Intra-person multi-task learning method for chronic-disease prediction.

Area of Science:

• Medical Informatics
• Machine Learning in Healthcare
• Computational Medicine

Background:

• Clinical data accumulation supports personalized medicine and diagnosis.
• Chronic diseases share characteristics, enabling prediction of multiple conditions from patient data.

Purpose of the Study:

• To propose an intra-person multi-task learning framework for jointly predicting correlated chronic diseases.
• To improve model performance and stability in chronic disease prediction.

Main Methods:

• Data preprocessing and feature selection using bidirectional recurrent imputation for time series (BRITS) and LASSO.
• Development of single-task Convolutional Neural Network-Long Short-Term Memory (CNN-LSTM) models.
• Implementation of a novel intra-person multi-task learning CNN-LSTM framework for simultaneous prediction of multiple chronic diseases.

Main Results:

• The multi-task learning framework demonstrated superior stability and accuracy compared to single-task models and baseline recurrent networks.
• The model's flexibility and generalization were validated across different time steps.

Conclusions:

• Intra-person multi-task learning offers a more robust approach for predicting multiple correlated chronic diseases.
• The proposed CNN-LSTM framework advances the potential for accurate, personalized chronic disease management.