Development of a machine learning-based interface for insulin dependency prediction using clinical data.

Area of Science:

• Artificial Intelligence in Medicine
• Clinical Informatics
• Biomedical Data Science

Background:

• Diabetes mellitus presents a significant global health challenge.
• Early identification of insulin dependency is crucial for effective patient management.
• Artificial intelligence offers potential for improving diagnostic accuracy in complex diseases.

Purpose of the Study:

• To develop and evaluate an AI-based diagnostic system for identifying insulin dependency in diabetes mellitus.
• To compare the performance of various machine learning models using real-world clinical data.
• To identify key predictive features for early diabetes assessment.

Main Methods:

• Utilized a real-world clinical dataset of 100 anonymized patient records.
• Preprocessed data including handling missing values and feature encoding.
• Implemented and evaluated Logistic Regression, Random Forest, XGBoost, and LightGBM models using 5-fold cross-validation.
• Assessed model performance using accuracy, precision, recall, and F1-score.

Main Results:

• XGBoost demonstrated superior performance with an accuracy of 0.88, precision of 0.86, recall of 0.90, and F1-score of 0.88.
• LightGBM also showed strong results (accuracy 0.85, F1-score 0.84).
• Postprandial blood sugar (PPBS) and glycated hemoglobin (HbA1c) were identified as the most predictive features.

Conclusions:

• The developed AI system, particularly the XGBoost model, shows promise for early identification of insulin dependency in diabetes.
• Preliminary findings suggest AI can aid in timely diabetes intervention.
• Further validation on larger, multi-site cohorts is necessary before clinical implementation.