Artificial intelligence for predicting long-term outcomes in patients with colorectal cancer (a systematic review and meta-analysis)

Abstract

AIM: to evaluate the performance of artificial-intelligence algorithms in predicting long-term treatment outcomes in patients with colorectal cancer using clinical data alone to evaluate the performance of artificial-intelligence algorithms in predicting long-term treatment outcomes in patients with colorectal cancer (CRC) using clinical data alone. MATERIALS AND METHODS: a systematic search (2015–2024) was conducted in PubMed, Science Direct, MedRxiv, BioRxiv and Google Scholar. Original studies that applied machine-learning or deep-learning techniques exclusively to clinical variables for predicting CRC recurrence were included. Of 657106 records screened, 43 met the eligibility criteria; 12 were entered into a meta-analysis. Pooled area under the ROC curve (AUC), heterogeneity metrics (I², τ², Q-test), publication bias and sensitivity were assessed. Robustness was examined with a leave-one-out analysis. RESULTS: a systematic search (2015–2024) in PubMed, Science Direct, MedRxiv, BioRxiv and Google Scholar. Original studies that applied machine-learning or deep-learning techniques exclusively to clinical variables for predicting CRC recurrence were included. Of 657106 records screened, 43 met the eligibility criteria; 12 were entered into a meta-analysis. Pooled area under the ROC curve (AUC), heterogeneity metrics (I², τ², Q-test), publication bias and sensitivity were assessed. Robustness was examined with a leave-one-out analysis. CONCLUSION: AI models show promising accuracy in predicting colorectal cancer recurrence, supporting their potential utility in clinical decision-making. Nevertheless, further validation in large-scale, prospective studies is required before widespread clinical implementation.

Ähnliche Arbeiten