Machine learning for predicting colon cancer recurrence.

Area of Science:

• Oncology
• Data Science
• Medical Informatics

Background:

• Colorectal cancer (CRC) is a major global health issue with high recurrence rates.
• Early detection and intervention are crucial for managing CRC recurrence.
• Machine learning (ML) offers advanced tools for data-driven insights in cancer care.

Purpose of the Study:

• To develop and evaluate ML algorithms for predicting colon cancer (CC) recurrence.
• To identify key demographic, clinicopathological, and laboratory factors associated with CC recurrence.
• To enhance personalized risk stratification for CC patients.

Main Methods:

• Retrospective analysis of 396 colon cancer patients (2010-2021).
• Application of ML algorithms to predict CC recurrence.
• Evaluation using AUC, accuracy, recall, precision, and F1 scores.

Main Results:

• Identified significant risk factors: sex, CEA, tumor location/depth, invasion, and lymph node status.
• CatBoost Classifier achieved 0.92 AUC and 88% accuracy.
• Key predictors included CEA, albumin, N stage, weight, and blood counts.

Conclusions:

• ML integration enables personalized risk stratification and improved clinical decisions for CC.
• Early identification of high-risk patients can lead to more effective treatments.
• This study highlights ML's transformative potential in precision oncology for CC management.