Artificial intelligence models for survival prediction in colorectal cancer: a systematic review of time-to-event approaches.

Background: Colorectal cancer (CRC) is a leading cause of cancer-related mortality worldwide, with marked heterogeneity in patient survival. Conventional clinicopathologic staging provides limited individualized prognostic information. Artificial intelligence (AI) and machine learning methods have enabled survival-aware modeling approaches that integrate complex clinical, imaging, and molecular data to improve time-to-event survival prediction. This systematic review aimed to synthesize current evidence on AI-based models for survival prediction in CRC. Methods: A systematic search of PubMed, Scopus, Web of Science Core Collection, IEEE Xplore, and Google Scholar was conducted in accordance with the PRISMA 2020 guidelines. Studies were included if they applied AI methods to model time-to-event survival outcomes in CRC. Records were independently screened, data were extracted, and risk of bias and applicability were assessed using the PROBAST + AI tool. Results: Eight retrospective cohort studies published between 2021 and 2025, encompassing 11 811 patients, met the inclusion criteria. The included studies evaluated diverse data modalities, including clinical variables, radiomics, histopathology, transcriptomics, and genomics. Model performance was moderate to high, with concordance index values ranging from approximately 0.70 to 0.85. All studies demonstrated effective risk stratification of patients into distinct survival groups. Models incorporating high-dimensional imaging or molecular data generally outperformed those based on clinical variables alone. The overall risk of bias was low to unclear, with no study rated as high risk. Conclusion: AI-based time-to-event survival models demonstrate potential prognostic value for prognostic stratification in CRC. Further prospective studies, standardized reporting, and external validation are required to support clinical translation.