An explainable Artificial Intelligence software system for predicting diabetes.

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Healthcare
• Diabetes Technology

Background:

• Minimally invasive blood glucose monitoring causes discomfort.
• Effective diabetes surveillance is crucial for managing medical expenses.
• Existing methods require blood extraction, posing challenges for continuous monitoring.

Purpose of the Study:

• To develop an Explainable Artificial Intelligence (XAI) system for non-invasive glucose level analysis.
• To create an intelligible machine learning model for predicting and explaining glucose level outcomes.
• To improve diabetes surveillance through a comfortable and accurate monitoring system.

Main Methods:

• Utilized Bi-directional Long Short-Term Memory (Bi-LSTM) and Convolutional Neural Network (CNN) for abnormal glucose level analysis.
• Acquired glucose level data via glucose oxidase (GOD) strips placed on the body.
• Converted signal data to spectrogram images for classification (low, average, abnormal glucose).
• Trained an individualized monitoring model using labeled spectrogram images and an XAI-driven architecture.

Main Results:

• The proposed XAI model effectively tracks real-time glucose levels.
• Performance evaluation using confusion matrix metrics demonstrated high accuracy.
• The model successfully identified individuals with elevated glucose levels.

Conclusions:

• The developed XAI system offers a promising non-invasive approach for diabetes surveillance.
• The integration of Bi-LSTM, CNN, and XAI enhances the intelligibility and accuracy of glucose monitoring.
• This technology has the potential to significantly improve the management of diabetes and reduce healthcare costs.