Hybrid Quantum Neural Networks for Enhanced Breast Cancer Thermographic Classification: A Novel Quantum-Classical Integration Approach
Riza Alaudin Syah, Irwan Alnarus Kautsar, Haza Nuzly bin Abdull Hamed, Gunawan Witjaksono · 2026
Read Hybrid Quantum Neural Networks for Enhanced Breast Cancer Thermographic Classification: A Novel Quantum-Classical Integration Approach by Riza Alaudin Syah, Irwan Alnarus Kautsar, Haza Nuzly bin Abdull Hamed, Gunawan Witjaksono (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-16953.txt.
Introduction
Syah and three colleagues blend quantum circuits with classical convolutional networks to classify breast cancer from heat images. Classical deep learning struggles with complex thermal patterns in this task. Their hybrid design encodes features with a four-qubit variational circuit and attention, then fuses them classically, showing stronger convergence and feature representation than leading standard models in their tests.
Catalog record
| CREATOR | Riza Alaudin Syah, Irwan Alnarus Kautsar, Haza Nuzly bin Abdull Hamed, Gunawan Witjaksono |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 23792 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2604.16953
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Hybrid Quantum Neural Networks for Enhanced Breast Cancer Thermographic Classification: A Novel Quantum-Classical Integration Approach Riza Alaudin Syah, Irwan Alnarus Kautsar, Haza Nuzly bin Abdull Hamed, Gunawan Witjaksono Abstract Breast cancer diagnosis through thermographic image analysis remains a critical challenge in medical AI, with classical deep learning approaches facing limitations in complex thermal pattern classification tasks. This paper presents a novel Hybrid Quantum Neural Network (HQNN) architecture that integrates quantum computing principles with classical convolutional neural networks for enhanced breast cancer classification. Our approach employs parameterized quantum circuits with multi-head attention mechanisms for quantum-aware feature encoding, coupled with classical convolutional layers for comprehensive pattern recognition. The quantum component utilizes a 4-qubit variational circuit with strongly entangling layers, while the classical component incorporates advanced attention mechanisms for feature fusion. Experimental validation on breast cancer thermographic data demonstrates substantial performance improvements over state-of-the-art classical architectures, with the quantum-enhanced approach exhibiting superior convergence dynamics and enhanced feature representation capabilities. Our findings provide evidence for quantum advantage in medical image classification through classical simulation, establishing a framework for quantum-classical hybrid systems in healthcare applications. The methodology addresses key challenges in quantum machine learning deployment while maintaining computational feasibility on near-term quantum devices. I Introduction Breast cancer represents the second most common cancer globally, with approximately 2.3 million new cases diagnosed annually [ 1 ] . Early detection through accurate thermographic analysis significantly improves patient survival rates, with 5-year survival rates exceeding 90% when detected in early stages [ 2 ] . However, manual interpretation of thermographic images is time-consuming, subjective, and requires extensive expertise, creating bottlenecks in clinical workflows and potential for diagnostic errors [ 3 ] . Deep learning approaches, particularly Convolutional Neural Networks (CNNs), have demonstrated remarkable success in medical image classification tasks [ 4 , 5 ] . EfficientNet architectures have emerged as state-of-the-art models, achieving superior performance through compound scaling of network depth, width, and resolution [ 6 ] . However, classical deep learning approaches face inherent limitations in feature representation capacity and may struggle with complex pattern recognition in high-dimensional medical data. Quantum machine learning has emerged as a promising paradigm that leverages quantum mechanical phenomena to enhance computational capabilities [ 7 , 8 ] . Hybrid Quantum Neural Networks (HQNNs) combine classical neural network architectures with quantum circuits, potentially offering exponential advantages in feature space exploration and pattern recognition [ 9 , 10 ] . The mathematical foundation of quantum circuits in HQNNs can be expressed through unitary transformations applied to quantum states. Consider a quantum circuit with n qubits initialized in state |\psi_{0}\rangle . The circuit applies a sequence of parameterized quantum gates U(\theta) to produce the final state: |\psi(\theta)\rangle=U_{L}(\theta_{L})\cdots U_{2}(\theta_{2})U_{1}(\theta_{1})|\psi_{0}\rangle (1) where \theta=\{\theta_{1},\theta_{2},\ldots,\theta_{L}\} represents the trainable parameters of the quantum circuit. The expectation value of a measurement operator \hat{O} provides the quantum circuit output: \langle\hat{O}\rangle=\langle\psi(\theta)|\hat{O}|\psi(\theta)\rangle (2) This quantum output is then integrated with classical neural network layers to form the hybrid architecture, enabling the exploitation of both quantum superposition and classical feature extraction capabilities. Despite theoretical