Prognostic prediction model for triple-negative breast cancer using artificial intelligence and ultrasound radiomics.
[OBJECTIVE] To develop and validate an artificial intelligence (AI)-driven ultrasound radiomics model for predicting postoperative recurrence and metastasis in patients with triple-negative breast can
APA
Zilalan Y, Fan J, et al. (2026). Prognostic prediction model for triple-negative breast cancer using artificial intelligence and ultrasound radiomics.. Frontiers in oncology, 16, 1654953. https://doi.org/10.3389/fonc.2026.1654953
MLA
Zilalan Y, et al.. "Prognostic prediction model for triple-negative breast cancer using artificial intelligence and ultrasound radiomics.." Frontiers in oncology, vol. 16, 2026, pp. 1654953.
PMID
41815549
Abstract
[OBJECTIVE] To develop and validate an artificial intelligence (AI)-driven ultrasound radiomics model for predicting postoperative recurrence and metastasis in patients with triple-negative breast cancer (TNBC) receiving standardized therapies.
[METHODS] We conducted a retrospective study of 668 female TNBC patients (treated 2013-2018). Univariate and multivariate logistic regression were first used to screen significant clinicopathological variables for baseline assessment and to inform model development. Radiomic features were automatically extracted from pretreatment ultrasound images using PyRadiomics following tumor segmentation. A radiomics signature was constructed by integrating LASSO for feature selection with a support vector machine (SVM) classifier. The model's performance was evaluated using the area under the receiver operating characteristic curve (AUC), calibration curves, decision curve analysis (DCA), and confusion matrices.
[RESULTS] The ultrasound radiomics model showed high predictive accuracy for any recurrence/metastasis, with an AUC of 0.9458 in the training cohort and 0.8983 in the validation cohort. For distinguishing between locoregional recurrence and distant metastasis, the model achieved AUCs of 0.9341 and 0.8824 in the training and validation cohorts, respectively. Calibration and decision curve analyses confirmed the model's robust predictive capability and potential clinical utility.
[CONCLUSION] This study demonstrates that an AI-enhanced ultrasound radiomics model can effectively predict postoperative recurrence and metastasis patterns in TNBC, offering a promising non-invasive tool to support personalized prognosis assessment.
[METHODS] We conducted a retrospective study of 668 female TNBC patients (treated 2013-2018). Univariate and multivariate logistic regression were first used to screen significant clinicopathological variables for baseline assessment and to inform model development. Radiomic features were automatically extracted from pretreatment ultrasound images using PyRadiomics following tumor segmentation. A radiomics signature was constructed by integrating LASSO for feature selection with a support vector machine (SVM) classifier. The model's performance was evaluated using the area under the receiver operating characteristic curve (AUC), calibration curves, decision curve analysis (DCA), and confusion matrices.
[RESULTS] The ultrasound radiomics model showed high predictive accuracy for any recurrence/metastasis, with an AUC of 0.9458 in the training cohort and 0.8983 in the validation cohort. For distinguishing between locoregional recurrence and distant metastasis, the model achieved AUCs of 0.9341 and 0.8824 in the training and validation cohorts, respectively. Calibration and decision curve analyses confirmed the model's robust predictive capability and potential clinical utility.
[CONCLUSION] This study demonstrates that an AI-enhanced ultrasound radiomics model can effectively predict postoperative recurrence and metastasis patterns in TNBC, offering a promising non-invasive tool to support personalized prognosis assessment.