This paper presents a systematic ablation and evaluation integrity study of hybrid quantum-classical neural networks for seven-class skin lesion classification on the HAM10000 dermoscopy dataset. We study the effect of the evaluation protocol on a hybrid quantum-classical pipeline by comparing two partitioning schemes on the same model and data. The first is a leakage-prone scheme that partitions the pool of augmented images, and the second is a verified scheme that partitions original lesion identities. Under the leakage-prone scheme, augmented variants of the same physical lesion appear on both sides of the split. An audit found that 1,095 of 4,810 lesions (22.8%) crossed the split boundary, artificially inflating all previously reported metrics. We rebuild the pipeline with a verified lesion-level partition consisting of 4,152 training, 522 validation, and 517 original test images. The training set is augmented to 10,503 images while ensuring zero image-level and lesion-level overlap. We then re-run every experiment across five random seeds with per-class Wilson confidence intervals and paired significance tests on a fixed held-out test set that is never used for tuning. Under the corrected protocol, the strongest configuration features a ResNet50 backbone with concatenation fusion, focal loss, and test-time augmentation. This setup reaches 78.41% $\pm ~0.74$ test accuracy and a macro F1 of 0.804 against a verified baseline of 76.17% $\pm ~1.91$ . However, under the leakage-prone protocol, this exact same pipeline yields 94.57%. A parameter-matched classical control, in which a 36-parameter multilayer perceptron replaces the variational quantum circuit (VQC), is statistically indistinguishable from the hybrid model (78.14% versus 78.41%, p = 0.81). Additionally, a controlled fusion comparison shows that concatenation is not significantly better than weighted fusion. We therefore report no detectable advantage from the quantum branch at this scale as a controlled negative result rather than a failure, and we characterize VQC trainability under noiseless and depolarizing simulation. Finally, a split-audit survey of recent quantum-hybrid HAM10000 studies indicates that reported accuracies above 90% are associated with undisclosed or leakage-prone evaluation protocols. The contribution of this work is methodological: a reproducible leakage audit, a verified correction protocol, an honest multi-seed baseline, a parameter-matched classical control, and evidence that evaluation protocols rather than architectures drive the headline differences reported in this literature.
This paper presents a systematic ablation and evaluation integrity study of hybrid quantum-classical neural networks for seven-class skin lesion classification on the HAM10000 dermoscopy dataset. We study the effect of the evaluation protocol on a hybrid quantum-classical pipeline by comparing two partitioning schemes on the same model and data. The first is a leakage-prone scheme that partitions the pool of augmented images, and the second is a verified scheme that partitions original lesion identities. Under the leakage-prone scheme, augmented variants of the same physical lesion appear on both sides of the split. An audit found that 1,095 of 4,810 lesions (22.8%) crossed the split boundary, artificially inflating all previously reported metrics. We rebuild the pipeline with a verified lesion-level partition consisting of 4,152 training, 522 validation, and 517 original test images. The training set is augmented to 10,503 images while ensuring zero image-level and lesion-level overlap. We then re-run every experiment across five random seeds with per-class Wilson confidence intervals and paired significance tests on a fixed held-out test set that is never used for tuning. Under the corrected protocol, the strongest configuration features a ResNet50 backbone with concatenation fusion, focal loss, and test-time augmentation. This setup reaches 78.41% ±0.74 test accuracy and a macro F1 of 0.804 against a verified baseline of 76.17% ±1.91 . However, under the leakage-prone protocol, this exact same pipeline yields 94.57%. A parameter-matched classical control, in which a 36-parameter multilayer perceptron replaces the variational quantum circuit (VQC), is statistically indistinguishable from the hybrid model (78.14% versus 78.41%, p = 0.81). Additionally, a controlled fusion comparison shows that concatenation is not significantly better than weighted fusion. We therefore report no detectable advantage from the quantum branch at this scale as a controlled negative result rather than a failure, and we characterize VQC trainability under noiseless and depolarizing simulation. Finally, a split-audit survey of recent quantum-hybrid HAM10000 studies indicates that reported accuracies above 90% are associated with undisclosed or leakage-prone evaluation protocols. The contribution of this work is methodological: a reproducible leakage audit, a verified correction protocol, an honest multi-seed baseline, a parameter-matched classical control, and evidence that evaluation protocols rather than architectures drive the headline differences reported in this literature.
