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

Area of Science:

• Cardiovascular diagnostics
• Artificial Intelligence
• Quantum Computing

Background:

• Heart failure is a major global health concern, causing millions of deaths annually.
• Current diagnostic methods lack early detection capabilities and efficient intervention planning.
• Classical machine learning models struggle with complex data and lack interpretability.

Purpose of the Study:

• To develop a novel hybrid classical-quantum neural network for improved cardiovascular diagnostics.
• To address limitations of classical machine learning in handling complex, high-dimensional data.
• To leverage quantum computing for enhanced accuracy and interpretability in heart disease detection.

Main Methods:

• Introduction of the Kolmogorov-Arnold Classical-Quantum Dual-Channel Neural Network (KACQ-DCNN).
• Integration of Kolmogorov-Arnold Network (KAN) components with quantum circuits for learnable activation functions.
• Evaluation using a 4-qubit, 1-layered KACQ-DCNN model against 37 benchmark models.

Main Results:

• The KACQ-DCNN achieved 92.03% accuracy and 94.77% ROC-AUC score, outperforming benchmark models.
• Ablation studies showed a synergistic effect, improving accuracy by approximately 2% compared to Multilayer Perceptron (MLP) variants.
• Demonstrated improved heart disease detection accuracy and interpretable insights via LIME and SHAP.

Conclusions:

• The KACQ-DCNN offers a significant advancement in cardiovascular diagnostics.
• The model provides interpretable insights and robust uncertainty quantification.
• This hybrid approach paves the way for more reliable and transparent clinical decision-making in cardiology.