Hybrid Quantum-Classical CNN Enhances Breast Tumor Classification Accuracy
A new study published on arXiv introduces a hybrid Quantum-Classical Convolutional Neural Network (QCNN) designed to improve the binary classification of breast tumors using the BreastMNIST dataset. The proposed architecture combines classical convolutional feature extraction with two distinct four-qubit quantum circuits: an amplitude-encoding variational quantum circuit and an angle-encoding circuit with circular entanglement. These quantum components generate feature embeddings that are fused with classical features to create a joint feature space for final classification. To ensure a fair comparison, the hybrid model was parameter-matched against a baseline classical CNN, with both trained under identical conditions using the Adam optimizer. Experimental results from five independent runs demonstrated that the hybrid QCNN achieved statistically significant improvements in accuracy over the classical counterpart, validated by a Wilcoxon signed rank test (p = 0.03125) and a large effect size (Cohen's d = 2.14). This research highlights the potential of leveraging quantum entanglement and feature fusion in medical imaging tasks and establishes a statistical framework for evaluating hybrid quantum models in biomedical applications.
Wire timeline
Hybrid Quantum-Classical CNN Enhances Breast Tumor Classification Accuracy
A new study published on arXiv introduces a hybrid Quantum-Classical Convolutional Neural Network (QCNN) designed to improve the binary classification of breast tumors using the BreastMNIST dataset. The proposed architecture combines classical convolutional feature extraction with two distinct four-qubit quantum circuits: an amplitude-encoding variational quantum circuit and an angle-encoding circuit with circular entanglement. These quantum components generate feature embeddings that are fused with classical features to create a joint feature space for final classification. To ensure a fair comparison, the hybrid model was parameter-matched against a baseline classical CNN, with both trained under identical conditions using the Adam optimizer. Experimental results from five independent runs demonstrated that the hybrid QCNN achieved statistically significant improvements in accuracy over the classical counterpart, validated by a Wilcoxon signed rank test (p = 0.03125) and a large effect size (Cohen's d = 2.14). This research highlights the potential of leveraging quantum entanglement and feature fusion in medical imaging tasks and establishes a statistical framework for evaluating hybrid quantum models in biomedical applications.
cs.AI updates on arXiv.org