on the accumulated evidence of the benefits of these procedures; however, indications are neither exclusive nor definitive. It is still possible to think outside the box to enlarge them. Hence, interventional oncologists have to be aware of the promises but also the limits of their treatments. It is essential that the interventional oncologist engages himself or herself in fundamental, transversal or clinical research to develop, improve and confirm the therapeutic armamentarium of IO.
To contribute to the further evolution of current guidelines, the interventional oncologist has to defend the minimal invasiveness of his techniques. They cause less pain, fewer side effects and shorter recovery times. Performing these procedures under local anesthesia or mild sedation on an outpatient basis may shorten the time required for treatment and hospitalization. It may even improve the quality of life of patients. Beyond the cost of the devices, which must evolve to be less binding, the rationale behind this is that global costs of these procedures could be less than conventional treatments. This has to be proved by setting up broad medico-economic studies. Expansion of the current indications might occur if oncologists or surgeons are involved in the process and willing to support it. But, it has to be kept in mind that the final goal of interventional oncologists is better cancer care than the mere comparison of techniques. As cancer is a multifaceted disease, interventional oncologists can further provide innovative and effective minimally-invasive solutions to better care for the patients in a multidisciplinary team effort and thereby be recognized as the Fourth Musketeer of the cancer care. This abstract was taken from a previously published work [1]. PET imaging (PET/CT) is routinely used in evaluation of patients with various malignancies including lymphoma. Many of these patients also undergo a separate MRI examination as a problem solving tool. These PET (PET/CT) and MRI examinations are usually performed with temporal delay (at a different time and date). Introduction of hybrid PET/MR systems over the last 5 to 10 years now enables simultaneous or near simultaneous acquisition of PET and MRI information. The role of hybrid PET/MRI is being investigated by many groups in various oncologic applications such as cancer detection, staging, and assessment of treatment response. These hybrid PET/MR systems provide not only metabolic PET information but multi-parametric MRI information (such as diffusion and perfusion weighted imaging) which is temporally and spatially correlated. To test, validate, and clinically implement these systems and to take advantage of all of its various capabilities and functionalities we need to understand: different components of the system and how they have been modified from conventional PET and MRI how do these systems differ from each other and from conventional PET and MRI how to optimize protocols and implement them to answer clinically relevant questions.
We will discuss: ( to CE-CT, in particular for response assessment after chemo-or immune-chemotherapy. There are two sets of rules for lymphoma response evaluation: the Lugano classification [1], and the recently published RECIL classification [2]. For FDG-avid lymphomas, both rely chiefly on the so-called 5-point Deauville score, which is based on a comparison of the lesions' FDG uptake on PET with the FDG uptake of reference tissues, especially the liver. While both classifications distinguish between complete remission, partial remission, stable disease and progressive disease, RECIL includes "minimal response" as an additional category. Apart from this semi-quantitative evaluation, glucose metabolism on [18F]FDG-PET can also be measured quantitatively through calculation of standardized uptake values (SUV), metabolic tumor volume (MTV) and total lesion glycolysis (TLG), which may carry prognostic information [3,4]. Apart from [18F]FDG-PET/CT, whole-body diffusion-weighted MR imaging (DWI) has been proposed as an alternative technique, providing results that are almost, though not quite, as good as those achieved with [18F]-FDG-PET/CT, both in terms of pre-therapeutic staging and treatment response assessment [5,6]. It is based on the restriction of water movement in hypercellular tumors such as lymphomas, and is thus an indirect way of assessing cell density and size, and extracellular space tortuosity.  [7]. A definitive advantage over PET/CT lies in the fact that delayed timepoint PET, which improves detection of some indolent lymphomas [8], can be easily integrated into the PET/MRI work-flow. Among novel PET radiotracers used in lymphomas, [68Ga]Ga-Pentixafor, which targets CXCR4 chemokine receptors [9], has great potential, because CXCR4, which mediates cell chemotaxis and migration, is overexpressed in lympho-proliferative diseases. This abstract has been previously published [10]. Pediatric Lymphoma, both Hodgkin and non-Hodgkin subtypes, has served as the model in pediatric oncology for the use of functional imaging to guide response-based treatment strategies. Over the past 10 years the incorporation of functional imaging, mainly FDG-PET, into standard response assessment algorithms has allowed responsebased protocols to be developed with the goal of reducing late affects related to chemo-and radiation therapy. The purpose of this presentation is to provide an overview of the imaging techniques and response definitions used in Pediatric Lymphoma. We will review both Hodgkin and non-Hodgkin lymphoma, understanding that while many similarities exist, important differences are also notable. The use of PET/MRI will be discussed along with a review of quantitative techniques to refine definitions of FDG-PET response. Finally we will conclude with a brief discussion of surveillance imaging and new tracers and techniques in development for imaging Pediatric Lymphoma.
Bone marrow is one of the largest organs in the body and is visible in every magnetic resonance (MR) imaging study. It is composed of a combination of hematopoietic red marrow and fatty yellow marrow, and its composition changes throughout life in response to normal maturation (red to yellow conversion) and stress (yellow to red reconversion). MR imaging is highly sensitive for detection of altered marrow signal intensity, and the T1-weighted spin-echo sequence provides the most robust contrast between yellow marrow and disease. Heterogeneous red marrow and red marrow hyperplasia can mimic marrow disease, but should be distinguished from neoplastic replacement (leukemia, lymphoma, primary bone sarcomas, hematogenous metastases) and expected posttreatment changes (radiation therapy, chemotherapy, colony-stimulating factor, bone marrow transplant). Nonneoplastic edema-like processes can also alter marrow signal intensity, including trauma, infection, inflammation (chronic recurrent multifocal osteomyelitis, juvenile inflammatory arthritis), altered biomechanics, and chronic regional pain syndrome. Unfortunately, MR imaging findings are often nonspecific and overlap among many of these vastly different causes.
PET-CT has a high sensitivity when assessing marrow infiltration in pediatric malignancies. Advances in radiologic modalities may obviate the use of invasive, painful, and costly procedures like bone marrow biopsy. Furthermore, biopsy results are limited by insufficient tissue or the degree of marrow infiltration (diffuse vs. focal disease). PET-CT can improve the precision of biopsy when used as a guiding tool. Therefore, a definitive diagnosis is reliant on a combination of imaging findings, clinical evaluation, laboratory assessment, and occasionally tissue analysis.

