Premium
SU‐E‐QI‐07: Early Evaluation of Tumor Response to Therapy Via More Accurate Apparent Diffusion Coefficient Maps
Author(s) -
Ma B,
Li X,
Kuang Y
Publication year - 2014
Publication title -
medical physics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.473
H-Index - 180
eISSN - 2473-4209
pISSN - 0094-2405
DOI - 10.1118/1.4888987
Subject(s) - voxel , effective diffusion coefficient , nuclear medicine , false positive paradox , diffusion mri , medicine , logistic regression , radiology , magnetic resonance imaging , mathematics , statistics
Purpose: We investigated the impact of more accurately computed apparent diffusion coefficient (ADC) maps, generated by registration of high‐b‐value and low‐b‐value diffusion weighted MR volumes, on early evaluation of tumor response to therapy. Methods: As in previous work, voxel‐by‐voxel ADC changes in diffusion MR exams during the course of cancer treatment have been shown to be capable of predicting therapeutic response earlier and more accurately than conventional methods. Here we further improved the prediction accuracy using better estimated ADC values. Due to the different susceptibility effects of the different b‐field acquisitions, the low‐ and high‐b‐value diffusion weighted MR volumes can be markedly out of alignment. Nonlinear registration was employed to align these two volumes and more accurate ADC values were produced. Therapeutic efficacy was then evaluated using the voxel‐by‐voxel ADC change analysis on the registered pre‐ and post‐treatment ADC maps. Results: We have applied the described method to 14 sets of pre‐ and post‐treatment diffusion MRI volumes of breast cancer patients. The post‐treatment exams were acquired in average five weeks after the initiation of treatment. Voxel‐by‐voxel ADC change analysis with logistic regression (p=0.009) demonstrated that the probability of a responsive treatment is significantly associated with increased ADC values. There were 2 false positives and 1 false negative (sensitivity=80%, specificity=75%). Using more accurate ADC values estimated with nonlinear registration, ADC change analysis yielded 1 false positive and 1 false negative for a sensitivity of 90% and a specificity of 75%. Conclusion: The study suggests registration of high‐b‐value and low‐b‐value diffusion weighted MR volumes ensures their voxel correspondence and the subsequent estimation of ADC values are more accurate than ADC computation without registration. The improved ADC maps could increase the accuracy of voxelby‐ voxel ADC change analysis of registered interval exams in early assessment of treatment response. NIH/NIGMS U54 GM104944, Lincy Endowed Assistant Professorship