0870
IVIM Analysis in Response Evaluation of Osteosarcoma Treated with Neoadjuvant Chemotherapy: Correlation with Histopathological Necrosis1Center for Biomedical Engineering, Indian Institute of Technology, Delhi, India, New Delhi, India, 2Radio Diagnosis, All India Institute of Medical Sciences, New Delhi, India, New Delhi, India, 3Department of Physics, Indian Institute of Technology, Delhi, India, New Delhi, India, 4Medical Oncology, IRCH, All India Institute of Medical Sciences, New Delhi, India, New Delhi, India
Synopsis
Histological necrosis is the current gold standard for response evaluation in osteosarcoma treated with neoadjuvant chemotherapy (NACT). However it is applicable only after tumor resection on completion of NACT. Thus, a non-invasive early marker of NACT response is desirable. We performed NACT response prediction and evaluation using Intra-voxel Incoherent Motion (IVIM) Diffusion weighted MRI and histogram analysis. IVIM parameters and its histogram analysis revealed clinically useful information in characterizing and predicting chemotherapy response to Osteosarcoma.
Purpose
Histopathological necrosis is the current gold standard for response evaluation in osteosarcoma treated with neoadjuvant chemotherapy (NACT), but can be performed only after completion of NACT. Intra-voxel Incoherent Motion (IVIM) analysis is reported to have a significant role as quantitative non-invasive imaging biomarker for various clinical applications1,2. Objective of this study was to investigate the role of IVIM parameters as surrogate markers of response prediction and evaluation early in the course of NACT.Methods
IVIM dataset for twenty patients (n=20;M:F=17:3;Age=16.6±2.2years; Metastatic:localized=13:7) with osteosarcoma were acquired. All patients underwent pre-operative NACT consisting of 3 cycles of Cisplatin and Doxorubicin3 every three weekly. The gold standard used was histopathological examination of operative specimen and histological necrosis≥90% was considered as good response to NACT. IVIM datasets were acquired at three time-points – pre-NACT (t0), after 1stNACT (t1) and after 3rdNACT (t2) using free breathing Spin Echo-Echo Planar imaging (SE-EPI) with varying gradient strengths at 11 b-values(0,10,20,30,40,50,80,100,200,400,800 s/mm2).
Tumor volume at different time-points was determined separately using region of interest (ROI) drawn manually by a radiologist (>10years of experience) across the tumor. Apparent diffusion coefficient(ADC) and IVIM parameters such as Diffusion coefficient(D), Perfusion coefficient(D*), Perfusion fraction(f) were estimated in tumor volume at different time-points (t0,t1&t2) using a novel methodology, bi-exponential model with Total Variation (TV) penalty function (BE+TV)4. It has been shown that BE+TV may be more reliable for IVIM analysis compared to voxel-wise fitting of the IVIM bi-exponential (BE) model1. Histogram analysis5 was performed on diffusion parameters (ADC,D,D*&f) and compared with histological necrosis. Nonparametric two-sample Mann-Whitney-Wilcoxon (MWW) test was applied to evaluate significant (p<0.05) difference between good responders (GR) and poor responders (PR) groups and paired t-test was used to evaluate significant (p<0.05) changes in parameters across different time-points. Receiver-operating-characteristic curve (ROC) analysis was performed to evaluate the performance of different parameters for predicting chemotherapy response.
Results
According to histopathological assessment, four patients(20%) were classified as GR and sixteen patients(80%) as PR (GR:PR=4:16). Representative images of one patient each from GR and PR groups at different time-points are presented in Figure 1&2. At t0, average ADC & D in tumor volume were observed as ADC:1.37±0.31×10-3mm2/s & D:1.28±0.27×10-3mm2/s respectively and average D* & f in tumor volume were observed as D*:26.39±10.5×10-3mm2/s & f:13.99±2.44% respectively among all patients. Histogram analysis of IVIM parameters in tumor at different time points for GR and PR groups are summarized in Table1.