Introduction
Cancer development is characterised by the development of several biological capabilities (cancer hallmarks), which include sustained proliferative signalling, evading growth suppressors, resisting cell death, replicative immortality, inducing angiogenesis, and the activation of invasion mechanisms and metastasis. Mutations in multiple hallmarks are needed for tumor development, progression, spread and therapy resistance. Imaging techniques can map these hallmarks, showing tumour heterogeneity in their distribution, and in so doing have impacts on tumour diagnosis and management. Precision oncology requires detailed knowledge of changing tumor biology, which biopsies (either targeted/liquid) cannot provide, especially for serial & heterogeneous assessments in therapy settings. The ability of functional & molecular imaging to enable noninvasive, quantitative measurements of tumor biology in-situ, allows of the full disease burden to be assayed serially. In so doing, imaging of therapy targets, drug pharmacokinetics and pharmacodynamics can provide markers of therapy efficacy, which are complementary to tissuebased biomarkers.
Tumour proliferation and clonal mutations Cancers growth arises from cumulative, non-lethal genetic damage in mechanisms that normally control cell proliferation. Mutations and epigenetic DNA methylation enhance cellular circuits able to maintain proliferation, despite 'bottlenecks' of lethal mutations, immunity and environmental (including treatment related) barriers. At the same time, mutations also drive other cellular machinery needed for sustaining growth, progression and therapy resistance. It is possible to observe tumour cellularity and cell viability with diffusion MRI, and to quantify diffusion imaging using ADC values (x10 -3 mm 2 /s; μm 2 /s). The mechanisms of programmed tumor cell death (PCD) affects ADC changes in response assessment settings. For example, necroptosis results in the death of large numbers of cancer cells, including stroma destruction. There is much inflammation and extracellular oedema accompanying cell and nuclear lysis (membrane disruption) which results in marked increases in free water pool and higher ADC values. On the other hand, apoptotic cell death results in fewer cells being lost without adjacent cell/stroma damage. Organelles and proteins in apoptotic bodies are intact. Lipid droplets are present and little inflammation is observed. Therefore, ADC change depends on balance between apoptosis, autophagy & tumour repopulation, and only modest ADC rises maybe seen. Angiogenesis Neoangiogenesis is critical for tumour growth and driven by low tissue oxygen levels. Developing neovasculature is disorganised, with irregular and tortuous vessels. Structural abnormalities of tumour vessels are the basis for differential contrast enhancement. The clinical deployment of anti-vascular drugs has made imaging of angiogenesis important in assessing therapeutic efficacy. Traditional anatomic imaging is insensitive to detect early changes as antiangiogenic therapies are cytostatic. Multiple quantitative imaging techniques can depict the functional properties of neoangiogenic blood vessels. Two major dynamic contrast enhancement MR techniques are commonly used: dynamic susceptibility (T2*W DSC-MRI) and dynamic contrast enhancement (T1W DCE-MRI).

Metabolic reprogramming in cancer
The oncogenes promoting cancer cell proliferation are involved in the metabolic transformation via HIF-1 α /MYC, PI3/AKT, mTOR & AMPK pathways. Mitogenic signals promote nutrient uptake & synthesis of building blocks of DNA, RNA, proteins and lipids. Heterogeneity in environment pO 2 , pH levels and nutrient availability combine with altered tumor metabolism, resulting in a continuous supply of building blocks that allow cancer cells to survive and proliferate under selective pressures. Common metabolic events in cancer include (1) anaerobic glycolysis even in conditions of high oxygen tension (Warburg phenomenon),

