Answer justification in medical decision support systems based on Bayesian classification.

Area of Science:

• Artificial Intelligence
• Medical Informatics
• Bayesian Statistics

Background:

• Computer programs often lack transparency in their decision-making processes.
• Medical decision support systems require clear explanations for clinical trust and adoption.
• Bayesian classification is a common method for probabilistic reasoning in AI.

Purpose of the Study:

• To develop a novel method for automated answer justification in AI.
• To enable computer-supported decision aids in medicine to explain their conclusions.
• To enhance the interpretability of Bayesian classification models.

Main Methods:

• Analyzing prior and conditional probabilities to identify key factors influencing outcome probabilities.
• Developing an automated approach for answer justification.
• Applying the method to a Bayesian classification-based computer decision aid for stroke classification.

Main Results:

• The proposed method successfully identifies factors driving posterior probability rankings.
• Explanations generated are understandable and clinically plausible.
• Demonstrated utility in a real-world medical decision support scenario (stroke classification).

Conclusions:

• Automated answer justification is feasible and beneficial for medical AI.
• The probability analysis method provides transparent and clinically relevant explanations.
• This approach can improve the reliability and usability of AI-driven medical decision aids.