A hybrid quantum algorithm for multi-resolution arrhythmia classification

Hybrid quantum algorithms for ECG image-based arrhythmia detection are increasingly used, but their performance under lower resolutions and varied training initialisations remains unclear. This study evaluates a hybrid quantum algorithm combining a fully fine-tuned ResNet50 backbone with a 10-qubit variational quantum circuit for six-class ECG arrhythmia classification. The algorithm was assessed on 24,799 held-out ECG images at controlled content resolutions of 80, 112, 160, 224 and 320 pixels and compared with a conventional ResNet50 classifier and a closely parameter-matched classical 10-dimensional bottleneck. At 224 and 320 pixels, all three approaches reached ceiling-level classification performance. Differences became more apparent as spatial information was reduced. At 112 pixels, evaluation across three independent training seeds produced a mean accuracy of 0.9690 ± 0.0124 for the hybrid quantum algorithm, compared with 0.9396 ± 0.0044 for ResNet50 and 0.9580 ± 0.0354 for the classical bottleneck. Corresponding macro-F1 scores were 0.8638 ± 0.0417, 0.8235 ± 0.0062 and 0.8783 ± 0.1033, respectively. Although the hybrid algorithm achieved the highest mean accuracy, the relative performance of the hybrid and classical bottleneck architectures varied across training seeds, indicating sensitivity to optimisation randomness. At 160 pixels, the architectures approached ceiling performance, while at 80 pixels the hybrid algorithm and ResNet50 showed comparable classification behaviour. These findings indicate that the quantum pathway does not provide a consistent classification advantage across resolutions or random initialisations, but remains competitive under moderate spatial-information reduction. Without patient identifiers or external validation, results reflect image-level benchmark performance. Broader generalisability requires patient-independent and external validation.