Particle filter-based parameter estimation algorithm for prognostic risk assessment of progression in non-small cell...

Area of Science:

• Oncology
• Medical Informatics
• Machine Learning

Background:

• Non-small cell lung cancer (NSCLC) poses a significant health threat, with electronic medical records offering potential for risk assessment and recurrence reduction.
• Traditional machine learning models face challenges in improving prediction accuracy and utilizing new patient data effectively as sample sizes grow.

Purpose of the Study:

• To develop a dynamic NSCLC postoperative survival risk assessment model that continuously updates parameters using new patient data.
• To enhance model accuracy and effectively integrate characteristic data from subsequent patients for improved risk prediction.

Main Methods:

• A novel approach combining particle filtering and parameter estimation was employed to build the NSCLC survival risk model.
• Empirical analysis experiments were conducted to demonstrate the feasibility and evaluate the performance of the proposed method.

Main Results:

• The developed model achieved an overall accuracy of 92% and a recall of 71% for deceased patients.
• Compared to traditional machine learning models, the particle filter-based approach improved accuracy by 2% and recall for deceased patients by 11%.

Conclusions:

• The particle filter-based model effectively utilizes subsequent patient data, offering greater relevance for individual patient risk assessment.
• This method supports precision medicine by providing a more accurate and adaptive approach to NSCLC survival risk prediction.