Purpose:

Quantitative analysis of Intravoxel incoherent motion (IVIM) effect of diffusion weighted imaging (DWI) can assess both diffusion and perfusion component of tissue separately^1,2. Though Bi-exponential model (BE) and its variations have been used widely in research, however there is no standard methodology that is in clinical use². To the best of our knowledge, IVIM quantitative analysis in primary bone tumors has not been reported and no reference IVIM parameter values in bone tumors are available. The purpose of this study was to perform quantitative IVIM parameter estimation using state-of-the-art IVIM analysis method - BE with adaptive Total Variation (TV) penalty function (BE+TV)³ in a cohort of twenty patients with Osteosarcoma.

Methods:

IVIM dataset for twenty patients (n=20, M:F=18:2, Age=16.0±2.9yrs), with Osteosarcoma were acquired under the Institutional Review Board approved protocol (IEC-103/05.02.2016,RP-26/2016) before commencement of chemotherapy. The acquisition was performed on a 1.5T Philips Achieva MRI scanner with Spin Echo-Echo Planar imaging sequence at 11 b-values (0,10,20,30,40,50,80,100,200,400,800 s/mm²), with TR/TE=7541/67msec, matrix size=192×192, slice-thickness/Gap=5mm/0.5mm, field-of-view=250×250mm², voxel-size= .98/3.52/5.0mm. Apparent diffusion coefficient (ADC) was estimated using mono-exponential (ME) model for b≥200s/mm²; as perfusion component is dominant in lower b-values. IVIM parameters Diffusion coefficient (D), Perfusion coefficient (D*), Perfusion fraction (f) were estimated using state-of-the-art IVIM analysis method, BE with adaptive Total Variation Penalty function (BE+TV)³. Commonly used IVIM analysis methodologies like BE model², segmented techniques with 2-parameter fitting^4,5, and 1-parameter fitting^4,5 optimize the signal fitting at each voxel independently; overlooking the spatial context of the signal resulting in high noise in estimated parameters². As spatial homogeneity in the tissue is expected physiologically, the image gradient based penalty Total Variation⁶ was incorporated in the BE+TV method³ to reduce non-physiological spatial inhomogeneity. BE+TV applies non-linear least square optimization for data fitting and the desired spatial homogeneity in the parametric images was achieved by updating the parametric images (D,D*,f) iteratively with the corresponding adaptive penalty function TV to produce qualitatively and quantitatively improved IVIM parameter estimation³. Region of interest (ROI) for tumor and normal tissue were demarcated for each slice manually by a radiologist (>9yrs of experience in cancer MR imaging) for all patients (Figure1.a&b). Tumor ROIs were drawn to include only solid tumor part, excluding bone and necrosis areas. Mean and standard deviation for the estimated quantitative parameters were calculated in tumor and normal tissue volumes. Nonparametric two-sample Kolmogorov-Smirnov (KS) test was perform among estimated parameters in normal and tumor tissue to evaluate statistically significant (p<0.05) difference. Coefficient-of-variation (CV) was calculated as a measure of heterogeneity to evaluate the regional variation within the tissue. Analysis methods were implemented in an in-house built analysis toolbox implemented in MATLAB^® (MathWorksInc.,v2013,Philadelphia,USA).

Results:

Figure1.c shows the IVIM signal fitting with BE+TV method in tumor and normal tissue ROIs presented in Figure1.b. Mean values of diffusion and perfusion parameters for tumor and normal ROI volume for 20 patients are presented in Table1. Both ADC & D were in close range to each other and were observed to be lower in tumor (12.76±2.5x10^-4mm²/s and 12.48±2.3x10^-4mm²/s respectively) than normal tissue (14.79±2.63x10^-4mm²/s and 14.25±2.59x10^-4mm²/s respectively); while both D* & f (30.21±9.48x10^-3mm²/s and 13.43±2.29% respectively) were higher in tumor than normal tissue (26.27±9.32x10^-3mm²/s and 10.71±2.75% respectively). Average inter-patient variability in estimated ADC,D,D*,f were 2.53x10^-4mm²/s, 2.3x10^-4mm²/s, 9.48 x10^-3mm²/s and 2.29% respectively. Average intra-patient variations in estimated ADC,D,D*,f in tumor volume across all data were observed as 4.38±0.79x10^-4mm²/s, 4.7±0.76x10^-4mm²/s, 25.68±3.55x10^-3mm²/s and 8.82±2.39% respectively. CV for all parameters in normal and tumor tissue volume is represented in Table2. Average CV for estimated ADC,D,D*,f in normal and tumor tissue volume were observed to be 25-34%, 27-39%, 110-118% and 66-77% respectively. Estimated parametric maps for ADC,D,D*,f for one representative patient are shown in Figure2.

Discussion:

Estimated ADC (mean:12.76±2.5x10^-4mm²/s; range:7.08-17.58x10^-4mm²/s), and estimated D (mean:12.48±2.3x10^-4mm²/s; range:7.47-16.94x10^-4mm²/s) were in agreement with each other and were comparable with reported ADC values in Osteosarcoma⁷. Though values for IVIM parameters (D,D* &f) in Osteosarcoma are not widely studied in literature, CV for D,D* &f in tumor were comparable with reported studies at the signal to noise ratio (SNR=20) similar to clinical setup^3,5. Higher intra-patient variability in estimated D* and f in tumor volume might be due to the high heterogeneous nature of the tumor. Quantitative IVIM parameters using BE+TV method might be helpful in evaluating changes in tumor heterogeneity in the course of treatment.

Conclusion:

IVIM parameters estimated using state-of-the-art IVIM analysis method - BE with adaptive Total Variation penalty function in Osteosarcoma are reported that are useful for IVIM analysis in bone tumors in the course of treatment and can serve as reference for future simulation studies.

Acknowledgements

Authors would like to thank the Government of India for the funding support required for the study. EBK was supported with the research fellowship funds from Ministry of Human Resource Development, Government of India.