The value of artificial intelligence in ultrasound imaging for predicting molecular subtypes of breast cancer: a meta-analysis

Background The study aims to integrate an evaluation of the accuracy and validity of ultrasonography based artificial intelligence (AI) algorithms for predicting the molecular subtypes of breast cancer patients through a meta-analysis. Methods A search of the PubMed, Embase, Web of Science, and Cochrane Library databases was performed to locate relevant literature, and the reported studies before February 2026 were included. We evaluated the quality of the studies included by utilizing the Quality Assessment of Diagnostic Accuracy Studies (QUADAS) questionnaire. Two evaluators independently searched for literature and assessed the quality of literature included in the study. Results A total of thirteen studies assessing a number of 13615 patients of breast cancer were included. The results demonstrated that ultrasonic imaging combined with artificial intelligence algorithm has promising accuracy and effectiveness to predict molecular subtypes of breast cancer. The pooled sensitivity and specificity were 0.89 (95%CI: 0.82-0.93) and 0.82 (95%CI: 0.77-0.86), respectively. Additionally, the diagnostic odds ratios (DOR), positive likelihood ratio (PLR) and negative likelihood ratio (NLR) were 32.10 (95%CI: 18.60-55.38), 4.84 (95%CI:3.80-6.17), and 0.14 (95%CI: 0.09-0.22). The area under the curve (AUC) was 0.91 (95%CI: 0.88-0.93). Publication bias was not significantly observed. Conclusions Ultrasonic imaging based on artificial intelligence algorithm has good performance and application prospects for forecasting breast cancer molecular subtypes. This technique can help establish the molecular subtype of breast cancer before operation, offering effective help for the treatment plan. It may reduce unnecessary biopsy, which is anticipated to become a meaningful implement in clinical application. Systematic review registration https://www.crd.york.ac.uk/prospero/display_record.php?ID=CRD42024599983 , identifier CRD42024599983.