AI and Cancer Research ── Breast ── 2026-10-13
Breast cancer: treatment, trials and AI
- Abstract
BACKGROUND: The heterogeneity of breast cancer is a key factor influencing treatment decisions and patient prognosis. As a crucial cell proliferation marker, the expression level of the Ki-67 protein is closely associated with tumor aggressiveness, recurrence risk, and prognosis. The goal of this study was to develop and validate an integrated radiomics model based on digital breast tomosynthesis (DBT) that incorporates both habitat subregions and peritumoral features to preoperatively predict Ki-67 expression status in breast cancer. METHODS: A total of 146 patients with invasive breast cancer who underwent preoperative DBT were retrospectively enrolled in the study and divided into training (n=102) and testing (n=44) cohorts. Intratumoral regions were manually segmented, whereas peritumoral regions (2, 4, and 6 mm) were automatically delineated using mask padding. Moran's index clustering was used to divide the tumors into four habitat subregions. Radiomics features were extracted from the intratumoral regions, habitat subregions, and peritumoral regions. Following feature selection, radiomics models were established using machine learning algorithms. A combined model was subsequently developed by integrating habitat subregions and peritumoral radiomics features with clinical factors. The area under the curve (AUC) was used to assess the performance of radiomics models. RESULTS: In predicting Ki-67 expression status, the rbf_support vector machine (SVM) model, constructed from habitat radiomics features, outperformed the other models, with AUC values of 0.841 and 0.820 in the training and testing cohorts, respectively. The combined model slightly increased the AUC values to 0.849 and 0.852, respectively. CONCLUSIONS: A combined model integrating habitat, peritumoral radiomics features, and clinical factors demonstrated optimal accuracy in predicting Ki-67 expression levels in breast cancer; this model may help clinicians develop personalized treatment plans for breast cancer patients.
Journal IF-equivalent: 1.9 (OpenAlex 2-year mean citedness, value as of 2026-10-12, retrieved 2026-10-13; not the official Clarivate IF)Reference: Gong J, Zhuang Z, Liu S, Xu L, Liu H, Zhu H, et al. Moran’s index habitat and peritumoral radiomics based on digital breast tomosynthesis for predicting Ki-67 status in breast cancer. Quant Imaging Med Surg. 2026 Oct;16(10):807-7. doi:10.21037/qims-2026-1-0047.Checked: Abstract only