Machine learning methods for predicting early recurrence in Ta stage bladder cancer and comparison with conventional statistical methods

Comprehensive Summary

This study compared conventional statistical models with machine learning (ML) approaches to predict early recurrence within two years in patients with Ta-stage non-muscle invasive bladder cancer. Researchers retrospectively analyzed clinical, demographic, imaging, and pathological data from patients treated between 2018 and 2021, dividing them into recurrence and non-recurrence groups. Body mass index, ASA score, and the presence of visible blood in urine were identified as significant predictors of early recurrence. The conventional Cox-regression model achieved an AUC of 0.66, while ML models performed better, with AUCs of 0.75 (random forest), 0.87 (logistic regression), and 0.74 (k-nearest neighbors). The results suggest that machine learning models may provide more accurate predictions of early recurrence than traditional statistical methods.

Outcomes and Implications

These findings indicate that machine learning tools could improve risk stratification and follow-up planning for patients with early-stage bladder cancer. More accurate predictions may allow clinicians to personalize surveillance schedules and treatment decisions, potentially improving outcomes and reducing unnecessary procedures.