Comparison of MR cytometry methods in predicting immunohistochemical factor status and molecular subtypes of breast cancer.

BACKGROUND

First evaluation of the performance of MR cytometry incorporating transcytolemmal water exchange in predicting immunohistochemical factor status and molecular subtypes of breast cancer.

PATIENTS AND METHODS

We prospectively enrolled 90 breast cancer patients in the study. For each participant, pulsed gradient spin-echo (PGSE) with diffusion time of 70 ms and oscillating gradient spin-echo (OGSE) diffusion-weighted imaging of 25 Hz and 50 Hz were performed on a 3T MRI scanner. Time-dependent apparent diffusion coefficients (ADC) and microstructural parameters including cell diameter , intracellular volume fraction , water exchange rate constant , and apparent extracellular diffusivity were calculated. Single- and multi-variable logistic regression analyses were performed to evaluate their performance in identifying immunohistochemistry (IHC) factor status and molecular subtypes. The area under the receiver operating characteristic curve (AUC) was computed.

RESULTS

The multi-variable regression models generated from MR cytometry-derived metrics provided higher AUC compared to those from time-dependent ADC metrics, . 0.744 . 0.645 for estrogen receptor (ER), 0.727 . 0.688 for progesterone receptor (PR), 0.734 .0.623 for HER2, and 0.679 . 0.633 for Ki67, 0.751 . 0.644 for Triple-Negative Breast Cancer (TNBC), 0.819 . 0.765 for HER2-enriched, 0.730 . 0.659 for Luminal A, 0.633 . 0.633 for Luminal B. MR cytometry with transcytolemmal water exchange (JOINT and EXCHANGE) outperformed the original one with the impermeable model (IMPULSED) in predicting PR (0.727 . 0.705), HER2 (0.734 . 0.689), Ki67 (0.679 . 0.646), TNBC (0.751 . 0.748) and HER2-enriched (0.819 . 0.739), Luminal A (0.730 . 0.666), Luminal B (0.633 . 0.630).

CONCLUSIONS

MR cytometry outperformed conventional ADC measurements in clinical breast cancer subtyping. Incorporating transcytolemmal water exchange further enhanced classification accuracy.