Abstract Recent studies have reported competitive predictive performance from hybrid quantum--classical learning models across biomedical and healthcare tasks, motivating growing interest in Quantum Machine Learning (QML) for medical data. However, predictive performance does not explain how quantum models arrive at their decisions. The combination of learned representations, parameterized quantum circuits, entangling operations, and noise introduce immense challenge for obtaining human-interpretable and circuit-level explanations. We present \textbf{FQXAI}, a unified semantic and circuit level explainability framework for hybrid quantum--classical medical learning. FQXAI transforms heterogeneous biomedical inputs into an eight-dimensional proxy supervised semantic concept space, mapping these representations into differentiable Gaussian fuzzy linguistic states, and generates model-level IF--THEN rationales. Exact Shapley values are computed over all
28=256 semantic coalitions, enabling exhaustive attribution without Monte Carlo sampling. At the circuit level, \textbf{SVQX} quantifies the contributions of CNOT operations under controlled depolarizing noise and identifies negatively contributing gates as pruning candidates. A Jacobian-based projection further transfers semantic attributions to the original ECG and imaging domains. Experiments on MIT-BIH Arrhythmia, PneumoniaMNIST, and PathMNIST demonstrate the framework across temporal, radiological, and histopathological modalities. The models achieved accuracies of 0.890, 0.768, and 0.726, while exact FQ-SHAP achieved AUDC values of 0.780, 0.786, and 0.423, respectively. Under 5\% simulated depolarizing noise, SVQX-guided pruning reduced CNOT counts from 14 to 11, 12, and 10, with corresponding accuracy increases of 2.20, 1.74, and 1.51 percentage points. These results demonstrate the feasibility and efficacy of coupling semantic explanations with noise-aware circuit analysis for auditable hybrid quantum medical learning.
Deep learning vision systems excel at pattern recognition yet falter when inputs are noisy or the model must explain its own confidence. Fuzzy inference, with its graded memberships and rule transparency, offers a remedy, while parameterized quantum circuits can embed features in richly entangled Hilbert spaces with striking parameter efficiency. Bridging these ideas, this study introduces a innovative Highly Quantized Fuzzy Neural Network (HQFNN) that realises the entire fuzzy pipeline inside a shallow quantum circuit and couples the resulting quantum signal to a lightweight CNN feature extractor. Each image feature is first mapped to a single qubit membership state through repeated angle reuploading. Then a compact rule layer refines these amplitudes, and a clustered CNOT defuzzifier collapses them into one crisp value that is fused with classical features before classification. Evaluated on standard image benchmarks, HQFNN consistently surpasses classical, fuzzy enhanced and quantum only baselines while using several orders of magnitude fewer trainable weights, and its accuracy degrades only marginally under simulated depolarizing and amplitude damping noise, evidence of intrinsic robustness. Gate count analysis further shows that circuit depth grows sublinearly with input dimension, confirming the model's practicality for larger images. These results position the model as a compact, interpretable and noise tolerant alternative to conventional vision backbones and provide a template for future quantum native fuzzy learning frameworks.