Application of the Naïve Bayesian Classifier to optimize treatment decisions

Area of Science:

• Oncology
• Medical Informatics
• Machine Learning

Background:

• Radiotherapy (RT) is a common treatment for brain tumors.
• Accurate prediction of cancer relapse or progression is crucial for patient management.
• Existing methods for risk assessment may benefit from advanced computational tools.

Purpose of the Study:

• To evaluate the accuracy, specificity, and sensitivity of the Naïve Bayesian Classifier (NBC) for predicting individual cancer relapse risk.
• To assess the utility of NBC in patients undergoing postoperative radiotherapy for brain tumors.

Main Methods:

• Analysis of data from 142 brain tumor patients treated with RT (2000-2005).
• Utilized 96 binary attributes related to disease, patient, and treatment for NBC model training.
• Validated classifier performance using leave-one-out and cross-validation, and a clinical test with 35 patients.

Main Results:

• Achieved high classification accuracy (84%), specificity (0.87), and sensitivity (0.80).
• NBC demonstrated consistent performance in both training and clinical evaluation phases.
• The classifier effectively calculated individual probabilities for relapse or progression.

Conclusions:

• The Naïve Bayesian Classifier (NBC) is a valuable tool for supporting risk assessment in brain tumor patients post-radiotherapy.
• NBC can aid clinicians in identifying patients at higher risk of relapse or progression.
• This predictive capability can inform personalized treatment decisions and improve patient outcomes.