Interpretable Analysis of Oral Mucosal Lesions Using U-Net and Quantum Approaches

Oral mucosal lesions are abnormal changes found in the tissue of the mouth. They can be painful and may include abnormal cell growths, as well as rare tongue and hard- palate disorders. They may result from infections, inflamma- tory processes, trauma, autoimmune conditions, or precancerous changes. A recent study found a 59.6% prevalence of oral mucosal lesions among elderly individuals, highlighting a signif- icant association with factors such as age over 65, male gender, tobacco habits, and denture usage. Thus, lesions can significantly impact a patient's quality of life, necessitating accurate diagnosis and timely intervention. In this study, we use a combination of two advanced technologies-U-Net, a deep learning model, and quantum computing-to analyze medical images of oral health. We use a publicly available dataset containing labeled images classified as ‘Lesion’ and ‘Oral Cavity,’ focusing only on lesion detection. Our results show that while the U-Net model achieves 95.02% segmentation accuracy, the quantum approach using QCNN achieves 74.5% accuracy. Although the quantum model does not outperform U-Net in segmentation, it demonstrates potential for future research in quantum-enhanced medical imaging. To measure how well our method works, we use different evaluation techniques, including Dice Coefficient, Intersection over Union (IoU), accuracy, precision, recall, and F1-score. Our results indicate that while quantum computing has shown high accuracy in previous research for classification tasks, its application to segmentation-particularly in oral lesion detection-faces challenges such as lower accuracy, quantum noise, and hardware constraints. This research highlights the potential of integrating AI and quantum approaches for medical diagnosis while acknowledging the need for further improvements in quantum seamcntation models.