Before NACT, skewness and kurtosis of ADC and D were observed to be higher among GRs (skewness:1,1 & kurtosis:2.4,2 respectively) than PRs (skewness:0.6,0.4 & kurtosis:1.7,1.5 respectively) and during NACT these showed greater reduction in GR than PR group (skewness:80-90% vs 10-30% reduction & kurtosis:30-84% vs 23-44% reduction for ADC & D). Among GRs average D* and its variance showed significant reduction during NACT (28x10-3mm2/s vs 16x10-3mm2/s,p=0.01 and 8x10-4 vs 4x10-4,p=0.01 respectively) than PR group (26x10-3mm2/s vs 20x10-3mm2/s,p=0.07 and 6x10-4 vs 5x10-4,p=0.16 respectively) and skewness & kurtosis of D* were observed to be significantly increased among GRs (0.8 vs 1.8,p=0.02 and 0.7 vs 3.3,p=0.04 respectively) during NACT than PR group (1 vs1.4, p=0.12 and 0.7 vs1.5,p=0.41 respectively). After 1stNACT, mean energy(measuring homogeneity) of D* among GRs was observed to be significantly lower than PRs (1.69x10-3 vs 3.85x10-3,p=0.01); although higher among GRs(7.4x10-3 vs 4.13x10-3) after completion of NACT. Average f among both the groups showed similar decrease during NACT(12% vs 14%); although variance and energy of f among GRs were observed to be significantly higher than PRs (1.41x10-2 vs 0.6x10-2,p=0.01 and 0.42x10-3 vs 0.15x10-3,p=0.02 respectively) at t1.
Pre-NACT, using ROC curve analysis, skewness of ADC & D (Sn:75%;Sp:63%,88%; Th:1.15,0.71 respectively) and variance of D* (Sn:75%;Sp:88%;Th:0.8x10-3) individually showed AUC of 0.75-0.78; while in combination resulted in Sn:75%,Sp:94% with AUC=0.83 in predicting good response to NACT (Figure3.a). After 1stNACT, energy of ADC & D* (Sn:75%;Sp:56-63%;Th:1.16x10-3,0.35 x10-3 respectively) and variance, energy of f (Sn:75%;Sp:56-75%;Th:0.58x10-2,0.19x10-3 respectively) individually produced AUC of 0.73-0.77 and jointly showed Sn:75%,Sp:88% with AUC=0.91 in predicting NACT response (Figure3.b).
Discussion
During chemotherapy ADC and D among GRs became more negatively skewed and flattened with increased energy indicating increased diffusion and comparatively more homogeneity in tumor than PR group. Skewness, kurtosis of D* observed to be significantly increased (positively skewed) among GRs indicating greater reduction of perfusion in tumor than the PR group. Energy of D* and f among GRs showed increment during NACT indicating plausibly homogeneous perfusion pattern in tumor than PR group. Energy measuring homogeneity was observed as a useful histogram parameter for predicting NACT response early in the course of NACT.Conclusion
IVIM parameters and its histogram parameter, skewness & energy, may be useful in response prediction and evaluation of osteosarcoma treated with NACT.Acknowledgements
Authors would like to thank the Government of India for the funding support required for the study. E.B.K. was supported with the research fellowship funds from Ministry of Human Resource Development. Authors would also like to thank the nursing and support staff for helping in patient recruitment and scanning.References
1. Le Bihan D. et al. Separation of diffusion and perfusion In Intravoxel Incoherent Motion MR Imaging. Radiology. Aug, 1988; 168(2):497-505.
2. Koh D.M. et al. Intravoxel Incoherent Motion in Body Diffusion-Weighted MRI: Reality and Challenges. AJR. June, 2011;196: 1351–1361.
3. Geller DS, Gorlick R. Osteosarcoma : A review of diagnosis , management , and treatment strategies. Clin Adv Hematol Oncol. 2010;8(10):705-718.
4. Baidya Kayal E. et al. Quantitative Analysis of Intravoxel Incoherent Motion ( IVIM ) Diffusion MRI using Total Variation and Huber Penalty Function. Med Phys. 2017;44(11):5489-5858. doi:10.1002/mp.12520.
5. Mardia KV. Measures of multivariate skewness and kurtosis with applications. Biometrika. 1970; 57(3): 519-530.
Figures
