An expert system for assigning patients into clinical trials based on Bayesian networks

Area of Science:

• Medical Informatics
• Artificial Intelligence in Medicine

Background:

• Patient assignment to clinical trials is complex, relying on specific inclusion/exclusion criteria.
• Physician decision-making requires extensive knowledge and data analysis for eligibility determination.

Purpose of the Study:

• To investigate an expert system utilizing Bayesian networks to assist physicians in clinical trial eligibility determination.
• To assess the feasibility and performance of this expert system in real-world clinical protocols.

Main Methods:

• Development of an expert system employing Bayesian networks, a probabilistic approach.
• Implementation and testing of the system across three distinct clinical trial protocols.
• Evaluation of the system's ability to score eligibility with complete and incomplete patient data.

Main Results:

• The expert system demonstrated feasibility across the implemented clinical protocols.
• The system accurately calculated eligibility scores when all patient data was available.
• The system effectively predicted patient eligibility even with missing evidence, guiding physicians to suitable trials.

Conclusions:

• The developed expert system is a feasible tool for streamlining clinical trial patient assignment.
• The system's probabilistic approach and learning capabilities enhance the accuracy and efficiency of eligibility determination.
• This technology can significantly aid physicians in identifying the most appropriate clinical trials for their patients.