An Approach of Integrating Domain Knowledge into Data-Driven Diagnostic Model

Area of Science:

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

Background:

• General practitioners face challenges in accurate and efficient diagnosis of common diseases.
• Reducing diagnostic errors and physician workload is crucial for improving healthcare quality.
• Existing diagnostic tools may lack the ability to integrate comprehensive domain knowledge.

Purpose of the Study:

• To develop and evaluate a neural network model for classifying potential diagnoses from ambulatory healthcare data.
• To introduce a novel method for integrating domain knowledge into neural network training for medical diagnosis.
• To assess the model's performance against a baseline in terms of knowledge consistency and generalization.

Main Methods:

• Development of a neural network model for disease classification.
• Utilizing ambulatory healthcare data for model training and evaluation.
• Implementation of a novel approach for incorporating domain knowledge into the neural network architecture and training process.

Main Results:

• The developed neural network model demonstrated superior performance compared to the baseline model.
• The model achieved higher knowledge consistency, aligning diagnostic outputs with established medical expertise.
• Enhanced model generalization was observed, indicating robustness across different data variations.

Conclusions:

• The proposed neural network model offers a promising tool for assisting general practitioners in diagnosing common diseases.
• Integrating domain knowledge into AI models can significantly improve diagnostic accuracy and reliability.
• This approach has the potential to reduce misdiagnoses and alleviate workload in primary care settings.