| Neoadjuvant therapy (NAT) can effectively reduce the tumor stage in patients with locally advanced gastric cancer, thereby improving long-term outcomes; however, therapeutic responses vary markedly among individuals. Accurate preoperative prediction of response to NAT and timely optimization of treatment strategies are central to the development of individualized perioperative treatment plans. In recent years, immune checkpoint inhibitors combined with neoadjuvant chemotherapy have been increasingly applied in locally advanced gastric cancer and have significantly improved pathological response rates. Meanwhile, the increasing complexity of tumor response patterns in the context of immunotherapy has posed new challenges for efficacy prediction and assessment. Conventional CT imaging and pathological tumor regression grading systems play important roles in evaluating the response to neoadjuvant therapy; however, the former has limited ability to reflect tumor biological heterogeneity, whereas the latter is restricted to postoperative assessment, and neither can fully meet the demands for precise and dynamic prediction. With advances in precision medicine and artificial intelligence, multimodal predictive models that systematically integrate imaging phenotypes, pathological features, and molecular biological information across multiple dimensions offer new opportunities for minimally invasive and accurate efficacy prediction. |
