IEEE World AI IoT Congress (AIIoT)
Quantum Convolutional Neural Network for Mammography Classification
Qinwen Deng, Priyank Kalgaonkar
Abstract
Breast cancer remains a leading cause of cancer-related mortality among women worldwide, where early detection via mammography is critical for improving survival rates. While classical convolutional neural networks (CNNs) achieve high accuracy in medical imaging, they require extensive computational resources and large training datasets. This work investigates the feasibility of applying a quantum convolutional neural network (QCNN) for mammogram classification within constraints compatible with near-term quantum hardware. We present a comparative study systematically evaluating two classical dimensionality reduction techniques (Principal Component Analysis and Autoencoding) and two parameterized quantum circuit ansätze (SU(4) and U5) across two public datasets (MIAS and mini-DDSM).Experimental results demonstrate that the QCNN achieves the best accuracy of 81% on the MIAS dataset using autoencoder with the SU(4) circuit. These findings suggest that QCNNs are a feasible approach for mammography classification, achieving reasonable accuracy with fewer than 50 trainable parameters, highlighting their potential in settings where labeled data is scarce.