advantages, practical implementation of quantum machine learning in medical applications remains underexplored. Most existing studies focus on toy datasets or simplified problems, lacking comprehensive comparison with state-of-the-art classical approaches on real medical imaging tasks [ 11 ] . Furthermore, the optimal quantum circuit design, qubit requirements, and performance trade-offs for medical image classification remain unclear. This study addresses these gaps by conducting a systematic comparison between HQNNs and classical EfficientNet architectures for breast cancer thermographic image classification. Our investigation aims to: (1) quantify the performance advantages of quantum-enhanced approaches, (2) analyze training dynamics and convergence patterns, (3) evaluate computational requirements and scalability, and (4) provide practical guidelines for quantum machine learning deployment in medical diagnostics. II Related Work The intersection of quantum computing and medical image analysis has gained significant attention in recent years. Schuld and Killoran [ 10 ] provided a comprehensive review of quantum machine learning algorithms, highlighting potential advantages in feature mapping and optimization landscapes. Their work established theoretical foundations for quantum-enhanced pattern recognition. In medical imaging applications, several studies have explored quantum approaches. Li et al. [ 12 ] investigated quantum convolutional neural networks for medical image segmentation, demonstrating improved performance on brain tumor datasets. However, their evaluation was limited to small-scale problems with simplified quantum circuits. For breast cancer classification specifically, classical approaches have achieved remarkable success. Spanhol et al. [ 13 ] established benchmark datasets for breast cancer thermographic image analysis, enabling standardized evaluation protocols. Araujo et al. [ 14 ] achieved 83.3% accuracy using traditional CNN architectures, while more recent work by Kassani et al. [ 15 ] reported 98.13% accuracy using ensemble methods. Quantum machine learning applications in healthcare have shown promising preliminary results. Mangini et al. [ 16 ] demonstrated quantum advantage in drug discovery applications, while Garg and Ramakrishnan [ 17 ] reviewed quantum computing applications in bioinformatics. However, direct comparison with state-of-the-art classical methods on real medical datasets remains limited. The gap in existing literature lies in comprehensive evaluation of quantum-enhanced approaches against established classical benchmarks using realistic medical imaging datasets. Our work addresses this by providing systematic comparison between HQNNs and EfficientNet on breast cancer classification tasks. III Methodology III-A Dataset and Experimental Setup III-A 1 Dataset Source and Characteristics We utilized the ”Breast Cancer Detection Using Thermography” dataset from Kaggle [ 18 ] , containing 762 thermal images ( 480\times 640 pixels, PNG format) for binary classification of normal vs. malignant breast tissue. The dataset provides thermographic images captured at standardized conditions with consistent temperature mapping. From this dataset, we selected 262 representative samples (132 normal, 130 malignant) to ensure balanced class distribution and computational feasibility for quantum circuit training. Fig. 1: Augmented training samples from the breast cancer thermography dataset. Top row: normal tissue samples (label 0) with applied data augmentations including horizontal flips, rotations, and color jittering. Bottom row: malignant tissue samples (label 1) with identical augmentation transforms, showcasing the thermal pattern diversity used for HQNN training. The selected 262 samples were split using stratified sampling: 209 training samples (105 normal, 104 malignant) and 53 validation samples (27 normal, 26 malignant), maintaining an 80:20 ratio. A separate test set of 50 samples was reserved for final evaluation. Figure 1 demonstrates the augmented training samples with applied data transformations from both normal and malignant classes. All experiments were conducted on macOS with MPS acceleration for consistent hardware conditions. III-A 2 Data Preprocessing and Augmentation Images underwent comprehensive preprocessing for thermal image analysis: (1) Resizing to 64\times 64 pixels for computational efficiency while preserving thermal patterns, (2) Intensity normalization using ImageNet statistics (mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]) for transfer learning compatibility, (3) Temperature mapping standardization to ensure consistent thermal range representation. Training samples were augmented using: random horizontal flipping ( p=0.5 ), rotation ( \pm 10^{\circ} ), and thermal contrast adjustment (brightness =0.2 , contrast =0.2 ). This increased the effective training set from 209 to approximately 1,045 samples while maintaining thermographic validity. Validation and test sets remained unaugmented for unbiased evaluation. III-A 3 Experimental Configuration All models (HQNN, EfficientNet-B0, ResNet-50) were trained with identical hyperparameters for fair comparison: batch size 16, initial learning rate 0.0001 (selected via grid search over [0.001, 0.0001, 0.00001]), exponential decay