Developing an Explainable Artificial Intelligence System for the Mobile-Based Diagnosis of Febrile Diseases Using...

Area of Science:

• Artificial Intelligence in Healthcare
• Medical Diagnostics
• Explainable AI (XAI)

Background:

• Machine learning models in healthcare often lack transparency, hindering trust and adoption in critical decision-making.
• Mobile health platforms offer potential for widespread diagnostic accessibility but require interpretable AI for clinical utility.

Purpose of the Study:

• To propose and evaluate a mobile-based explainable artificial intelligence (XAI) platform for diagnosing febrile illnesses.
• To integrate local interpretable model-agnostic explanations (LIME) and generative pre-trained transformers (GPT) for enhanced model transparency and user comprehension.

Main Methods:

• Developed a mobile AI system using random forest for disease diagnosis.
• Integrated LIME for interpreting model predictions and GPT-3.5 for generating natural language explanations.
• Evaluated diagnostic performance for malaria, urinary tract infections, respiratory tract infections, typhoid fever, and HIV/AIDS.

Main Results:

• The model achieved high performance for malaria (85% precision, 91% recall, 88% F1-score).
• Moderate performance was observed for UTIs and RTIs, while typhoid fever and HIV/AIDS detection showed limitations requiring further fine-tuning.
• LIME identified key symptoms influencing diagnoses, and GPT-3.5 provided clear explanations, improving system trustworthiness.

Conclusions:

• The mobile XAI platform enhances transparency and trustworthiness in febrile illness diagnosis.
• Integration of LIME and GPT-3.5 aids clinical decision-making and improves user comprehension.
• The system shows promise for improving patient outcomes and reducing healthcare burden through accessible, explainable diagnostics.