Conclusions
Multiparametric imaging enables biologic assessments of cancer because of its multidimensional and quantitative nature. Imaging biomarkers are multispectral (multiple imaging biomarkers reflecting multiple biologic properties), multispatial and temporally resolved. Multiparametric imaging enables more accurate detection, localization, characterization and response assessments. Precision oncology (right treatment for the right patient, at the right time, for the right duration) requires appropriately validated imaging biomarkers (and tissue BMs Head and neck cancer treatment Head and neck squamous cell carcinoma is the most common cancer in the head and neck. Concurrent chemoradiotherapy is the standard non-surgical treatment for advanced stage disease but~25-50% of cancers are resistant to treatment. Therefore, there is a need for response assessment at primary and nodal sites early in the course of treatment. Treatment options for patients with non-responding cancers include radiotherapy escalation, addition of targeted agents, early cessation of additional induction chemotherapy regimens, or change to surgical treatment, while potential options for patients with responding cancers include de-escalation of treatment to reduce side effects. In addition, response assessment early after the end of treatment (~6 weeks) is desirable to allow more timely detection of patients with residual cancer in the head and neck who would benefit from salvage surgery or a radiotherapy boost. Imaging modalities and assessment parameters Assessment of treatment response has traditionally been based on change in tumour size, but size reduction during treatment lags response and may not be an accurate indicator of long-term outcome. Furthermore, in the post-treatment setting residual masses may be caused by residual cancer or benign post-treatment inflammation and scarring. MRI and FDG-PET provide valuable additional functional parameters for response assessment. These two modalities are being used in the post-treatment setting and are under intense investigation for intra-treatment assessment. This lecture will emphasise the role of MRI which has the advantage of providing both anatomical and functional parameters in the same examination. The emphasis will be on functional MRI using DWI but will also include DCE-MRI, MRS and CEST, and the complimentary role of functional MRI and FDG-PET in multiparametric imaging. Standardisation of functional MRI techniques remains problematic (imaging protocols, map analysis methods and outcome measures) and is complicated by an increasing array of MRI parameters and differing thresholds. Despite these limitations there are promising trends in early response assessment and these are summarised below. Early intra-treatment response assessment DWI is the most promising functional MRI technique. In the first few days of treatment cancers may show a transient decrease in ADC (cell swelling), but thereafter responding cancers show an increase in ADC as cells die, this increase being aided in the early phase by a radiation-induced increase in perfusion. Differences between responders and non-responders are detected early after the start of treatment (~2 -3 weeks), responders tending to have a significantly higher rise in ADC (> 14% -24%) and shift of ADCs to the high ADC range, compared to non-responders. Further research is still ongoing regarding the exact timing of the first intra-treatment scan and the value of serial imaging, but a fall in ADC after an initial rise (repopulation of tumour cells) is linked to an unfavourable response. There is less research into DCE-MRI and FDG-PET, but early intra-treatment rises in K trans and Ve (2 weeks), and larger decreases in FDG activity (1-2 weeks) have been observed in responders compared to nonresponders. Very early results using CEST imaging observe a rise in the levels of amide protons in non-responders, while MRS at 2 weeks has not been shown to predict response. Early post-treatment response assessment Post-treatment assessment is based on anatomical imaging, a focal residual mass of ≥1 cm and/or reduction in size (<50% primary site and < 80-90% nodal site) indicating high likelihood of residual cancer. However, anatomical criteria have low PPVs which are problematic, especially in the early post-treatment assessment of large necrotic HPV-related nodes which take time to resolve. Multiparametric imaging of head and neck cancer has a role in early post-treatment response assessment, and is close to being used for early intra-treatment response assessment. Advances in response assessment will accelerate as new functional imaging techniques emerge and more features can be extracted and analysed from existing large data sets using computer assisted interpretation (radiomics). There is compelling evidence that recent advances in computer science will play a vital role in future medicine assisting diagnosis and treatment (Rahmesh AN 2004). Decision support systems are already in routine use in hospitals for selecting most appropriate imaging methods and imaging protocols according to the latest international guidelines. Computer science has important potential to facilitate and analyse US, CT and MRI images in oncology e.g. automated cancer detection, segmentation and characterization as well as quantitative therapy response assessment. Various artificial intelligence (AI) techniques including e.g. neuronal networks, fuzzy logic, evolutionary computation, deep learning or computer vision are currently emerging. Radiomics refers to sophisticated computer algorithms that extract numerous quantitative imaging features (mostly from ROIs preselected by radiologists). Various applications are currently emerging including e.g. automated detection and characterization of lung, breast and prostate cancer. Novel healthcare informatics and communication technology open amazing opportunities for improving cancer imaging. With the increasing demand for cancer imaging, ever greater demand for expert knowledge in oncologic radiology is developing. But oncologic imaging specialists are usually located only in larger hospitals and dedicated care cancer centers, which, in turn, are located only in larger metropolitan areas. As a consequence, staffing shortage in rural, lowresource and less-educated areas will be inevitable. For a hospital administration it may even be easier to purchase high-performance imaging devices than to recruit imaging specialists. Accordingly, actual imaging opportunities may lie dormant. Support of imaging decisions for best possible utilization of local radiologic capabilities is feasible using the novel communication technology (DiPiro PJ 2017). Oncologic radiologists are in particular familiar with the situation that imaging studies acquired at other institutions are transmitted to their centers for second-opinion reporting and follow-up assessment. An extended radiology consulting service can accordingly include recommendation and implementation of state-of-the-art imaging protocols, quality control, image interpretation and documentation of the imaging studies according to most recent guidelines. Even clinical decision support by participation in interdisciplinary tumor boards sessions is on principle possible from a technological point of view. The growth of the internet with broad band capabilities has made teleradiology already a daily experience for many radiologists. National telemedicine programmes are currently under development in order to share expertise for improved cross-border medical services (Saliba V 2012). For the time being, however, major obstacles still hinder this progress including (1) legal aspects including data privacy, (2) financing, (3) local infrastructure and resources and (4) cultural factors (Saliba V 2012). Cross-border telemedicine combined with AI may have significant impact on improved care of cancer patients, particularly in rural areas and developing countries where expert knowledge is not available.
Exploiting the opportunities of AI, consulting oncologic radiologists may act as remote "information specialists" providing access to a specialist service (Jha S 2016). Current research on AI is emerging considerably, and further advancements in healthcare informatics and the built-up of supra-regional networks will help to overcome barriers. Novel IT technologies open the window to improving cancer imaging and making a high-quality imaging service available for more patients. Imaging in cancer has become an indispensable tool for treatment decision making, its success has been driven to a large extent by technical innovation in the past decades, which allows us to pinpoint small volume disease, characterise cancers and their biology, assess the effectiveness of therapies and provide novel prognostic information. However, the pace of change poses a challenge to our discipline, as new technologies are often released before recent introductions can be thoroughly tested or validated. Nonetheless, promising developments have often filtered quickly into clinical practice. Cancer imaging is at the cusp of another major revolution, which will be driven by artificial intelligence/ machine learning and big data.
