AI-mediated ultrasound radiomics in the diagnosis and treatment of triple-negative breast cancer: research progress and future challenges

Triple-negative breast cancer (TNBC) is a subtype of breast cancer with strong invasiveness, limited treatment options, and poor prognosis. Its early accurate diagnosis and individualized treatment are major challenges faced by the clinic. With the rapid development of artificial intelligence (AI) technology, AI-mediated ultrasonic radiomics provides new ideas for non-invasive diagnosis and treatment of TNBC. This technology integrates machine learning (ML), deep learning (DL), and radiomics methods to achieve high-throughput extraction of quantitative features from ultrasonic images and construct a predictive model capable of characterizing tumor heterogeneity. Currently, AI-driven ultrasonic radiomics for TNBC diagnosis has evolved from basic differential diagnosis to multi-subtype classification, with its diagnostic performance further improved via multimodal image fusion. For prognostic assessment, the models effectively predict patients' disease-free survival (DFS) and overall survival (OS) by integrating intratumoral and peritumoral texture features, clinicopathological indicators, and other relevant factors. Nevertheless, the translation of this technology into routine clinical practice faces multiple challenges: insufficient standardized data protocols, limited model interpretability, and lack of rigorous multicenter validation studies. In the future, research on the establishment of a standardized radiomics workflow among different devices and medical centers should be given priority as well as the research on the construction of high-performance AI models with good interpretability, and multicenter prospective clinical studies should be carried out to verify its clinical value.