Acta informatica medica : AIM : journal of the Society for Medical Informatics of Bosnia & Herzegovina : casopis Drustva za medicinsku informatiku BiH · 2025
MRI-Derived Radiomics For Classifying Breast Cancer Molecular Subtypes: a Modeling Approach.
СтатьяНазвание и аннотация: Перевод в очереди
Обработка статьи · этапы, ошибки и управление
Загрузка состояния…
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
Hue Tran Thi, Huong Nguyen Thu, Long Tran Quoc, Hung Nguyen Duy
Аннотация
BACKGROUND: Breast cancer is a biologically heterogeneous disease with four major molecular subtypes that determine prognosis and treatment strategies. MRI-based radiomics provides a non-invasive method to predict these subtypes by quantifying tumor heterogeneity.
OBJECTIVE: To develop and validate a logistic-regression model using MRI-derived radiomic features to predict four molecular subtypes of invasive breast cancer: luminal A (LA), luminal B (LB), HER2-enriched (HER2), and triple-negative breast cancer (TNBC).
METHODS: A retrospective cohort of 169 patients with histologically proven invasive breast carcinoma who underwent pre-treatment dynamic contrast-enhanced MRI (DCE-MRI, 3.0 T) was analyzed. Tumors were manually segmented; radiomic texture features were extracted with LIFEx and standardized by z-score normalization. Feature selection was performed using L1-regularized logistic regression (LASSO). Four one-vs-rest logistic-regression models were trained with 5-fold cross-validation. Performance metrics included AUC, sensitivity, specificity, accuracy, and precision.
RESULTS: The models achieved AUCs of 0.840 (TNBC), 0.788 (HER2), 0.661 (LA), and 0.635 (LB). TNBC showed the highest accuracy (0.923), whereas LB had the lowest sensitivity (0.393). Confusion matrices revealed good classification for TNBC and HER2 but frequent misclassification between LA and LB. TNBC-related features were largely intensity- and entropy-based.
CONCLUSION: MRI-derived radiomic signatures can non-invasively differentiate breast-cancer molecular phenotypes, with particularly strong performance for TNBC and HER2. Although LA-LB separation remains limited, the LASSO-logistic-regression framework offers moderate-to-high diagnostic accuracy and potential value as a complementary precision-oncology decision-support tool.
OBJECTIVE: To develop and validate a logistic-regression model using MRI-derived radiomic features to predict four molecular subtypes of invasive breast cancer: luminal A (LA), luminal B (LB), HER2-enriched (HER2), and triple-negative breast cancer (TNBC).
METHODS: A retrospective cohort of 169 patients with histologically proven invasive breast carcinoma who underwent pre-treatment dynamic contrast-enhanced MRI (DCE-MRI, 3.0 T) was analyzed. Tumors were manually segmented; radiomic texture features were extracted with LIFEx and standardized by z-score normalization. Feature selection was performed using L1-regularized logistic regression (LASSO). Four one-vs-rest logistic-regression models were trained with 5-fold cross-validation. Performance metrics included AUC, sensitivity, specificity, accuracy, and precision.
RESULTS: The models achieved AUCs of 0.840 (TNBC), 0.788 (HER2), 0.661 (LA), and 0.635 (LB). TNBC showed the highest accuracy (0.923), whereas LB had the lowest sensitivity (0.393). Confusion matrices revealed good classification for TNBC and HER2 but frequent misclassification between LA and LB. TNBC-related features were largely intensity- and entropy-based.
CONCLUSION: MRI-derived radiomic signatures can non-invasively differentiate breast-cancer molecular phenotypes, with particularly strong performance for TNBC and HER2. Although LA-LB separation remains limited, the LASSO-logistic-regression framework offers moderate-to-high diagnostic accuracy and potential value as a complementary precision-oncology decision-support tool.