Radiologists should prepare to embrace these developments, by active engagement with stakeholders to define the role of machines in radiological workflows, so that machines can work synergistically with and support humans to be better and more effective radiologists. There is an immense opportunity for radiologists to become imaging physicians, who are more active in directing patient care and management alongside our clinical colleagues. The coming era of growing reliance on machines and potential information overload from multi-modal and integrated diagnostics can lead to a highly de-personalised radiological profession when paradoxically providing increasingly personalised medical data. In such an environment, we run the risk of becoming mere providers of information, rather than communicators of knowledge or wisdom. We should use artificial intelligence to free up time to enable synthesis of new knowledge from the information we gather, which is subsequently distilled into wisdom for clinical practice. Radiologists have become increasingly isolated from direct patient contact, and there is chance to re-orient this relationship, which will help foster greater understanding of our profession and better patient-centred engagement. To maintain the relevance of our profession, it is important for radiologists to re-think their role in the new healthcare enterprise. Our decisions today will shape the practice of tomorrow. In the past few years, great improvements have been made to achieve tumour control of primary and secondary liver malignancies.
To date, hepatic resection is considered the most effective treatment option for metastases, CCC, and early stages of HCC. However, tumour ablation has gained considerable ground as an alternative to surgery. And transarterial chemoembolization (TACE) is the standard of care as treatment for HCC in intermediate stage, but it may also have potential for treatment of liver metastases. Highly efficient combinations of chemotherapeutic and antiangiogenic may lead to complete disappearance of metastases (histologic complete remission) or at least they will prolong life in palliative situations. Thus, the role of imaging not only to plan treatment, but also for assessment of therapy response and follow-up is of utmost importance [1]. Contrast-enhanced multiphasic MDCT is widely used in the evaluation of HCC after local tumour treatment [2]. Necrotic lesions appear in general as low-density areas with a thin rim of contrast enhancement in the arterial and portal venous phase, which may represent granulation tissue. Another frequent finding is the presence of perilesional parenchymal enhancement, which likely represents a benign physiologic hemodynamic response [3]. In contrast, residual tumour after incomplete tumour ablation or TACE may display focal or nodular peripheral enhancement in the arterial or portal venous phase. At MRI, DWI has been found be helpful to detect early tumour recurrence, together with dynamic gadolinium-enhanced imaging (with subtraction). In the hepato-biliary phase of liver-specific MR contrast agents, tumour necrosis as well as viable tumour may display low signal intensity (due to lack of hepatocellular uptake). A variety of extrahepatic primary malignancies spread to the liver via the arterial or portal venous blood supply to seed metastases. Colorectal cancer patients with a limited number of liver-only metastases clearly benefit from surgical resection. Tumour ablation (either by radiofrequency, microwave, electroporation or laser) has shown great promise as an alternative for patients, who refuse surgery or in whom surgery resection is contra-indicated for medical reasons. As a joint procedure in combination with surgery it helps to achieve parenchyma-sparing local tumour control in patients with deep lesions. Side-by-side comparison of the pre-and post treatment CT scan (with the knowledge of the procedure performed) may help in the confident diagnosis of residual tumour post treatment: it shows if the size of necrosis in the post treatment scan actually covers completely the tumour seen on the pre-contrast scan. At MRI, necrosis post tumour ablation may show low signal intensity on T2-weighted images and high signal intensity on T1-weighted images due to coagulation necrosis. Dynamic T1-weighted GRE pulse sequences are recommended to show perfusion of vital tumour, which helps to assess therapy response. However, a recent study demonstrated that a single T1w GRE pulse sequence with fat-suppression is very accurate in predicting treatment efficacy after tumour ablation [4]. Volumetric portal venous phase enhancement as a sign of response has been shown to be associated with survival [5]. Gadolinium enhancement is also key to detect tumour recurrence. DWI has surpassed T2w fatsuppressed TSE imaging for detection of local tumour recurrence and new intrahepatic metastases.
Recently dynamic contrast-enhanced perfusion MRI assessment of tumour response to anti-angiogenesis drugs was described. These results suggest that dynamic contrast-enhanced MRI may serve as a precise biomarker to monitor pharmacological response to angiogenesis inhibitors than measurements of tumour diameter. In summary, contrast-enhanced multi-phasic MDCT and MRI are the best tools to assess tumour response to therapy. Multiparametric MRI with diffusion-weighted imaging and dynamic gadolinium-enhanced sequences (and perfusion assessment) is very sensitive to assess local therapy response. In patients with treated liver metastases, MDCT has the advantage to monitor liver and extrahepatic disease. Hepatocellular carcinoma (HCC) is the second most common cause of cancer-related mortality worldwide. When recognized at early stages, HCC can be cured with appropriate therapy, which may involve resection, percutaneous ablation, transarterial chemoembolization or orthotopic liver transplantation (OLT). Since 2001, multiple expert panels in different parts of the world have advocated the noninvasive diagnosis of HCC if certain imaging criteria are met. Therefore, accuracy of classifying lesions as HCC or non-HCC on imaging studies is critical to patient management. All of the consensus guidelines for the diagnosis of HCC prior to LI-RADS have important shortcomings, which supports the need for LI-RADS. None of the guidelines prior to LI-RADS defines or illustrates the pertinent imaging features on which the noninvasive diagnosis of HCC is based. LI-RADS not only establishes standardized imaging feature terminology, it precisely defines the imaging features and provides illustrations for clarification. Secondly, none of the guidelines prior to LI-RADS has tried to clarify the probability that a lesion not fulfilling the criteria for definite HCC might be HCC. LI-RADS addresses this shortcoming by providing a comprehensive system to categorize each liver observation based on the likelihood that it is HCC, a non-HCC malignancy or a benign abnormality. A number of recent studies have provided data validating the appropriateness of the LI-RADS categories. LI-RADS is of particular importance in the United States and Western Europe where OLT is considered the treatment of choice for patients with cirrhosis and HCC. Currently, approximately 25% of liver transplants performed in the United States are on patients with HCC, because patients with cirrhosis who develop HCC are given additional priority on the liver transplant waiting list independent of their liver function. Thus, the conservative approach of LI-RADS to the noninvasive diagnosis of HCC, which favors specificity over sensitivity, makes sense in the United States. The rationale is that if a patient with an imaging diagnosis of HCC can potentially receive higher priority on the transplant waiting list than a patient with poorer liver function but no HCC, then we need to be as certain as possible that the diagnosis is accurate. Such considerations are similar in Western Europe where 12% of liver transplants are performed on patients with HCC. An additional advantage of LI-RADS over other consensus guidelines for the diagnosis of HCC is that by standardizing terminology, interpretation, reporting and imaging management LI-RADS improves communication among radiologists, hepatologists and surgeons, thus enhancing the opportunity to improve patient outcomes.
Poster Abstracts Learning objectives: -To highlight radiological features of common paediatric extracranial neoplasms correlated with demographic, clinical and pathological information; -To illustrate demographically atypical presentations.

