Leveraging a hybrid convolutional gated recursive diabetes prediction and severity grading model through a mobile app

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare
• Computational Biology

Background:

• Diabetes mellitus presents significant morbidity and mortality, necessitating early detection to prevent complications.
• Existing machine learning models for diabetes prediction face challenges with accuracy, reliability, and data imbalance.
• Dependable diabetes prediction models are crucial for timely intervention and improved patient outcomes.

Purpose of the Study:

• To introduce a novel deep learning mechanism, the convolutional gated recurrent unit (CGRU), for accurate diabetes detection and severity assessment.
• To address the limitations of current machine learning models in predicting diabetes.
• To enhance the extraction of temporal and spatial characteristics from data for improved prediction accuracy.

Main Methods:

• A deep learning framework utilizing the convolutional gated recurrent unit (CGRU) was developed for diabetes prediction.
• Data preprocessing involved imputation, feature removal, and normalization on the BRFSS dataset.
• Clustering algorithms were employed to classify diabetes severity levels based on patient characteristics and identified patterns.

Main Results:

• The proposed CGRU model achieved a high accuracy rate of 99.9% in diabetes prediction.
• The CGRU framework demonstrated superior performance compared to state-of-the-art approaches like Attention-based CNN and Ensemble ML models.
• Clustering algorithms proved beneficial in identifying subtle patterns for more accurate and reliable severity level classification.

Conclusions:

• The CGRU model significantly enhances early diabetes detection and diagnosis, leading to better healthcare outcomes.
• The proposed deep learning approach offers a more accurate and reliable method for diabetes prediction and severity assessment.
• Future research should focus on validating the CGRU model on diverse datasets to further establish its generalizability.