Optimization of Cervical Cancer Screening: A Stacking-Integrated Machine Learning Algorithm Based on Demographic,...

Area of Science:

• Oncology
• Machine Learning
• Public Health

Background:

• Current cervical cancer risk prediction models rely on clinical indicators, limiting their application.
• There is a need for accessible methods to identify high-risk individuals for optimized screening.
• Stacking-integrated machine learning (SIML) offers enhanced predictive performance by combining multiple algorithms.

Purpose of the Study:

• To develop and validate a SIML model for identifying women at high risk of cervical cancer.
• To utilize demographic, behavioral, and historical clinical factors for risk prediction.
• To improve the efficiency of cervical screening strategies and medical resource allocation.

Main Methods:

• A SIML algorithm was developed using screening data from 858 women.
• Data were split into training (80%) and testing (20%) sets for model development and validation.
• Random forest and logistic regression identified predictive features; 12 ML algorithms were compared, with the best forming the SIML model.

Main Results:

• The random forest model identified 18 predictive features, with hormonal contraceptive use, pregnancies, smoking years, and sexual partners being most significant.
• The SIML algorithm demonstrated superior performance with an Area Under the Curve (AUC) of 0.877, sensitivity of 81.8%, and specificity of 81.9%.
• SIML outperformed other tested machine learning methods in predicting cervical cancer risk.

Conclusions:

• The developed SIML model accurately identifies women at high risk for cervical cancer.
• This predictive tool enables personalized cervical screening programs by optimizing intervals and care plans.
• The model effectively leverages demographic, behavioral, and clinical data for risk stratification.