Artificial intelligence predicts the progression of diabetic kidney disease using big data machine learning

Area of Science:

• Medical Informatics
• Nephrology
• Machine Learning

Background:

• Diabetic kidney disease (DKD) is a major complication of diabetes.
• Early detection of DKD progression is crucial for timely intervention.
• Existing methods for predicting DKD aggravation have limitations.

Purpose of the Study:

• To develop and validate an AI-powered predictive model for diabetic kidney disease (DKD) aggravation.
• To utilize natural language processing and big data machine learning on electronic medical records (EMR).
• To assess the model's accuracy in predicting DKD progression over a 6-month period.

Main Methods:

• A predictive model was constructed using AI and big data machine learning on EMR data from 64,059 diabetes patients.
• AI extracted features from the past 6 months and identified 24 key factors using a convolutional autoencoder.
• A predictive model incorporating 3,073 features, including time-series data, was built using logistic regression analysis.

Main Results:

• The AI model achieved 71% accuracy in predicting DKD aggravation.
• A significant increase in hemodialysis incidence was observed in the DKD aggravation group over 10 years.
• The model demonstrated the ability to detect DKD progression.

Conclusions:

• AI can effectively support clinical judgment in predicting diabetic kidney disease (DKD) aggravation.
• The developed AI model shows promise for early detection and intervention in DKD.
• This approach may contribute to reducing the incidence of hemodialysis in diabetic patients.