Content organisation:
Wilms' tumour is the commonest renal tumour in children, characterised by large mass lesions, often with the characteristic 'claw sign' , within the renal parenchyma. Peak incidence is 3-4 years, but the tumour occasionally occurs in adults. The demographics of renal tumours are discussed. Rhabdomyosarcoma is pathologically diverse with a variety of cell types, and can occur at any site, although head & neck and pelvic tumours account for the majority of cases. It is anatomically heterogeneous and may occur in young adults. In paediatrics, hepatoblastoma is the most frequent liver tumour. However, as age increases, a wider differential of liver tumours emerges which will be illustrated. Typically "adult tumours" may occur in children, for example nasopharyngeal carcinoma and other head & neck tumours, adrenal carcinoma and a variety of sarcomas. Examples will be presented.

Conclusions:
Correlation with demographic information is a fundamental of paediatric tumour diagnosis. However it should be recognised that typically adult tumours may occasionally present in children, and vice versa.

P2
Prevalent diagnostic discrepancy patterns found in chest x-ray error cases SJJ Touyz, M MacKenzine, K Kim, P Janousek, G Grazvydas Pennine Acute Hospitals NHS Trust, Manchester, UK Correspondence: SJJ Touyz (sarah.touyz@doctors.org.uk) Cancer Imaging 2018, 18(Suppl 1):P2 Background: Diagnostic radiological errors are a recognised issue, with~30% error rates in chest x-rays(CXR) often leading to medicolegal problems. Regular discrepancy meetings analyse errors and are encouraged for quality improvement by local trusts and the Royal College of Radiologists. Reviewing error cases improves diagnostic outcomes through education.

Aims:
This study retrospectively reviews CXR error cases referred to local discrepancy meetings at a general district hospital in an acute multisite trust in the UK. The aims are to: (i) iidentify recurrent patterns of errors for quality improvement, (ii) highlight patterns/areas needing further analysis during reporting, (iii) and scrutinise current practice of discrepancy meetings.

Methods:
Retrospective analysis of 51 CXR error cases from discrepancy meetings was done from 2009-2017 using PACS/CRIS. A database with various parameters was made for statistical analysis.

Results:
The average patient age was 68.5, with 56% males and 44% females. 76% had a history of smoking.>60% of discrepancy errors were perceptual, 29% cognitive and 8% communicative. Most errors were primary malignancies(>55%) and infections(14%). Missed lesions were in the peripheral(41%), hilar(30%) and parahilar(24%) regions. 40% of cases were in the upper zones. The most common referred symptom was cough, and most referrals came from A&E. Conclusion: Discrepancy meetings educate reporters and hope to improve diagnostic accuracy (particularly for malignancies). Perceptual errors were the most common error made in CXR reporting. Peripheral and upper zones in smoking patients referred with cough need extra scrutiny when reporting. Improvements to current discrepancy meeting methodology is encouraged to optimise outcomes and effectiveness of the meetings and ensure quality improvement for future patient outcomes.

