A Machine Learning and Bayesian Belief Network Approach to Predicting Cervical Cancer Risk: Implications for Risk...

Area of Science:

• Computational Oncology
• Medical Informatics
• Digital Health

Background:

• Cervical cancer poses a significant global health burden, requiring improved risk stratification and early detection methods.
• Existing predictive models often struggle with data complexities like missing values and class imbalance.

Purpose of the Study:

• To develop and validate a comprehensive predictive framework for cervical cancer using machine learning (ML) and Bayesian Belief Networks (BBNs).
• To enhance early detection and risk stratification for cervical cancer through advanced computational approaches.

Main Methods:

• Analysis of a cohort of 858 patients.
• Integration of data science techniques including multiple imputation, feature selection, and imbalance mitigation.
• Application of ML algorithms and BBNs, specifically Bayesian Additive Regression Trees (BART).

Main Results:

• The combined ML model achieved 95.6% accuracy, 0.958 AUROC, and 0.945 F1-score.
• The BBN model demonstrated 91.3% sensitivity and 86.8% specificity.
• The framework showed high predictive performance across various cervical cancer screening tests.

Conclusions:

• The proposed framework demonstrates technical efficacy for cervical cancer prediction.
• The study highlights the potential for integrating AI and ML into clinical decision-support systems.
• Interdisciplinary collaboration is crucial for developing effective AI healthcare solutions and advancing precision medicine.