Dynamic contrast-enhanced MRI-derived intratumoral heterogeneity quantification score: Improving lymphovascular invasion and invasive breast cancer recurrence-free survival predictions.

INTRODUCTION To investigate whether preoperative magnetic resonance imaging (MRI)-based quantification of intratumoral heterogeneity (ITH) in invasive breast cancer (IBC) can predict lymphovascular invasion (LVI) and recurrence-free survival (RFS).

METHODS In this retrospective cohort study, 223 patients with histopathologically confirmed IBC who underwent preoperative dynamic contrast-enhanced MRI were analyzed. ITH scores were calculated using texture features and pixel distribution. Peak tumor enhancement phase-driven clustering defined tumor subregions for radiomics and habitats feature extraction. Six machine-learning models integrating MRI morphological features (MRI-MF) and significant indicators contributed to the development of the prognostic nomogram. Performance was assessed via area under the receiver operating characteristic curve (AUC), Cox proportional hazards regression, and Kaplan-Meier analyses.

RESULTS In the validation set, the ITH score exhibited superior predictive accuracy (AUC: 0.791, 95 % confidence interval [CI]: 0.706-0.870) compared to habitats (AUC: 0.762, 95 % CI: 0.676-0.837) and radiomics (AUC: 0.739, 95 % CI: 0.648-0.826) analyses. The CatBoost-driven hybrid model, combining ITH and MRI-MF, achieved the highest performance (AUC: 0.869, 95 % CI: 0.763-0.948). Kaplan-Meier analysis indicated significantly shorter RFS for patients with LVI (+), high ITH scores, and high nomogram risk (all p < 0.005).

CONCLUSION The ITH score predicts LVI and RFS in patients with IBC, with the CatBoost-driven hybrid model improving personalized prognosis and treatment.

IMPLICATIONS FOR PRACTICE The findings of this study underscore the potential of noninvasive ITH quantification for precise risk assessment, with possible implications for preoperative decision-making in IBC management.