QuMo: Benchmarking a Quantum Moiré-Based Classifier for Brain Tumor Diagnosis

Brain tumor diagnosis from MRI remains a critical challenge in neuro-oncology, requiring accurate detection of subtle morphological differences across tumor types. Recent advances in quantum machine learning (QML) have introduced new possibilities for representing high-dimensional medical imaging data through quantum states, yet most existing models lack mechanisms to efficiently capture non-local spatial hierarchies and fine-grained structural variations essential for distinguishing glioma subtypes. To address these limitations, we present QuMo, a Quantum Moiré-based classifier that integrates quantum-enhanced feature extraction with Moiré pattern-inspired modulation kernels. Unlike conventional CNNs or standard variational quantum classifiers, our approach embeds spatial interference patterns directly into the quantum feature mapping pipeline, enabling compact yet expressive representations. We implement QuMo within a hybrid quantum–classical architecture using parameterized quantum circuits (PQCs) optimized by variational algorithms, and evaluate it on the BraTS brain tumor dataset. To ensure rigorous and reproducible comparisons, we benchmarked QuMo using the QUARK framework, a standardized benchmarking suite for quantum applications. QuMo was evaluated against a range of quantum and hybrid baselines, including PQC, VQC, QNN, QSVM, QBM, QCNN, and QFT-based classifiers, with metrics aggregated over multiple independent runs. Our results show that QuMo achieves higher classification accuracy, improved parameter efficiency, and reduced quantum resource requirements compared to all baselines. Benchmarking further demonstrates that QuMo exhibits lower variance and greater robustness across repeated trials. Moiré-based quantum kernels enhance sensitivity to heterogeneous tumor boundaries and subtle structural distortions, offering interpretable, scalable, and resource-efficient QML models for precision neurodiagnostics. This establishes geometric interference–driven quantum encodings as a promising direction for nextgeneration medical imaging analytics. The code will be publicly available at https://github.com/akortheanchor/QuMo.