Automated analysis of dermoscopic images can support benign-versus-malignant skin lesion classification, but the practical usefulness of hybrid quantum-classical models remains unclear under current hardware constraints. This paper presents an experimental comparison between a classical MobileNetV2-based baseline and a hybrid pipeline that combines convolutional feature extraction with a variational quantum circuit for binary classification on the publicly available HAM10000 dataset. In the hybrid approach, image features extracted by the classical backbone are projected to a low-dimensional latent space and encoded into a trainable quantum circuit. Both models are evaluated on disjoint train, validation, and test splits, with model selection for the quantum pipeline performed on the validation set. The classical baseline clearly outperformed the hybrid model. The selected hybrid configuration reached a balanced accuracy of 0.582 and an area under the receiver operating characteristic curve (AUC) of 0.610 on the held-out test set. We also validated inference on IBM Quantum backends, confirming compatibility with execution on real quantum hardware. Taken together, these results suggest that the hybrid pipeline is better viewed as a feasibility study under current noisy intermediate-scale quantum (NISQ) limitations rather than as a replacement for strong classical image classifiers.
Abstract Accurate skin lesion classification in real-world deployment remains difficult when models are trained on a single dataset and must generalize across heterogeneous devices and patient cohorts. Quantum computing can enrich representation learning via variational transformations in high-dimensional Hilbert spaces, yet clinically reliable and domain-robust hybrid quantum dermoscopy remains insufficiently explored. A shift-consistent and explanation-stable distributed hybrid quantum network (SCES-QNet) was developed to examine whether source-only cross-domain generalization, calibration, and attribution consistency could be improved within a unified dermoscopic classification framework. Distributed quantum branches were combined with prediction-consistency and attribution-equivariance regularization under strict source-only training. HAM10000 was used for model development, whereas ISIC 2019 and PAD-UFES-20 were reserved for external evaluation after class-taxonomy harmonization. Discrimination, calibration, and explanation-stability measures were assessed over five independently initialized runs. On HAM10000, near-ceiling performance was achieved (Macro-F1 = 98.4%, Balanced Accuracy = 98.6%, macro-AUROC = 0.998). Under source-only transfer, strong retention was maintained on ISIC 2019 and PAD-UFES-20 (Macro-F1 = 96.2% / 94.8%; macro-AUROC = 0.993 / 0.989). Calibration degradation remained controlled (ECE: 0.007 → 0.014 → 0.017), and explanation stability remained high across domains (ESS ≥ 0.872). Under the evaluated protocol, SCES-QNet was associated with strong discrimination, controlled calibration drift, and comparatively stable explanations across the selected datasets. The findings provide preliminary evidence for stability-aware hybrid quantum learning in dermoscopic classification; however, broader multi-center evaluation, verified cross-archive duplicate auditing, and execution on physical NISQ hardware remain necessary before clinical or hardware-level conclusions can be drawn.
This study presents a reproducible evaluation framework for hybrid quantum-classical neural networks (HQCNNs) in healthcare classification, rather than a new architecture. We assess a four-qubit HQCNN combining a compact classical encoder, a two-layer parameterized quantum circuit (PQC), and a classical readout (441 trainable parameters) against carefully tuned classical baselines on the Wisconsin Diagnostic Breast Cancer (WDBC) dataset under identical five-fold cross-validation. The work is framed as a single-dataset proof-of-concept: the contribution is a documented, shared-fold evaluation protocol with a parameter-matched classical control and a quantified epistemic-informativeness analysis, not a demonstration of general quantum advantage. The HQCNN reached 96.49±1.96% accuracy and 99.44±0.60% ROC-AUC. A parameter-matched classical multilayer perceptron (441 parameters) reached 95.08±1.81% accuracy; the HQCNN’s +1.41 percentage-point edge at equal capacity was not statistically significant (paired t, p=0.056). Across five shared folds, no HQCNN-versus-classical accuracy difference survived Holm–Bonferroni correction (all adjusted p≥0.625), so we report the HQCNN as competitive with, not superior to, strong tuned classical baselines. A multi-split depth ablation showed that circuit depth L∈{1,2,3} had no statistically detectable effect on accuracy (L=2 vs. L=3: Wilcoxon p=1.00); we therefore adopt two variational layers as a practical default rather than an optimum. Under a low-noise simulator (depolarising and amplitude-damping channels, p=0.01), accuracy was 96.49%, indicating robustness only at modest uniform error rates; realistic hardware noise is higher. We additionally apply Bayesian surprise as an epistemic-informativeness heuristic—not a formal generative model—to rank which findings are most worth building on. The framework offers a reproducible, documented evaluation procedure that can support cumulative comparison of hybrid quantum-classical models in healthcare.
