Time-dependent diffusion MRI as in vivo histology: microstructural biomarkers for breast cancer diagnosis and prognostic stratification

BACKGROUND: Time-dependent diffusion MRI (TD-dMRI) holds potential for characterizing tumor cell properties; however, its utility in diagnosing various breast lesions and evaluating Ki-67 expression remains unexplored. This study aimed to assess the utility of TD-dMRI for distinguishing benign breast lesions, ductal carcinoma in situ (DCIS) and invasive breast carcinoma (IBC), and correlating its derived parameters with molecular subtypes and histopathological features. METHODS: This retrospective study enrolled participants with breast lesions who underwent oscillating and pulsed gradient encoded TD-dMRI and conventional multi-parametric MRI from May 2024 to January 2025. The imaging microstructural parameters using limited spectrally edited diffusion (IMPULSED) model was used to extract microstructural parameters, including mean cell diameter (Dmean), cellularity, intracellular volume fraction (Vin), and extracellular diffusivity (Dex). Histopathological results (molecular subtypes, Nottingham grade, and Ki-67 index) served as the reference standard. Statistical analyses comprised receiver operating characteristic (ROC) curve analysis, correlation tests, and intergroup comparisons. RESULTS: This study included 104 patients with 108 breast lesions. Significant differences were observed in TD-dMRI parameters among benign breast lesions, DCIS and IBC, and between high and low expression of Ki-67 (P < .05). IBC exhibited the highest cellularity (3.30 ± 0.77 μm− 1), the largest Dmean (10.26 ± 1.53 μm), the highest Vin (30.42% ± 7.33%), and the lowest ADC value (0.83 ± 0.10 × 10− 3mm2/s). High Ki-67 expression exhibited higher cellularity (3.38 ± 0.70 μm− 1), Dmean (10.51 ± 1.38 μm), Vin (31.10 ± 6.97%) and lower Dex (2.11 ± 0.22 µm2/ms) and ADC value (0.84 ± 0.11 × 10− 3mm2/s) compared to low Ki-67 expression. ADCDWI demonstrated the highest diagnostic performance in identifying benign lesions (AUC = 0.886) and IBC (AUC = 0.917). Vin showed considerable efficacy in distinguishing high Ki-67 expression from low Ki-67 expression (AUC = 0.801). Luminal B tumors displayed higher Dmean (10.63 ± 1.47 μm), Vin (31.48 ± 7.96%), lower ADC (0.82 ± 0.11 × 10− 3mm2/s) and Dex (2.07 ± 0.28 µm2/ms). CONCLUSION: TD-dMRI noninvasively quantifies microstructural features of breast lesions, improving diagnostic accuracy and providing valuable correlations with molecular subtypes and proliferative activity, with promise for clinical translation in precision oncology. TRIAL REGISTRATION: Registry: Zhongnan Hospital of Wuhan University, TRN: 2,025,221 K, registration date: 27 August 2025, retrospectively registered.