- Case series
- Open Access
Improving the diagnosis of radiation necrosis after stereotactic radiosurgery to intracranial metastases with conventional MRI features: a case series
Cancer Imaging volume 22, Article number: 33 (2022)
The distinction between true disease progression and radiation necrosis after stereotactic radiosurgery to intracranial metastases is a common, but challenging, clinical scenario. Improvements in systemic therapies are increasing the importance of this distinction. A variety of imaging techniques have been investigated, but the value of any individual technique is limited.
Assessment should extend beyond simply the appearances of the lesion at a given timepoint, but also consider local anatomy and lesion evolution. Firstly, enlargement of a metastasis is affected by local anatomical boundaries, such as the dural reflections or cerebrospinal fluid spaces. In contrast, the radiation dose administered with stereotactic radiosurgery does not respect these anatomical boundaries and is largely concentric around the treated lesion. Therefore, new, non-contiguous enhancement across such a boundary can be confidently attributed to radiation necrosis. Secondly, the dynamic nature of radiation necrosis may result in a change in lesion shape, with different portions of the lesion simultaneously enlarging and regressing. Regression of part of a lesion indicates radiation necrosis, even if the overall lesion enlarges. This case series describes these two features and provides illustrative clinical examples in which these features allowed a confident diagnosis of radiation necrosis.
The distinction between true disease progression and radiation necrosis should extend beyond just the appearances of the lesion. More nuanced interpretation incorporating a relationship to anatomical boundaries and a change in shape can improve accurate diagnosis of radiation necrosis.
Intracranial metastases (IM) are a frequent challenge in managing patients with metastatic disease. While IM can respond to systemic therapies , intracranial progression despite extracranial response remains a common clinical scenario. Radiotherapy (RT), in particular stereotactic radiosurgery (SRS), provides an effective means of treating IM, with high rates of local control . The main limitation of SRS, however, is the risk of radiation necrosis (RN) affecting adjacent brain tissue, which occurs in 5–26% of lesions treated, most commonly 6–11 months after SRS . In addition to managing symptoms associated with RN, it is important to distinguish RN from true tumor progression (progressive disease, PD), as the management of these two entities differs substantially. It is expected that RN will regress spontaneously over time, though medical treatments (for example, corticosteroids or bevacizumab) or even surgical resection may be warranted for control of symptoms. Therefore, additional treatment due to an incorrect diagnosis of PD (for example, surgical resection, repeat SRS or a change in medical therapy) can lead to significant unnecessary morbidity. Similarly, a delayed diagnosis of PD may lead to worse patient outcomes, for example a need for more extensive surgical resection.
Accurate diagnosis of RN is notoriously challenging, however, due to an overlap in the conventional MRI (Magnetic Resonance Imaging) appearances with PD . This has led to the investigation of a variety of advanced MRI sequences, including Diffusion-Weighted Imaging, Perfusion and Spectroscopy . Zach et al. reported promising results with delayed-contrast MRI , though subsequent validation and clinical implementation have been limited at this stage. PET (positron emission tomography) using amino acid tracers such as FET (fluorine-18-fluoroethyl-L-tyrosine) may also be used [6, 7]. Nevertheless, each technique has its limitations, prompting more research into the augmentation of traditional imaging interpretation with artificial intelligence (AI) techniques . There has, however, been less work in this area than in some other areas of neuro-oncology; for example, there are fewer published papers on AI techniques for diagnosing post-treatment effects in IM than gliomas , despite IM being more common.
These techniques have typically focused on the imaging characteristics of the lesion in question, but the assessment should extend beyond simply the appearances of the lesion at a given timepoint. For example, RN preferentially involves the white matter rather than gray matter , and the risk is particular high in the corpus callosum . Two additional features, which we have observed to be specific for RN, relate to anatomical boundaries and a change in the shape of the lesion. This case series describes these two features, and provides illustrative examples from the authors’ own clinical practice in which these features allowed a more confident diagnosis of RN.
These cases were identified through the first author’s routine clinical practice at a tertiary level oncologic hospital. Institutional ethics committee approval was received for a waiver of patient consent. SRS was delivered either using a Varian TrueBeam linear accelerator with dynamic conformal arc therapy (DCAT) or with the Gamma Knife radiosurgery machine. The SRS schedules used in each case are specified in the respective figure legends.
