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

Area of Science:

• Urology
• Oncology
• Medical Informatics

Background:

• Predicting recurrence in Ta-stage bladder cancer is critical for patient management.
• Non-muscle invasive bladder cancers (NMIBC) require accurate prognostication for effective follow-up strategies.
• Early identification of recurrence risk in Ta bladder cancer is essential.

Purpose of the Study:

• To compare the predictive performance of machine learning (ML) models against conventional statistical methods.
• To evaluate the accuracy in predicting 2-year postoperative recurrence for Ta stage bladder cancer.
• To identify significant clinical and pathological factors associated with early recurrence.

Main Methods:

• Retrospective analysis of 184 patients with primary Ta bladder cancer between 2018-2021.
• Data included demographic, clinical, imaging, and pathological parameters.
• Prediction models were constructed using Cox-regression, random forest, logistic regression, and k-nearest neighbors (KNN) algorithms.

Main Results:

• Significant predictors for early recurrence included Body Mass Index, American Society of Anesthesiologists (ASA) score, and macroscopic hematuria.
• Conventional Cox-regression model yielded an Area Under the Curve (AUC) of 0.66.
• Machine learning models demonstrated superior performance: Random Forest (AUC=0.75), Logistic Regression (AUC=0.87), and KNN (AUC=0.74).

Conclusions:

• Machine learning models significantly outperform conventional statistical methods in predicting Ta bladder cancer recurrence.
• ML-based prediction tools can enhance the accuracy of recurrence risk assessment in NMIBC.
• The findings support the integration of ML into clinical decision-making for bladder cancer follow-up.