Automatic classification of pancreatic cancer from urinary biomarkers using equivariant quantum convolutional neural networks with hybrid optimization algorithm

Early detection of Pancreatic Ductal Adenocarcinoma (PDAC) is crucial to improve survival rates. This study proposes an automated classification framework using equivariant quantum convolutional neural networks (EQCNNs) optimized with a Hybrid Adam-Dingo and Quantum Artificial Hummingbird Algorithm (HybADOQAHA). Urinary biomarkers - creatinine, lymphatic vessel endothelial hyaluronan receptor 1 (LYVE1), regenerating islet-derived protein 1 beta (REG1B), and trefoil factor 1(TFF1) were analyzed from 590 samples comprising healthy, benign, and PDAC cases. Pre-processing with dual-feature filtering and feature extraction <i>via</i> lifted Euler characteristic transform enhanced data quality. The experimental results demonstrate better accuracy, precision, recall, specificity, and Area Under the Curve (AUC) compared with baseline models, establishing the proposed method as a promising non-invasive diagnostic tool for early PDAC detection.