Application of
MR
Cytometry in Predicting Pathologic Complete Response to Neoadjuvant Chemotherapy in Breast Cancer: A Comparative Study
Jie Ding, Diwei Shi, Shiyun Sun, Jinlong Zheng, Wentao Yang, Caixia Fu, Dazhi Chen, Junzhong Xu, Hua Guo, Chao You, Yajia Gu ABSTRACT
Background
MR cytometry imaging is developed to characterize in vivo microstructures. Its clinical value in predicting pathological complete response (pCR) after neoadjuvant chemotherapy (NAC) for breast cancer remains unknown.
Purpose
To implement MR cytometry to predict pCR for breast cancer and compare the performance of different quantitative models (IMPULSED, JOINT, and EXCHANGE).
Study Type
Prospective.
Subject
209 female patients from center A (50.32 ± 11.06 years) and B (48.05 ± 11.30 years).
Field Strength/Sequence
3 T; pulsed gradient spin‐echo diffusion‐weighted imaging (DWI) and oscillating gradient spin‐echo DWI of 17/33 Hz (center A) or 25/50 Hz (center B).
Assessment
Tumor regions were manually delineated. Time‐dependent apparent diffusion coefficients (ADC) and model‐derived microstructural parameters (including cell diameter, intracellular volume fraction, cellularity, extracellular diffusivity, and water exchange rate constant) were calculated before and after NAC. Four multivariate logistic regression models, that is, time‐dependent ADCs, IMPULSED, JOINT, and EXCHANGE, were constructed to predict pCR; performance of combining MR cytometry with ADCs, that is, IMPULSED‐ADC, JOINT‐ADC, and EXCHANGE‐ADC, was also evaluated.
Statistical Tests
Wilcoxon test, Mann–Whitney U test, Logistic regression analysis, DeLong's test, and Pearson correlation analysis, with statistical significance set at p < 0.05.
Results
Changes in intracellular volume fraction, cellularity, and water exchange rate constant before and after NAC exhibited significant differences between pCR and non‐pCR groups. The regression models of time‐dependent ADCs, IMPULSED, JOINT, and EXCHANGE yielded AUCs of 0.709, 0.716, 0.747, and 0.745 in predicting pCR for center A and 0.777, 0.809, 0.846, and 0.864 for center B; the combined models of IMPULSED‐ADC, JOINT‐ADC, and EXCHANGE‐ADC improved AUCs to 0.737, 0.761, and 0.758 for center A and 0.818, 0.860, and 0.868 for center B.
Data Conclusion
JOINT and EXCHANGE, which incorporate transcytolemmal water exchange, outperformed impermeable IMPULSED in predicting NAC for breast cancer, and predictive performance can be further improved by combining MR cytometry parameters with time‐dependent ADCs.
Evidence Level
2.
Technical Efficacy
Stage 2.