Quantum-Enhanced Deep Learning Approach for Lung Tumor Identification

Abstract

Lung cancer is one of the leading causes of cancer-related deaths worldwide, affecting millions of people every year. Early and accurate detection is significant for providing suitable treatment and improving patient survival. Computed Tomography (CT) scan images are widely utilized in lung cancer diagnosis because they provide detailed information about lung tissues and abnormalities. Recently, quantum computing has also been explored in medical image analysis owing to its potential to process complex and high-dimensional data more efficiently. Therefore, this study introduces a Quantum Computing-based Framework for Enhanced Lung Cancer Detection (QCF-ELCD) using CT images. The proposed QCF-ELCD framework involves advanced pre-processing steps to enhance the image quality and uniformity within the data. The hybrid feature extraction process integrates ConvNext architecture along with an efficient channel attention mechanism. This combination extracts highly discriminative features, which are then forwarded to the quantum fidelity-based classification algorithm. The classifier enhances the model's performance by identifying the class with the maximum quantum fidelity match. Moreover, the AdaBelief optimizer is employed to improve model parameters during training, contributing to improved classification performance. Additionally, Grad-CAM++ is utilized to visualize the discriminative regions learned by the proposed QCF-ELCD, thereby improving model interpretability and clinical transparency. The experimental analysis of the QCF-ELCD framework is conducted using the IQ-OTH/NCCD Lung Cancer Dataset and LIDC-IDRI Dataset, achieving promising classification accuracy of 95.13% and 94.23%, respectively. Consequently, the proposed QCF-ELCD framework demonstrates its potential as an effective tool for automatic lung cancer classification, outperforming existing models.