KACQ-DCNN: Uncertainty-Aware Interpretable Kolmogorov-Arnold Classical-Quantum Dual-Channel Neural Network for Heart Disease Detection.

Background: Heart failure remains a critical global health issue, contributing significantly to cardiovascular disease burden and accounting for approximately 17.8 million annual deaths worldwide. Traditional diagnostic approaches face substantial limitations in early detection and intervention planning.

Problem: Classical machine learning models struggle with complex, high-dimensional data, class imbalances, poor categorical feature representations, and lack interpretability due to their 'black box' nature. While quantum machine learning shows potential, existing hybrid models have yet to capitalize on quantum advantages for cardiovascular diagnostics fully.

Solution: We propose the Kolmogorov-Arnold Classical-Quantum Dual-Channel Neural Network (KACQ-DCNN), a novel hybrid architecture that integrates KAN components with quantum circuits, enabling univariate learnable activation functions that significantly improve function approximation with reduced complexity. Our comprehensive evaluation demonstrates that the KACQ-DCNN 4-qubit 1-layered model outperforms 37 benchmark models with 92.03% accuracy and 94.77% ROC-AUC score. Ablation studies confirm a synergistic effect of classical-quantum components with KAN, improving accuracy by approximately 2% compared to MLP variants.

Benefits: KACQ-DCNN improves heart disease detection accuracy while providing interpretable insights through model-agnostic explainability techniques like Local Interpretable Model-agnostic Explanations (LIME) and Shapley Additive exPlanations (SHAP). Additionally, conformal prediction techniques deliver robust uncertainty quantification, advancing the field toward more reliable, transparent, and clinically trustworthy cardiovascular diagnostic systems that facilitate timely and effective interventions.