Content organisation:
Osteochondromas are one of the commonest bone tumours, and are most often found at the ends of long bones. Often solitary, the development of multiple osteochondromas is characteristic of diaphyseal aclasis (or hereditary multiple exostosis) in which the risk of malignant transformation to chondrosarcoma is increased. Chondrosarcoma varies in histology and prognosis; mesenchymal and dedifferentiated tumours have the worst prognosis with high rates of local recurrence and metastasis. Multimodality imaging has a useful and complementary role in enabling lesion characterisation, detection of lesion extent and malignant change, surgical planning and staging. We will discuss imaging characteristics of osteochondromas and chondrosarcomas and their complications as well as pseudotumour that mimics chondrosarcomas.

Conclusions:
Although plain radiograph is critical in the initial assessment of these tumours which are often characteristic, additional features on CT and MR images are invaluable adjunct tools for optimising tumour characterisation. Understanding and recognising the spectrum of appearances of these tumours allow improved patient assessment and are vital for optimal clinical management, including diagnosis, biopsy, staging, treatment, and expectations for prognosis.  We propose diagnostic tools for evaluation of musculoskeletal tumours and pseudotumours. We included in the differential diagnosis lymphangioma, muscle hematoma, abscess, osteomyelitis.

Conclusion:
There are several diseases that can mimic musculoskeletal tumours. It is necessary to take them into account and know how to use diagnostic keys for an accurate diagnosis. Learning Objectives: Describe the main differential diagnosis of non-adenomatous adrenal masses. Identify the main imaging findings of these entities.

Content Organisation:
Imaging studies of the adrenal glands are usually necessary in three clinical circumstances: patients who may have received a clinical diagnosis of adrenal hormone hyperfunction, patients who have undergone imaging studies of the suprarenal glands to confirm the suspicion of a metastatic disease, and finally, patients who underwent imaging studies for other indications and were accidentally detected a lesion on these glands. These lesions -in symptomatic and asymptomatic patients -can be classified according to several criteria, namely: size, site of anatomical origin, morphological characteristics or biological behavior. The differential diagnoses of non-adenomatous adrenal masses that we describe are: Simple cysts Pheocromocytoma Adrenal hiperplasia Adrenal hemorrhage Neoplasms CT imaging is the preferred diagnostic imaging technique used for the detection and characterisation of adrenal lesions with reported specificity of identifying the mass as an adenoma being close to 100%, though the sensitivity is lower. Positron emission tomography is generally performed in patients with cancer and it yields information about the biochemical processes that may precede gross anatomic changes.

Conclusion:
The adrenal gland is involved in a variety of benign, malignant primary and metastatic tumours. The use of a correct study algorithm, with the appropriate and accessible diagnostic tools, allows the correct characterization of the injuries. The role of the radiologist is of great importance in the identification and diagnosis of tumours and tumour-like conditions in symptomatic and asymptomatic patients. Learning Objectives: To review the utility of high resolution MRI in identifying the correct surgical approach based on the relation of the tumour with mesorectal and extralevator planes. Content Organisation: Low rectal cancer ( < 5 cm from the anal verge) remains a global challenge due to high positive pathological circumferential resection margin rates principally attributed to anatomic considerations as the fact that the mesorectal fascia tapers downward at this level and higher perforation rates. Preoperative anatomically based High Resolution MRI systems may be able to predict which patients may be at risk from traditional surgery and enable patients to undergo a more radical excision, thus ensuring adequate resection margins. We will review: -MRI low rectal cancer plane staging.

Conclusion:
Anatomically based High Resolution MRI staging for low rectal cancer is an excellent tool to reduced margin positive and to improve outcome in patients with low rectal cancer. Group recommendations, whole-body MRI replaced skeletal surveys in myeloma assessment, because it is a highly sensitive examination with wide coverage that can both quantify disease extent and qualify treatment response without exposure to ionising radiation. New applications are subsequently being identified, e.g. assessment of metastatic prostate cancer and lymphoma staging. The high sensitivity and wide coverage associated with whole-body MRI however, comes at the cost of multiple incidental findings. We present a pictorial review of incidental findings encountered at our tertiary oncology centre in 2017, illustrating the spectrum of disease that can be encountered.

Conclusions:
-Incidental findings are commonly encountered with wholebody MRI due to its high sensitivity (19% of studies). -As radiologists, we need to be aware of the prevalence and spectrum of possible incidental findings that can be encountered while reporting these studies and know how to manage them so that potentially life-threatening diagnoses are not missed.

P11
Advanced Learning objectives: The main purpose of this exhibit is to correlate pancreatic adenocarcinoma CT imaging using TNM staging system to histopathological. Perform the diagnosis and a correct staging of the pancreatic tumour, establishing criteria of resectability, and assisting the surgeon in the preoperative planning before the finding of relevant anatomical variants.

Content organisation:
Decisions about diagnostic management and resectability should involve multidisciplinary assessment with appropriate highquality imaging studies to evaluate the extent of disease. Carefully timed scan acquisition allows accurate local and distant staging as well as assessment of local resectability. In addition, we evaluate vascular maximum intensity projection (MIP) images and curved MPR images obtained along the planes of the great vessels.
We review Multidetector volumetric computed tomography (MDCT) scans of 23 adult patients with known or suspected pancreatic adenocarcinoma performed between January 2009 and April 2018. Findings were then compared with surgical and pathologic results.
Our preliminary results showed difficulties to asses between similar tumour sizes, especially between T2 and T3 stages. However, we were successful in assessing T4 and M1 (86% average accuracy) and therefore determining unresectability of tumours.

