SHAP-Based Feature Augmentation and Stacking Ensemble Learning for ECG-Based Serum Potassium Abnormality Prediction

Area of Science:

• Biomedical Engineering
• Medical Informatics
• Machine Learning in Healthcare

Background:

• Electrocardiogram (ECG) signals are crucial for diagnosing serum potassium abnormalities.
• Restrictions on raw patient data in Taiwan limit medical signal reuse and collaboration.
• Developing alternative features for ECG analysis is essential for broader applications.

Purpose of the Study:

• To transform ECG signal information into high-level SHapley Additive exPlanations (SHAP) features.
• To evaluate SHAP features as auxiliary or alternative inputs for serum potassium abnormality prediction.
• To propose a framework integrating ECG submodels, SHAP features, and ensemble learning.

Main Methods:

• A time-series classification framework was developed using ECG wave-segment submodels.
• SHAP was applied to augment features based on submodel contributions.
• Principal Component Analysis (PCA) features were used for comparison.

Main Results:

• SHAP-based augmented features improved ECG-based serum potassium abnormality prediction.
• Models trained solely on SHAP features achieved performance comparable to or exceeding the baseline.
• SHAP features retained discriminative information, demonstrating potential as alternative features.

Conclusions:

• SHAP features offer a viable method for ECG-based serum potassium abnormality prediction without raw signal access.
• This approach facilitates medical data reuse, cross-institutional collaboration, and privacy risk assessment.
• The findings provide a foundation for future research in privacy-preserving medical AI.