Abstract

An accurate and early detection of uterine fibroid is important for effective clinical treatment and prevention of delays in diagnostics. Sometimes classical learning models face difficulties due to complex feature representation and generalization, specifically when trained under limited medical datasets. To work through with such problems, this study proposes an organised pipeline with hybrid quantum classical convolutional neural network that incorporates classical feature extraction with parameterized variational quantum circuit for enhanced high-dimensional feature representation and classification. The proposed work was evaluated along with a classical CNN under both non-augmented and augmented training constraints. Experimental results under non-augmented constraints, the structured pipeline achieved ACC, AUC, SE, SP of 89%, 96%, 98%, 77% for hybrid QCNN and 63%, 69%, 72%, 51% for CNN. Similarly, under augmented conditions, achieved ACC, AUC, SE, SP of 90%, 96%, 96%, 83% for hybrid QCNN and 67%, 75%, 67%, 67% for CNN respectively. In both constraints, sensitivity and specificity was improved in hybrid model and it reduces false negative cases. Throughout epochs, the hybrid model showed faster convergence and improved generalization stability compared to classical CNN. These outcomes highlight the effectiveness of quantum-based approaches into classical learning for complex image classification by improving diagnostic reliability, reducing false negatives, false positives and supporting more accurate uterine fibroid detection systems.

Authors

A. Kasthuri, J. Gowrishankar
Jain Deemed-to-be University, India

Keywords

Hybrid Quantum Convolutional Neural Network, Uterine Fibroid, Preprocessing, Segmentation, Data Augmentation