Conclusion:
Based on our analysis, we conclude that it is difficult to perform a correct anatomo-radiological correlation, where the greatest discrepancy lies in objectifying the actual size of the tumour and its extension. Nevertheless, promising results are obtained when evaluating resectability of the tumour since it has direct impact on the surgical/ clinical management decisions. TNM staging for pancreatic adenocarcinoma is not frequently included in radiology reporting but is a useful tool for the preoperative assessment of pancreatic adenocarcinoma. The imaging of neuroendocrine tumours has evolved considerably over the past 40 years. Discovered only within the last 45 years, Somatostatin (Ss) is a neuropeptide which exerts its effects on multiple cell types, all over the body. Approximately 10 years after the initial discovery of Ss, research into somatostatin receptor (SSTR) subtypes and the cellular pathways it affects really began to progress. There are five Ss receptor subtypes with different binding affinities for the neuropeptide; the most widely expressed receptors and therefore most useful for imaging are SSTR 2 and 5. The creation of a Ss analogue and the ability to label it with an appropriate radionuclide, has been influential in discovering what we know about SSTR expression in both normal tissue and in pathological entities. This includes neuroendocrine tumours and has helped to evaluate how such tumours behave. Traditional diagnostic imaging of neuroendocrine tumours using computed tomography (CT), has limitations in the sensitivity of tumour identification. Functional imaging using radiopharmaceuticals increases sensitivity significantly. Traditionally this has used gamma emitting radionuclides, namely Indium-111. Increasingly, positron emitting tomography (PET) imaging, using pharmaceuticals labelled with Gallium-68 is being performed due to the superior sensitivity in tumour detection. Furthermore, peptide receptor radionuclide therapy (PRRT) is a growing area of interest amongst both the nuclear medicine and oncology communities. This educational poster seeks to provide the reader with a background into the history of Ss imaging and the range of pathologies that may be imaged. It will serve as a balanced overview of the imaging modalities available to date. We shall focus on functional imaging and discuss the advantages and limitations of using traditional gamma radionuclides as well as the increasingly available PET radiopharmaceuticals. We discuss the growing field of PRRT and consider what the future holds for both the imaging and treatment of neuroendocrine tumours. Learning objectives:

P15
To review the spectrum of metastatic disease to the heart and pericardium in correlation with types of spread, imaging characteristics, anatomic predilection, and treatment options.

Content organisation:
Metastases are the most common malignant cardiac tumours and by far more common than primary cardiac tumours. They can spread by direct invasion from adjacent malignancy, such as lung or mediastinal cancer, by lymphatic, or hematogeneous dissemination. We will demonstrate each specific way of vascular, direct, and lymphatic metastatic spread to the heart with anatomic predilection of particular type of metastasis, using case by case approach. We will review most common primary malignancies, known to metastasise to the heart and pericardium, including: uterine leiomyosarcoma, lung cancer, thymic cancers, carcinoid, melanoma, breast cancer, and lymphoma. We will present diagnostic modalities and imaging characteristics of cardiac metastasis: -Diagnostic modalities (Echocardiography, Computed Tomography, Cardiac MRI, PET-CT, and Coronary Angiography) -Behavior and imaging characteristics of tumours, which are usually similar to the primary malignancy -Patterns of Gadolinium enhancement We will discuss surgical, radiation, and chemotherapy treatment options of cardiac and pericardial metastatic disease.

Conclusions:
Metastatic disease to the heart and pericardium remains under recognised despite accumulating evidence in literature and diagnostic capabilities of modern medicine. Cardiac metastasis is a devastating disease with poor prognosis, which almost always upstages the primary malignancy and significantly limits treatment options. Aim:

Sleep abnormalities in non-CNS malignant tumours
The aim of this study is to present our experience between the correlation of 18 fluorocholine (FCH) positron emission tomography (PET)/computed tomography (CT) findings and PSA levels in detecting local recurrence, lymphatic/haematogenous involvement in patients with prostate carcinoma. Methods: 1421 patients with prostate cancer were enrolled in our institution between July 2012 and May 2018. Patients were separated in two main groups: biochemical progression 1230 (86%) and 191 (14%) staging. Each group was divided in three subgroups according to PSA levels: ≤1 ng/ml; ≥ 1-5ng/ml and >5 ng/ml. All of them underwent "Dual phase" PET-CT consisting of initial pelvis starting 20 minutes after the injection of 18 F-Choline followed by 1 hour delayed PET of the whole body. Results: 2321 lesions showed increased uptake on FCH-PET and were interpreted as local recurrence or metastases. According to biochemical progression we identified that the majority of patients have PSA levels of >5 ng/ml (713 patients 58%) and the most frequent site of compromised was pelvic lymph nodes(34%), local recurrence(27%) and extrapelvic lymph nodes(23%). On the other hand, for staging patients we have found that most of them have PSA levels of ≤1 ng/ ml (105 patients 55%) and we have detected that they most commonly have local compromise (37%), pelvic lymph nodes (32%) and extrapelvic lymph nodes(15%).

Conclusions:
In our experience 18 FCH PET/CT is a useful tool to detect lesions in patients with biochemical progression but it seems to have limited value for staging. The pattern of lymphadenopathy was assessed using T1W, T2W, DWI and DCE MRI sequences. Size and morphological characteristics were defined as the key descriptors in predicting nodal involvement of the disease. Involved lymph node groups were also described. Data was analysed using SPSS v.20 statistical software and subjected to a Fisher's exact test for significance.

