Quantum-Assisted CNN Revolutionizes Breast Cancer Diagnostics

By Dr Silpaja Chandrasekar, PhDReviewed by Susha Cheriyedath, M.Sc.Oct 30 2024

In a paper published in Scientific Reports, researchers introduced a quantum hybrid classical convolutional neural network (QCCNN) approach aimed at improving breast cancer diagnosis. By leveraging quantum computing’s advanced data processing capabilities, this approach enhances diagnostic accuracy and efficiency, showcasing the potential of quantum-assisted methods in medical imaging.

The QCCNN has been designed to enhance learning efficiency and address the high computational demands of conventional CNNs by integrating a quantum convolutional layer.

Simulations conducted on datasets from the German Breast Cancer Study Group (GBSG), the Surveillance, Epidemiology, and End Results Program (SEER), and the Wisconsin Diagnostic Breast Cancer (WDBC) database demonstrated that QCCNN outperformed traditional CNN and logistic regression models. These results underscore QCCNN's potential to improve breast cancer diagnostics through more effective and resource-efficient processing.

Background

Machine learning (ML), particularly convolutional neural networks (CNNs), has significantly advanced breast cancer diagnostics, enabling early detection through superior image analysis. However, CNNs face challenges in computational efficiency when dealing with large, complex datasets, creating a need for more effective solutions.

Quantum computing, with its unparalleled parallel processing capabilities, presents new opportunities for accelerating and refining data analysis in medical diagnostics. While its application in breast cancer diagnosis is still emerging, quantum computing shows great potential to transform diagnostic speed and accuracy.

Innovative Model for Breast Cancer

This study investigates the potential of a QCCNN for breast cancer diagnosis, comparing its performance against traditional models like CNN and logistic regression. The analysis employed three datasets—GBSG, SEER, and WDBC—to train and validate each model’s diagnostic accuracy.

Sourced from Kaggle, these datasets provide comprehensive breast cancer data, including patient demographics and tumor features. Specifically, the GBSG dataset includes 686 samples with 11 prognostic features, SEER contains 4,024 samples of patients diagnosed between 2000 and 2017, and WDBC offers 569 samples with 30 tumor-specific features.

Data preprocessing involved normalizing feature ranges to ensure consistent model training across varying scales. The CNN architecture consisted of convolutional, pooling, and fully connected layers designed to capture and classify image features. Feature extraction used sliding filters over images, with pooling layers reducing spatial dimensions. Classification accuracy served as the primary performance metric.

In the QCCNN, a quantum convolutional layer was incorporated into this structure to combine quantum computations with classical layers, enhancing image recognition. Developed using the PennyLane framework, the QCCNN applied quantum circuits for feature encoding and entanglement operations. Both CNN and QCCNN models were optimized using the Adam optimizer with a 0.1 learning rate, with cross-entropy loss assessing prediction accuracy.

Logistic regression, used as a baseline, employed a simpler linear model for classification, estimating probabilities through a weighted sum of input features. This model used a sigmoid function to map features to probabilities, providing a point of comparison for evaluating the diagnostic capabilities of CNN and QCCNN. Logistic regression was optimized with the Adam optimizer at a 0.01 learning rate.

The experimental setup, using PyTorch and PennyLane, enabled efficient training and evaluation of QCCNN and CNN. The QCCNN aimed to leverage the strengths of both quantum and classical layers, showing promise for breast cancer diagnostics. Evaluation criteria included classification accuracy and cross-entropy loss, providing a comprehensive assessment of each model's performance based on true and false classifications.

Hybrid Quantum-Classical Diagnostic Model

This study validated the efficiency of the QCCNN model for breast cancer diagnosis by comparing it with CNN and logistic regression on the GBSG, SEER, and WDBC datasets. Results showed that QCCNN consistently achieved the highest accuracy, with 74.38 % on GBSG, 90.09 % on SEER, and 96.16 % on WDBC validation sets, outperforming the other models. Each model was trained on 70 % of the data and validated with the remaining 30 %, using accuracy as the primary performance metric.

These findings highlight the QCCNN model's potential in medical imaging, demonstrating the advantages of quantum computing for disease diagnosis. Unlike traditional methods, the QCCNN model enhances feature extraction and improves classification performance by incorporating quantum layers.

Built with the PennyLane framework, the QCCNN utilized quantum entanglement in its convolutional layers, significantly boosting diagnostic accuracy across datasets. This hybrid approach underscores the promise of quantum-classical models in medical image analysis, pointing to potential applications in complex diagnostic tasks where conventional models may face limitations.

Conclusion

In summary, this study introduced a QCCNN model for breast cancer diagnosis, demonstrating superior classification accuracy compared to traditional models like CNN and logistic regression across the GBSG, SEER, and WDBC datasets. Despite challenges in quantum circuit design and the existing gap between classical and quantum computing, the study highlights the promising potential of QCCNN in medical diagnostics.

Future research will focus on applying the model to breast cancer staging and further evaluating its computational efficiency, advancing its practical use in real-world settings.

Journal Reference

Xiang, Q. et al. (2024). Quantum classical hybrid convolutional neural networks for breast cancer diagnosis. Scientific Reports, 14:1, 1-13. DOI: 10.1038/s41598-024-74778-7, https://www.nature.com/articles/s41598-024-74778-7

Disclaimer: The views expressed here are those of the author expressed in their private capacity and do not necessarily represent the views of AZoM.com Limited T/A AZoNetwork the owner and operator of this website. This disclaimer forms part of the Terms and conditions of use of this website.