Research objects
This study was approved by our institutional animal care committee and hospital ethics committee. Following the methods of previous studies, VX2 tumor cells (Central Laboratory of Shanghai General Hospital) and male New Zealand white rabbits (Qingdao Administration for Market Regulation) were used to establish malignant bone tumor model (n = 50) [15, 16]. All rabbits were 2 months of age at tumor seeding and weighed 2.0–3.0 kg. The animals were maintained in a specific pathogen-free environment with a room temperature of 22 °C–25 °C and a circadian rhythm of 12/12 light/dark. Tumor nodules were cut into 1-2 mm3 tissue chunks and immersed in 0.9% sodium chloride injection. For implantation, the prepared VX2 tissue fragments were delivered into the medullary cavity at the level of the right tibial nodule at a depth of 1 cm. Follow-up examinations were performed when the tumor was larger than 10 mm in diameter after 3 weeks. Considering that the same animal could not be repeated for pathological examination, and that all animals used in this study were subjected to the same survival and exposure conditions, the animals with bone tumor were divided into pre-treatment and post-treatment groups according to the principle of randomized control with a ratio of 3:7. In the pre-treatment group, baseline MRI scans were performed after model establishment. The post-treatment group underwent the intensity modulated radiation therapy (total dose 10 Gy, radiation field 6 × 6 cm2) and was then subjected to longitudinal MRI scanning after 5 days. Animals were sacrificed at the end of MR examination and pathological examination was performed. The post-treatment group was further divided into slight effect group and considerable effect group according to the pathological assessment. The workflow is shown in Fig. 1.
MRI acquisition
The phantom experiment of rabbit malignant bone tumor model was tested before the MRI scanning to evaluate the accuracy and repeatability of the quantitative IVIM-DKI parameters used in this study. Details of this investigation are shown in Supplement A1. All MRI was acquired using a 3 T MRI system (Prisma, Siemens, and Erlangen, Germany) with an eight-channel rabbit coil (Chenguang, China). Ketamine (1.5 mg/kg) was injected into the thigh muscle to maintain rabbit anesthesia. The rabbits were place in a supine position and allowed to breathe freely during data acquisition. The sequences included sagittal T2-weighted fat-saturated (FS) images (echo time/repetition time, 96/3000 ms; FOV, 160 × 160 mm; matrix, 320 × 320; slice thickness, 3 mm; intersection gap, 1 mm; and number of excitations, 2) and sagittal T1-weighted FS fast spin-echo (FSE; echo time/repetition time, 22/737 ms; FOV, 160 × 160 mm; matrix, 320 × 320; slice thickness, 3 mm; intersection gap, 1 mm; and number of excitations, 2). IVIM and DKI were based on readout-segmented long variable echo-trains using the RESOLVE technique (GRAPPA, R = 2; echo time/repetition time, 55/3000 ms; FOV, 170 × 170 mm; matrix, 128 × 100; slice thickness, 3 mm; diffusion gradient directions, 3; number of excitations, 4; receiver bandwidth, 930 Hz/pixel; b values 0, 10, 20, 30, 40, 50, 80, 100, 150, 200, 400, 600, 800, 1000, 1500, and 2000 s/mm2; and scanning time: 7.4 minutes).
Histopathology analysis
After MRI was completed, all rabbits were sacrificed and the right tibia specimens were harvested. According to the direction and thickness of the MR scan, the largest sagittal plane of the tumor was selected, and the tumor was sliced into 3-mm thick slices. The gross pathology was recorded and hematoxylin-eosin (HE) staining was performed. The pathologists evaluated the tumor cell density in the solid area by point-to-point comparison with gross slice under high- power microscope (200×). Tumor cell density was calculated as: Tumor cell density = Areatumor cell nucleus/ Areastatistical field. Treatment response was assessed by classifying the degree of tumor regression into three grades based on the decrease in tumor cell density. Tumor regression of 0–40% was defined as grade I, 40–60% as grade II, and > 60% as grade III. Grade I (tumor cell regression < 40%) was defined as slight effect, while grade II or grade III (tumor cell regression > 40%) was defined as considerable effect [17]. All pathologic analyses were performed by one experienced senior pathologist with more than 5 years of experience.
Image postprocessing
Image analysis was performed using the Body Diffusion Toolbox (Siemens Healthcare GmbH, Erlangen, Germany) on the Siemens workstation (syngo.via). Two experienced radiologists analyzed the images independently, with any disagreements being resolved through consultation. It should be noticed that IVIM and DKI have different b-values selected for subsequent processing. The IVIM was fitted on the basis of images with b-values of 0, 10, 20, 30, 40, 50, 80, 100, 150, 200, 400, 600, 800, and 1000 s/mm2. In addition, DKI was fitted according to images with b-values of 0, 1000, 1500, and 2000 s/mm2. The fitting diagram was attached to Fig. 2. Considering the heterogeneity of the tumor, the radiologist manually defined the region of interest (ROI) of the tumor contour on the corresponding tissue diffusion (Dt) map of the highest b-value, based on IVIM-DKI images of the largest sagittal plane of the tumor, using the same level of HE-stained specimens as a guide (Fig. 3). The entire solid tumor volume is covered, carefully avoiding cystic, hemorrhagic and artifact areas, and plotted three times to obtain a final average. The ROI was then automatically replicated to other parameter maps, including pseudo-diffusion (Dp), perfusion fraction (fp) for IVIM, and mean diffusion coefficient (MD) and mean kurtosis (MK) for DKI.
Statistical analysis
All statistical analyses were performed using SPSS (version 22.0; Chicago, IL, USA) and MedCalc (version 19.0.7; Ostend, Belgium). The Kolmogorov–Smirnov test of normality was performed for analyzing normality at p value>0.05. The mean value of the preprocessing parameters and the early changes after 5 days of radiotherapy were compared between the pre-treatment and post-treatment group using independent sample t-tests. A one-way ANOVA and Tukey post hoc multiple comparisons were used to analyze the differences in each quantitative diffusion parameters between the pre-treatment, considerable effect, and slight effect malignant bone tumor groups. Correlations between the diffusion parameters of the IVIM and DKI models were analyzed using Pearson correlation. Binary logistic regression analysis was performed to establish a diagnostic model with a combination of different parameters including Dt (cellularity), fp (vascularity), and MK (microstructural complexity) of IVIM and DKI, according to the results of the analyses stated above. Receiver operating characteristics (ROC) curves were created and the areas under the curves (AUCs) were calculated to evaluate the diagnostic performance of each diffusion parameter and their combinations for differentiating between considerable effect and slight effect groups. A p value of < 0.05 was considered significant for the above analyses.