Deep learning-based LDL-C level prediction and explainable AI interpretation.

Area of Science:

• Biomedical Informatics
• Artificial Intelligence in Healthcare
• Cardiovascular Disease Risk Assessment

Background:

• Accurate prediction of low-density lipoprotein cholesterol (LDL-C) is crucial for cardiovascular disease risk assessment.
• Traditional formulas for LDL-C estimation have limitations in accuracy.
• Deep learning (DL) models offer potential for improved predictive capabilities.

Purpose of the Study:

• To investigate the efficacy of deep learning models for predicting LDL-C levels.
• To compare the performance of DL models against traditional LDL-C formulas and machine learning models.
• To explore the interpretability of DL model predictions using LIME.

Main Methods:

• Utilized deep learning models including Convolutional Neural Networks (CNN), Recurrent Neural Networks (RNN), and Long Short-Term Memory (LSTM).
• Employed a dataset from New York-Presbyterian Hospital/Weill Cornell Medical Center, including triglycerides (TG), total cholesterol (TC), and high-density lipoprotein cholesterol (HDL-C).
• Compared DL model performance with traditional methods (Sampson, Martin equations) and traditional machine learning (ML) models. Used Local Interpretable Model-Agnostic Explanations (LIME) for model interpretability.

Main Results:

• Deep learning models demonstrated higher accuracy in LDL-C prediction compared to traditional formulas.
• RNN and LSTM models exhibited superior performance over other DL models and traditional equations.
• DL models provided results closer to ML models, indicating strong predictive power.
• LIME analysis offered insights into DL model decisions, though requiring more computational effort than ML models.

Conclusions:

• Deep learning models are more effective than traditional methods for predicting LDL-C levels.
• The study highlights the potential of RNN and LSTM models for accurate LDL-C estimation in clinical settings.
• Findings support the use of DL models for improved cardiovascular disease risk assessment and treatment planning.