Results:
Patients with HIV demonstrated significantly more nodal disease than non-HIV patients at a p-value 0.0028. The number of nodal groups involved was also higher in the group of patients who had HIV infection.

Conclusion:
The results of this small study indicate that more nodal disease is likely to be encountered in the HIV patient than in the non-HIV patient during MRI assessment of cervical cancer. More studies will be required to establish the factors behind our observations.

P23
Evaluation of multiparametric MRI (Mpmri) at 3T in prostate cancer in a private outpatient setting M Pixberg 1 , J Ghafur 1 , D Hercher 2 , J LFS is a rare autosomal dominant heritable syndrome related to a germline mutation of the tumour suppressor gene TP53 that predispose carrying patients to a broad spectrum of cancers, including breast cancer, sarcoma of soft tissue and bone, brain tumours and adrenocortical carcinoma. Breast cancer are the more frequent tumours associated with LFS. Nonetheless, less is known about the distribution of histopathogical types of breast lesions in LFS. We will present the review of LFS cases with breast disease identified from the database of our institution (Institut Curie, Paris, France), including mainly breast carcinoma, phyllodes tumour, ductal in situ carcinoma and breast sarcoma. We will emphasise the radio-histological correlations of the various breast lesion types and present their different management.
We will discuss the current guidelines for imaging screening and surveillance of LFS patients.

Conclusions:
Breast tumours are the most frequent tumours associated with LFS with around 30% of patients diagnosed with cancer aged < 30 years. Breast lesions associated with LFS are not limited to invasive carcinoma. It is important for radiologists to be familiar with the wide spectrum of breast tumours in LFS, since LFS patients are exposed to a broad range of cancers occurring at a particularly young age making the screening a major clinical challenge.

P31
Role of computed tomography in predicting lesser sac/omental disease in patients with advanced ovarian carcinoma S Mukhopadhyay, S Sen, A Chandra, D Lingegowda, A Chatterjee, P Ghosh, A Mukhopadhyay, R Shrestha Department of Radiology and Gynaecological Oncology, Tata Medical Center, Kolkata, India Correspondence: S Mukhopadhyay (sumitmukhopadhyayrad@gmail.com) Cancer Imaging 2018, 18(Suppl 1):P31

Aims:
Lesser sac (LS) involvement is reported in 60-70% of patients undergoing cytoreductive surgery (PDS or IDS) for advanced (FIGO III/IV) epithelial ovarian cancer (AEOC) . We aimed to study whether computed tomography (CT) scan can accurately predict the lesser sac/ omental disease pre-operatively.
(n=11, 16.9%), and PIK3CA (n=13, 20%). There was no significant correlation between qualitative imaging features and genetic mutations after adjusting for multiple comparisons (p>0.05). The mutations in frequently occurring APC and TP53 were most prevalent in the clusters C1 and least prevalent in cluster C3. Similarly, KRAS mutation was most frequent in the C1. In general, mutations were most frequent in cluster C1 when compared to mutations in clusters C3 and C4.

Conclusion:
This pilot study showed no significant correlation between qualitative MRI imaging features and genetic mutations after adjustments for multiple comparisons. However, promising correlations were observed among quantitative features and genetic mutations. Further studies with larger sample size are warranted to further evaluate and validate our preliminary data.

O4
Diagnostic accuracy of magnetic resonance (MRI) to predict axillary lymph node metastases in patients with breast cancer A Jaipal, G J Bansal (gjbansal@gmail.com) The Breast Centre, UHL, Cardiff and vale UHB, Cardiff, UK Cancer Imaging 2018, 18(Suppl 1):O4 Aim: The aim of this study was to evaluate the diagnostic accuracy of preoperative Breast MRI for axillary lymph node (LN) metastases in two groups of patients with breast cancer. Those who received Neoadjuvant chemotherapy (NAC) prior to surgery were compared to those which proceeded straight to surgery (non NAC).

Materials and Methods:
Between 2011-2018, 198 patients with primary breast cancer were divided in two groups (NAC and non-NAC). The preoperative MRI results of axilla were compared to postoperative results in both groups. The axillary LN status (number of abnormal LNs, long axisshort axis ratio, loss of fatty hila and cortical thickness) was recorded. SPSS (ver 21) was used for data analysis and two-sided p<0.05 was taken as statistically significant.

Results:
There were 112 patients in the non-NAC group and 86 patients in NAC group. Sensitivity of MRI to detect axillary metastases was 80.95% and specificity was 94.51% in the non -NAC group versus 67.74% and 81.13% in the NAC group (p= 0.001). There was greater correlation between the number of abnormal nodes on MRI and postoperative results in the non-NAC group (p<0.001).

Conclusions:
MRI was less sensitive and specific for the assessment of axillary LNs in patients who received NAC prior to surgery compared to those who do not receive NAC. Low sensitivity of MRI in patients who receive NAC underscore the importance of post NAC biopsy of nodes that turn normal on MRI after NAC.

Conclusions:
Hospitalised oncology patients are at higher risk of acute renal AEs following contrast-enhanced CT compared to those without cancer. Understanding the complex bidirectional relationship between cancer and kidney function and application of mitigation strategies is an important consideration for oncology patients.