No intervention (observational study)

The goal of this prospective observational study is to evaluate how well a previously developed prediction model can estimate outcomes in adults with gastric cancer. The model combines information from computed tomography (CT) images obtained before treatment with routine clinical information. The main questions are how well the model predicts how long participants live after treatment begins and whether their cancer progresses. Participants will receive their usual medical care. Researchers will collect pretreatment CT images, basic clinical and tumor information, laboratory and tumor marker results, treatment information, and follow-up outcomes obtained during routine care. These data will be entered into the predefined prediction model, and the model's predictions will be compared with what actually happens during follow-up. The study will not change the participants' usual treatment. Model predictions will be used for research purposes only and will not be used to make treatment decisions.

This study is a multicenter retrospective clinical research, led by the First Affiliated Hospital of Wenzhou Medical University, and jointly conducted by other sub-centers. The aim is to develop an non-invasive artificial intelligence system for predicting the response and clinical outcomes of patients with unresectable hepatocellular carcinoma (uHCC) to the treatment with atezolizumab combined with bevacizumab (T+A). In response to the clinical situation where approximately half of uHCC patients do not respond to the standard T+A therapy and traditional invasive biopsy is unable to fully reflect the heterogeneity of the tumor microenvironment, this study plans to retrospectively collect the data of 400 patients who met the inclusion and exclusion criteria from January 2020 to November 2025. The study will systematically summarize multi-dimensional data such as enhanced CT images within one month before treatment, baseline characteristics, serum markers, liver disease factors, and tumor stage. By integrating these clinical features with deep learning imageomics features extracted from images, the research team is dedicated to constructing and validating a safe, non-invasive, and reproducible prediction model, with the aim of achieving precise identification of the benefit population before implementing immunotherapy combined with anti-angiogenic treatment, and providing a powerful intelligent tool support for optimizing clinical treatment decisions and improving patient survival prognosis.