Quantum-SpinalNet: Hybrid Deep Learning for Accurate Breast Cancer Detection
Overview
Quantum-SpinalNet, a novel hybrid deep learning framework, combines Swin ResUNet3+ for tumor segmentation with Deep Quantum Neural Network and SpinalNet for classification, achieving high accuracy in breast cancer detection from mammograms. Evaluated on CBIS-DDSM and DDSM datasets, it demonstrated superior performance metrics including 93.8% accuracy and a Dice coefficient of 0.89.
Background
Breast cancer is the most common and deadly cancer among women worldwide, with early detection critical to improving survival rates. Mammography remains the primary imaging modality for early diagnosis, but interpretation is prone to human error. Advances in machine learning and deep learning, particularly convolutional neural networks, have enhanced automated detection and classification of breast lesions. However, challenges remain due to limited annotated data, computational costs, and segmentation precision. Quantum-SpinalNet addresses these by integrating advanced preprocessing, transformer-based segmentation, and quantum-inspired classification.
Data Highlights
Metric
Value
Accuracy
93.8%
Sensitivity
94.1%
Specificity
92.7%
Precision
91.2%
F1 Score
92.6%
Dice Coefficient
0.89
Intersection over Union (IoU)
0.82
Key Findings
Quantum-SpinalNet integrates Swin ResUNet3+ for effective tumor segmentation with Deep Quantum Neural Network and SpinalNet for classification.
Advanced preprocessing techniques include CEAMF-based denoising, Z-score normalization, and context-aware contrast enhancement using spatial energy curves.
The model achieved 93.8% accuracy and 94.1% sensitivity on CBIS-DDSM and DDSM datasets, outperforming traditional hybrid models.
Dice coefficient of 0.89 and IoU of 0.82 indicate precise tumor localization and segmentation.
Quantum-SpinalNet reduces computational costs and improves border segmentation compared to prior transformer-based models like TransUNet and Swin-UNETR.
Clinical Implications
Quantum-SpinalNet offers a robust, interpretable tool to assist radiologists in mammographic breast cancer detection, potentially reducing diagnostic errors and workload. Its high segmentation and classification accuracy support improved clinical decision-making and early intervention. Integration into diagnostic workflows could enhance consistency and reliability of breast cancer screening.
Conclusion
Quantum-SpinalNet represents a significant advancement in AI-assisted breast cancer detection, combining state-of-the-art segmentation and classification techniques to deliver high accuracy and precise tumor localization. This framework holds promise for enhancing mammographic diagnostic performance and patient outcomes.
Related Resources & Content
World Health Organization 2022 -- Breast Cancer Statistics and Impact
CBIS-DDSM and DDSM Datasets -- Mammographic Image Repositories