Skin lesion classification from dermoscopic images remains difficult because diagnostically important lesions may share similar colours, textures, and structural patterns. Fine local details and broader lesion characteristics must therefore be considered together to achieve reliable multi-class classification. This study presents a noise-aware and statistically evaluated dual-branch hybrid quantum-classical framework that combines EfficientNet-B3 features, handcrafted dermatological descriptors, and parameterized quantum circuits. A 6-qubit branch processes micro-level textural information, while a 4-qubit branch represents broader structural characteristics. The resulting quantum features are combined with deep visual representations and handcrafted descriptors based on texture, colour, and lesion complexity. The framework was evaluated using stratified five-fold cross-validation and repeated experiments with different random initialization seeds. On HAM10000, it achieved a mean accuracy of 95.18% ±0.75 %, a mean balanced accuracy of 92.92% ±0.89 %, and a mean macro-F1 score of 0.904±0.009 . A separate training and evaluation experiment on ISIC-2019 produced an accuracy of 93.86% and a macro-F1 score of 0.913, indicating consistent performance across two independently evaluated dermoscopic datasets. The effect of non-ideal quantum operation was also examined by introducing depolarizing noise during evaluation. The model retained accuracies of 94.13% at p=0.01 and 93.21% at p=0.05 . These findings show that combining deep features, dermatological descriptors, and dual-scale quantum representations can improve the discrimination of visually similar lesion categories, including melanoma. The proposed framework provides a statistically reliable and noise-resilient approach for computer-aided dermoscopic image classification, although further external and clinical validation is required before practical clinical use.
Background/Objectives: Hybrid quantum–classical neural networks are reported to classify brain tumors on magnetic resonance imaging (MRI) accurately, but these reports rarely control for data leakage or test whether the quantum circuit contributes. We examined both issues. Methods: We audited version 2 of the Kaggle Brain Tumor MRI Dataset (7200 images) with mirror- and rotation-invariant duplicate detection and by matching every image to its source collections, which recovered patient identities for 4846 tumor images. A variational quantum circuit (VQC) was compared with size-matched, deeper and Fourier-surrogate classical controls using frozen features, end-to-end fine-tuning and training from scratch, all under patient-grouped five-fold cross-validation. Results: Of the 1600 test images, 1269 (79%) had a copy in the training partition, and 915 of the 917 figshare-derived test images shared a patient with it. On frozen features, a six-qubit VQC head was 6.0 percentage points (pp) less accurate than its classical control (95% confidence interval −6.8 to −5.2) and less accurate than a classical surrogate of its own function class. With fine-tuning, it stayed within about 1 pp of its control but was 0.7 pp less accurate (−1.2 to −0.1) after convergence-based training; a hybrid network trained from scratch was 8.3 pp less accurate than its control. Patient-grouped evaluation lowered accuracy by 3–8 pp relative to our earlier duplicate-grouped evaluation, and 76 glioma test images from another source remained poorly recognized. Conclusions: The quantum layer added cost but no accuracy, and leakage and provenance mattered more than the classifier head.
Abstract Quantum machine learning (QML) has recently been explored as a means of augmenting classical convolutional neural networks (CNNs) for medical image analysis, yet most existing hybrid architectures insert a single quantum component either at the input or the output of an otherwise classical pipeline, and few works empirically isolate which placement is actually responsible for performance change. This paper proposes QCQ-CNN, a Quantum-Classical-Quantum hybrid framework for brain tumor magnetic resonance imaging (MRI) classification in which a quantum convolutional (“quanvolutional”) layer performs early-stage feature extraction, a classical CNN backbone performs hierarchical representation learning, and a variational quantum circuit (VQC) performs final classification, so that quantum processing brackets the classical backbone at both ends. Using a real, held-out four-class brain tumor MRI dataset (glioma, meningioma, pituitary tumor, and no tumor), we implement and train four architectural variants under matched conditions — (A) a fully classical CNN baseline, (B) a quantum-convolution-plus-classical-CNN variant, (C) a classical-CNN-plus-quantum-head variant, and (D) the full proposed QCQ-CNN — and report actual experimental accuracy, precision, recall, and F1-score on a common test set, together with training curves, confusion matrices, and qualitative visualizations of the learned quantum feature maps. Our small-scale experiments show that quantum convolutional preprocessing alone improves macro-F1 over the classical baseline (0.204 to 0.382), while replacing the classical classification head with a variational quantum circuit degrades performance in both the quantum-head-only and the full QCQ-CNN configurations. We discuss this finding in the context of variational quantum circuit trainability at low qubit counts and limited data, and argue that our staged ablation methodology — rather than a single best-accuracy claim — constitutes the paper's principal novelty: a reusable protocol for empirically localizing where, in a CNN pipeline, quantum layers help versus hurt. We conclude with concrete design recommendations and directions for scaling QCQ-style frameworks. Keywords: Quantum Machine Learning, Hybrid Quantum-Classical Neural Networks, Quanvolutional Neural Network, Variational Quantum Circuit, Convolutional Neural Network, Brain Tumor Classification, Medical Image Processing