Chronic liver disease classification using deep learning with SHAP-optimized hybrid features.

Area of Science:

• Hepatology
• Medical Informatics
• Artificial Intelligence in Medicine

Background:

• Liver disorders present significant health risks, necessitating timely and accurate diagnosis for effective treatment.
• Machine learning (ML), especially deep learning, shows promise in developing advanced diagnostic tools for disease prediction.
• Current diagnostic methods can be improved with more accurate and efficient predictive capabilities.

Purpose of the Study:

• To develop and evaluate a novel predictive framework for accurate liver disease detection using deep neural networks (DNNs).
• To integrate feature ranking and projection-based algorithms within the DNN framework to enhance predictive performance.
• To improve model interpretability by applying SHapley Additive exPlanations (SHAP) to identify key predictive features.

Main Methods:

• A deep neural network (DNN) model was developed for liver disease prediction.
• Feature ranking and projection-based algorithms were integrated into the DNN framework.
• SHapley Additive exPlanations (SHAP) were employed for model interpretability and feature importance analysis.
• The model's performance was validated using 10-fold cross-validation.

Main Results:

• The proposed DNN model achieved an average accuracy (ACC) of 92.50% under 10-fold cross-validation.
• The DNN model demonstrated superior performance compared to traditional ML algorithms and existing state-of-the-art methods.
• SHAP analysis successfully identified influential features impacting liver disease prediction.

Conclusions:

• The novel DNN framework offers a highly accurate and interpretable approach for liver disease detection.
• The findings suggest significant potential for improving diagnostic accuracy and supporting early intervention strategies.
• This ML-based approach can enhance patient outcomes through more effective liver disorder management.