IM generally enlarge in one of two ways. Most commonly, IM enlarge in a relatively concentric manner, with progressive enlargement of a nodular lesion. Less commonly, metastases may extend along an anatomical surface, such as along the cortex or pial surface of the brain, or along the ependymal surface of the ventricular system. The frequency with which metastases enlarge in these two ways depends, in part, on the histology. For example, intracranial metastases from melanoma have a particular predilection to developing at the interface between the pia mater and the cortex , and curvilinear extension along the pial surface is a common occurrence. As such, the pattern of IM progression or recurrence is affected by the local anatomy. In contrast, the distribution of the radiation dose administered with SRS in general does not respect these anatomical boundaries, but rather is largely concentric around the treated lesion. This is illustrated by the isodose distributions as depicted in Figs. 3 and 6.
Several such anatomical boundaries exist in the brain. Firstly, this includes the major dural reflections: the falx cerebri (Figs. 1 and 2) and the tentorium cerebelli (Fig. 3). In our experience, the development of new enhancement on the other side of a dural reflection after SRS is strongly suggestive of RN rather than PD. A sulcus and CSF may provide a similar barrier (Figs. 4 and 5). In our experience of both above scenarios, the new enhancement occurring with RN preferentially involves the subcortical white matter, sparing the intervening cortex, and thus the new lesion can be anatomically distinct from, or non-contiguous with, the originally treated lesion. It is important, however, to ensure that there is no contiguous involvement of the intervening meninges, which would otherwise raise concern for PD.
Similar considerations exist for IM located close to the ventricular system (Figs. 6 and 7), though the MRI appearances differ to a degree, related to the anatomical differences compared to the more peripheral lesions discussed above. In contrast to more peripheral lesions, there is often no intervening grey matter between the treated IM and the white matter on the other side of the ventricle. As a result, in our experience, such RN lesions tend to develop in continuity, wrapping around the ventricle. Of note, and in contrast to metastatic disease, there is often no enhancement along the ventricle surface, producing an “open ring” appearance. This open ring appearance can also occur with treated metastases located away from the ventricular surface (see Fig. 1), related to the lower differential radiosensitivity of the cortex and the effect this in turn has on the MRI appearances (the ring potentially being open on the side of the cortex). Notably, in all our illustrated cases, the area of enhancement correlates geographically with the high radiation dose regions, as highlighted in Figs. 3 and 6.
Change in shape
The progression and evolution of IM tends to occur in a relatively simple manner. Prior to treatment, or if not responding to treatment, the entirety of an IM will enlarge, without regression of any particular components. While subsequent recurrence (and hence enlargement) may occur within only a portion of the initial lesion, often at its margin, any post-treatment regression will have already occurred. Thus, it would be very uncommon for different components of the same IM to enlarge and regress simultaneously. In contrast, RN is a more variable and dynamic process, affected by a variety of factors inherent to that particular part of the brain (or voxel), such as the dose received and the sensitivity to radiation effects. These factors vary from voxel to voxel, often leaving to a complex evolution of RN lesions. Radiologically, this can manifest as a change in the shape of the lesion, with different portions of the given lesions simultaneously enlarging and regressing (Figs. 7, and 8). Given that such evolution would not be expected for recurrent IM, such an appearance can be more confidently attributed to RN.
Given the challenges in accurately distinguishing between RN and PD, radiologists need to use all the tools at their disposal. A more nuanced approach beyond simply assessing the appearances of the lesion can provide greater diagnostic confidence. Anatomical boundaries, such as dural reflections and CSF, provide a barrier to enlargement of a metastasis, but not to RN. Thus, new or progressive enhancement on the other side of such a boundary suggests a diagnosis of RN. Another important tool at the radiologist’s disposal is follow-up imaging. The change in the shape of a lesion is a potential discriminating feature, and may manifest before the overall lesion has started to regress, or even despite overall enlargement of the lesion. Given the moderate incidence of RN after SRS in contrast to the low rate of local recurrence, a new or enlarging lesion at the site of previous SRS is statistically more likely to represent RN rather than PD. As a result, in our clinical practice we are hesitant to give a confident diagnosis of PD upon the initial identification of a new or enlarging lesion at the site of previous SRS. Instead, we have a low threshold for recommending short-interval follow-up imaging (supplemented by advanced sequences), at which time the features described may become evident.
