Abstract
Background: Breast cancer brain metastases (BCBMs) are clinically challenging, and treatment decisions are influenced by tumor biology. Because receptor profiles may differ between primary breast tumors and brain metastases and brain biopsy may be impractical, non-invasive imaging biomarkers may provide useful biologic correlates. We evaluated whether diffusion-weighted imaging (DWI)-derived apparent diffusion coefficient (ADC) histogram metrics from BCBM were associated with primary tumor estrogen receptor (ER), progesterone receptor (PR), and human epidermal growth factor receptor 2 (HER2) status; the Ki-67 proliferation index; and luminal status. Methods: This retrospective exploratory single-center study included 72 adults with BCBM who underwent standardized 1.5T brain magnetic resonance imaging. The largest lesion in each patient was segmented on ADC maps in FireVoxel. ADC histogram features, including percentiles, were extracted. Using primary tumor biomarker status as the reference, candidate metrics were screened by univariable logistic regression. Parsimonious multivariable models included age, log-transformed lesion volume, and a single selected ADC percentile scaled by ×10. Discriminatory performance was assessed using area under the receiver operating characteristic curve (AUC); thresholds were derived with the Youden index. No external validation was performed. Results: Low-percentile ADC metrics were associated with ER positivity, PR positivity, and luminal disease, whereas no meaningful ADC histogram discrimination was observed for HER2. In multivariable models, ADC10×10 predicted ER positivity (odds ratio [OR] 0.441; AUC 0.847) and PR positivity (OR 0.478; AUC 0.819). Ki-67 positivity was best predicted by ADC75×10 (OR 3.095; AUC 0.905), although this finding should be interpreted cautiously. Luminal status (non-luminal vs. luminal) was predicted by ADC10×10 (OR 2.251; AUC 0.832). Conclusions: ADC histogram analysis from DWI in BCBM showed exploratory associations with primary tumor hormone receptor status and luminal subtype, but not HER2. These findings support ADC histogram features as candidate imaging biomarkers, but the Ki-67 result and all model performance estimates require cautious interpretation and independent external validation in multicenter cohorts.