BJR open · 2026
Predictive value of digital breast tomosynthesis based intratumoral-peritumoral radiomics and clinical features for benign-malignant breast microcalcifications.
СтатьяНазвание и аннотация: Перевод пока не подготовлен
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.
Wei Junting, Li Jie, Pan Ruixin, Zhu Yun, Gao Zhizhen
Аннотация
OBJECTIVE: To investigate the value of clinical and radiomics features in distinguishing benign from malignant breast microcalcifications and develop a noninvasive model to reduce unnecessary biopsies.
METHODS: This retrospective study analyzed 181 microcalcification lesions from 172 patients. Lesions were divided into training/validation sets (7:3). Manual volumes of interest were delineated on digital breast tomosynthesis images. Radiomics features were extracted from intratumoral and peritumoral regions. Among 9 machine learning classifiers, the optimal performer was selected to develop 3 radiomics models: intratumoral model (RM0), intratumoral and peritumoral 1 mm model (RM1), intratumoral and peritumoral 3 mm model (RM3). Independent clinical risk factors identified by univariate and multivariate logistic regression were integrated with the optimal radiomics model to build a nomogram model (NM). Model performance was evaluated using receiver operating characteristic (ROC) curves, calibration curves, decision curve analysis, and clinical impact curves.
RESULTS: RM0 performed best. Independent risk factors included: Breast Imaging Reporting and Data System (BI-RADS) category (P < .001), history of diabetes (P = .023), and personal history of breast cancer (P = .034). In the training cohort, the areas under the ROC curve (AUC) were 0.884 (clinical model), 0.728 (RM0), and 0.942 (NM). In the validation cohort, AUCs were 0.881, 0.695, and 0.790, respectively.
CONCLUSIONS: Single clinical features showed potential predictive value for microcalcifications, while the integrated model might not demonstrate adequate discriminative power.
ADVANCES IN KNOWLEDGE: This study highlights the need for noninvasive differentiation of microcalcifications and suggests that while individual clinical factors have value, combined modeling requires further optimization.
METHODS: This retrospective study analyzed 181 microcalcification lesions from 172 patients. Lesions were divided into training/validation sets (7:3). Manual volumes of interest were delineated on digital breast tomosynthesis images. Radiomics features were extracted from intratumoral and peritumoral regions. Among 9 machine learning classifiers, the optimal performer was selected to develop 3 radiomics models: intratumoral model (RM0), intratumoral and peritumoral 1 mm model (RM1), intratumoral and peritumoral 3 mm model (RM3). Independent clinical risk factors identified by univariate and multivariate logistic regression were integrated with the optimal radiomics model to build a nomogram model (NM). Model performance was evaluated using receiver operating characteristic (ROC) curves, calibration curves, decision curve analysis, and clinical impact curves.
RESULTS: RM0 performed best. Independent risk factors included: Breast Imaging Reporting and Data System (BI-RADS) category (P < .001), history of diabetes (P = .023), and personal history of breast cancer (P = .034). In the training cohort, the areas under the ROC curve (AUC) were 0.884 (clinical model), 0.728 (RM0), and 0.942 (NM). In the validation cohort, AUCs were 0.881, 0.695, and 0.790, respectively.
CONCLUSIONS: Single clinical features showed potential predictive value for microcalcifications, while the integrated model might not demonstrate adequate discriminative power.
ADVANCES IN KNOWLEDGE: This study highlights the need for noninvasive differentiation of microcalcifications and suggests that while individual clinical factors have value, combined modeling requires further optimization.