A Learning Health Care System Using Computer-Aided Diagnosis.

Area of Science:

• Medical Informatics
• Clinical Decision Support
• Artificial Intelligence in Medicine

Background:

• Physicians intuitively use pattern recognition for diagnosis, but flawed probabilistic reasoning leads to diagnostic errors.
• Current computer-aided diagnosis support systems are underutilized and lack pattern recognition and base rate consideration.
• Diagnostic errors are a significant cause of mortality and healthcare costs.

Purpose of the Study:

• To review limitations of current computer-aided diagnosis support systems.
• To propose a conceptual framework for future diagnosis support systems.
• To introduce a novel knowledge representation model for capturing physician expertise.

Main Methods:

• Developing a novel knowledge representation model based on structured patient presentation patterns, including temporal and semantic interrelations.
• Advocating for crowdsourced, deidentified, structured patient pattern collection for distributed knowledge accumulation.
• Creating a self-growing and self-maintaining knowledge base from collective physician experience.

Main Results:

• The proposed model captures complex clinical information beyond simple symptoms and signs.
• Crowdsourced data collection enables continuous knowledge base growth and maintenance.
• The collective pattern map can provide disease base-rate estimates and real-time outbreak surveillance.

Conclusions:

• Future diagnosis support systems should incorporate physician knowledge through structured patient patterns.
• This approach can enhance diagnostic accuracy, provide disease prevalence data, and aid in outbreak detection.
• The system offers benefits for healthcare in resource-limited settings and medical education.