<![CDATA[The primary challenge in employing conventional CNNs for medical image classification lies in their high parameter complexity and the potential degradation in generalization efficiency on large-scale datasets. This study explores an alternative solution through a hybrid architecture, namely the HQCNN, applied to the PathMNIST dataset. In this schema, classical CNN layers function as a spatial feature extractor, which is then mapped into the quantum domain via angle encoding mechanisms on a 4-qubit configuration. The novelty of this research lies in the scalability testing of the model, conducted incrementally up to a volume of 130,000 data samples. Experimental results confirm the superiority of the HQCNN, achieving an accuracy of 83.0%, significantly outperforming the classical CNN model, which exhibited a decline in effectiveness at 77.7%. These findings indicate that the integration of quantum principles, such as entanglement via CNOT gates, enables sharper class separation and reduces the risk of overfitting in large scale medical datasets.]]>
Copyrights © 2026
