Frontiers in oncology · 2026
Radiomics analysis of biplanar ultrasound images can discriminate non-mass breast carcinoma from mastitis.
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.
Wu Qinfu, Liu Guangde, Xiao Mengqiang, Xie Shanghuang, Teng Wenhui, Dong Yang, Chen Xiaoyi, Liu Tianzhu, Huang Peikai
Аннотация
BACKGROUND: Non-mass breast carcinoma (NMBC) is a diagnostically challenging subtype with sonographic features that often overlap with mastitis, leading to misdiagnosis and delayed treatment. Unlike mastitis, which typically warrants conservative therapy, NMBC requires prompt oncological management, making accurate preoperative differentiation critically important. Using transverse, longitudinal, and fused biplanar ultrasound images, we constructed radiomics models to distinguish NMBC from mastitis and evaluated the incremental diagnostic value of combining radiomics features with clinical variables.
METHODS: We retrospectively analyzed data from 139 patients (NMBC, 63; mastitis, 76) and extracted radiomics features from manually segmented transverse and longitudinal ultrasound images. After feature stability screening and least absolute shrinkage and selection operator selection, we constructed three radiomics logistic regression models for transverse, longitudinal, and fused imaging data, respectively, and additionally developed clinical variable-based models (age and BI-RADS) and combined clinical-radiomics models. The models were trained and internally validated using a 70/30 split, performance was assessed based on area under the curve (AUC), accuracy, sensitivity, specificity, and decision curve analysis (DCA) values.
RESULTS: In the training cohort, the fusion model (combining features from transverse and longitudinal images) outperformed single-plane models (AUC = 94.2%, accuracy = 87.6%). However, this advantage diminished in the validation cohort, where the transverse, longitudinal, and fusion radiomics models showed comparable performance (AUC = 0.730, 0.823, and 0.800, respectively; accuracy = 69.0%, 78.6%, and 78.6%, respectively). DeLong test revealed no significant differences among the models in the validation cohort. The combined clinical-radiomics model outperformed both radiomics and clinical variable-based models (AUC: 0.861-0.884 in the validation cohort); likelihood ratio tests confirmed that radiomics features and clinical variables contributed independently to the prediction. Calibration and DCA revealed that the fusion model showed superior clinical utility in the training cohort but demonstrated limited generalizability in the validation cohort.
CONCLUSION: Biplanar ultrasound imaging-based radiomics models show promise for distinguishing NMBC from mastitis. Although the fusion model excelled in the training cohort, it offered limited advantage internal hold-out validation. Integrating clinical variables with radiomics features offers added diagnostic value beyond either approach alone, suggesting that a combined clinical-radiomics strategy may be more clinically viable.
METHODS: We retrospectively analyzed data from 139 patients (NMBC, 63; mastitis, 76) and extracted radiomics features from manually segmented transverse and longitudinal ultrasound images. After feature stability screening and least absolute shrinkage and selection operator selection, we constructed three radiomics logistic regression models for transverse, longitudinal, and fused imaging data, respectively, and additionally developed clinical variable-based models (age and BI-RADS) and combined clinical-radiomics models. The models were trained and internally validated using a 70/30 split, performance was assessed based on area under the curve (AUC), accuracy, sensitivity, specificity, and decision curve analysis (DCA) values.
RESULTS: In the training cohort, the fusion model (combining features from transverse and longitudinal images) outperformed single-plane models (AUC = 94.2%, accuracy = 87.6%). However, this advantage diminished in the validation cohort, where the transverse, longitudinal, and fusion radiomics models showed comparable performance (AUC = 0.730, 0.823, and 0.800, respectively; accuracy = 69.0%, 78.6%, and 78.6%, respectively). DeLong test revealed no significant differences among the models in the validation cohort. The combined clinical-radiomics model outperformed both radiomics and clinical variable-based models (AUC: 0.861-0.884 in the validation cohort); likelihood ratio tests confirmed that radiomics features and clinical variables contributed independently to the prediction. Calibration and DCA revealed that the fusion model showed superior clinical utility in the training cohort but demonstrated limited generalizability in the validation cohort.
CONCLUSION: Biplanar ultrasound imaging-based radiomics models show promise for distinguishing NMBC from mastitis. Although the fusion model excelled in the training cohort, it offered limited advantage internal hold-out validation. Integrating clinical variables with radiomics features offers added diagnostic value beyond either approach alone, suggesting that a combined clinical-radiomics strategy may be more clinically viable.