While recent research into improving the diagnosis of RN has focused on augmenting the assessment with AI techniques, the subtle features described here are likely to be less amenable to accurate identification and characterisation by AI techniques. Going forward, it is important that the information assessed by AI-based techniques is not confined to the static imaging appearances alone, but also incorporates data on lesion location and temporal evolution, such as the features described here. The value of this approach has been shown in the use of deep learning for the prediction of glioma molecular subtype, in which cropping to the lesion led to lower prediction capability, attributed to the loss of information on tumor location and the correlation this has with molecular subtype [13, 14]. This highlights the need for an understanding of the expected imaging appearances, including the relevant sequences, when building and validating a robust AI model.
Nevertheless, emerging AI-based tools will remain only one part of the overall armamentarium, and the input of the neuroradiologist remains critical. Indeed, a multidisciplinary approach is important in these complex cases, and a variety of additional factors warrant consideration. These include the RT dose (for example if the patient has received repeat SRS or sequential whole-brain RT and SRS, both of which will increase the risk of RN) and neurosurgical considerations (for example the additional morbidity if the lesion were to enlarge). In addition, there have been some suggestions that simultaneous immunotherapy increases the likelihood of RN , though a recent systematic review suggests that there is no increased risk . The time since SRS and the SRS technique used are also relevant, given suggestions that RN may occur earlier after Gamma Knife SRS than linear accelerator-based SRS .
We have encountered other similar cases in our clinical practice, but have been limited in the number which could be adequately presented, and thus specifically selected those which provide the most robust and broad demonstration of the concepts. Nevertheless, the selective nature of this case series is an inherent limitation. As such, we are unable to provide typical measures of a diagnostic test such as sensitivity and specificity. In our experience, these features are quite specific for RN, and this is supported by biological plausibility. However, our experiences also suggest that the sensitivity of these features is likely to be modest, which will provide some challenges to more formal evaluation. This is most intuitive for the anatomical boundaries feature, as only a minority of metastases are located in such a location. Nevertheless, it would be worthwhile including these features in future work assessing a multi-parametric approach to the diagnosis of RN, in particular determining the frequency of these features and confirming their suspected high predictive value.
The distinction between PD and RN after SRS for IM is an increasingly common, but challenging, clinical scenario. While a variety of imaging techniques have been investigated, the value of any individual technique is limited. Assessment should, however, extend beyond simply the appearances of the lesion at a given timepoint, but also consider the tumor location and evolution. In particular, lesion development or enlargement across anatomical boundaries and a change in shape over time are both features which can increase confidence for a diagnosis of RN.
Availability of data and materials
Progressive disease / true tumor progression
Magnetic resonance imaging
Positron emission tomography
Tawbi HA, Boutros C, Kok D, Robert C, McArthur G. New Era in the Management of Melanoma Brain Metastases. Am Soc Clin Oncol Educ Book. 2018;38:741–50.
Minniti G, Clarke E, Lanzetta G, Osti MF, Trasimeni G, Bozzao A, et al. Stereotactic radiosurgery for brain metastases: analysis of outcome and risk of brain radionecrosis. Radiat Oncol (London, England). 2011;6:48.
Lee D, Riestenberg RA, Haskell-Mendoza A, Bloch O. Brain Metastasis Recurrence Versus Radiation Necrosis: Evaluation and Treatment. Neurosurg Clin N Am. 2020;31(4):575–87.
Shah R, Vattoth S, Jacob R, Manzil FF, O’Malley JP, Borghei P, et al. Radiation necrosis in the brain: imaging features and differentiation from tumor recurrence. Radiographics. 2012;32(5):1343–59.
Zach L, Guez D, Last D, Daniels D, Grober Y, Nissim O, et al. Delayed contrast extravasation MRI: a new paradigm in neuro-oncology. Neuro Oncol. 2015;17(3):457–65.
Galldiks N, Stoffels G, Filss CP, Piroth MD, Sabel M, Ruge MI, et al. Role of O-(2-(18)F-fluoroethyl)-L-tyrosine PET for differentiation of local recurrent brain metastasis from radiation necrosis. J Nucl Med. 2012;53(9):1367–74.
Lasocki A, Hicks RJ. How we read: the combined use of MRI and novel PET tracers for the characterisation and treatment planning of masses in neuro-oncology. Cancer Imaging. 2019;19(1):57.
Kim HY, Cho SJ, Sunwoo L, Baik SH, Bae YJ, Choi BS, et al. Classification of true progression after radiotherapy of brain metastasis on MRI using artificial intelligence: a systematic review and meta-analysis. Neuro Oncol Adv. 2021;3(1):vda080.
Bhandari A, Marwah R, Smith J, Nguyen D, Bhatti A, Lim CP, et al. Machine learning imaging applications in the differentiation of true tumour progression from treatment-related effects in brain tumours: A systematic review and meta-analysis. J Med Imaging Radiat Oncol. 2022. https://doi.org/10.1111/1754-9485.13436. Online ahead of print.
Choi S, Hong A, Wang T, Lo S, Chen B, Silva I, et al. Risk of radiation necrosis after stereotactic radiosurgery for melanoma brain metastasis by anatomical location. Strahlenther Onkol. 2021;197(12):1104–12.
Mohammadi AM, Schroeder JL, Angelov L, Chao ST, Murphy ES, Yu JS, et al. Impact of the radiosurgery prescription dose on the local control of small (2 cm or smaller) brain metastases. J Neurosurg. 2017;126(3):735–43.
Lasocki A, Khoo C, Lau PKH, Kok DL, McArthur GA. High-resolution MRI demonstrates that more than 90% of small intracranial melanoma metastases develop in close relationship to the leptomeninges. Neuro Oncol. 2020;22(3):423–32.
Cluceru J, Interian Y, Phillips JJ, Molinaro AM, Luks TL, Alcaide-Leon P, et al. Improving the noninvasive classification of glioma genetic subtype with deep learning and diffusion-weighted imaging. Neuro Oncol. 2022;24(4):639–52.
Lasocki A, Anjari M, Ӧrs Kokurcan S, Thust SC. Conventional MRI features of adult diffuse glioma molecular subtypes: a systematic review. Neuroradiology. 2021;63(3):353–62.
Colaco RJ, Martin P, Kluger HM, Yu JB, Chiang VL. Does immunotherapy increase the rate of radiation necrosis after radiosurgical treatment of brain metastases? J Neurosurg. 2016;125(1):17–23.
Gagliardi F, De Domenico P, Snider S, Roncelli F, Pompeo E, Barzaghi LR, et al. Role of stereotactic radiosurgery for the treatment of brain metastasis in the era of immunotherapy: A systematic review on current evidences and predicting factors. Crit Rev Oncol Hematol. 2021;165: 103431.
Navarria P, Clerici E, Carta G, Attuati L, Picozzi P, Franzese C, et al. Randomized Phase III Trial Comparing Gamma Knife and Linac Based (EDGE) Approaches for Brain Metastases Radiosurgery: Results from the Gadget Trial. Int J Radiat Oncol Biol Phys. 2018;102(3):S143–4.
Dr Arian Lasocki was supported by a Peter MacCallum Cancer Foundation Discovery Partner Fellowship.
Ethics approval and consent to participate
Ethics committee approval was received from the Peter MacCallum Cancer Centre for a waiver of patient consent (project ID QA/79844/PMCC).
Consent for publication
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.
About this article
Cite this article
Lasocki, A., Sia, J. & Stuckey, S.L. Improving the diagnosis of radiation necrosis after stereotactic radiosurgery to intracranial metastases with conventional MRI features: a case series. Cancer Imaging 22, 33 (2022). https://doi.org/10.1186/s40644-022-00470-6
- Intracranial metastases
- Magnetic Resonance Imaging
- Radiation necrosis
- Stereotactic radiosurgery