Research & Projects

Abstract

Around the half of the population in the world are affected by oral diseases, making it one of the most common health conditions. Quantum implementation in medical domain has revealed its potential and versatile applicability especially in medical imaging. This paper explores oral disease identification using hybrid quantum-classical neural networks (HQCNN) and quantum convolution neural networks (QCNN). Our work investigates the possibilities of quantum machine learning in processing complicated dental image data and the contributions it can make in oral healthcare. We implemented a hybrid and a pure QNN leveraging Qiskit framework and a whole dataset of annotated oral disease dataset. Our 8 qubit structured QCNN model and 2 qubit architecture of HQCNN model extract the image features by encoding the features into quantum circuits enabling more expressive demonstration employing fewer parameters. The final result showcases that QCNN and HQCNN perform better than CNNs in disease classification and promise better accuracy, generalization and computational efficiency. This experiment highlights a pioneering step in applying quantum inspired models for oral diagnostics, identifying promising avenues for improving oral healthcare worldwide.

Keywords: Quantum Machine Learning, Hybrid Quantum-Classical Neural Networks (HQCNN), Quantum Convolution Neural Networks (QCNN), Oral Diseases Detection, Parameterized Quantum Circuits, Principal Component Analysis (PCA), Angle Encoding

https://www.preprints.org/manuscript/202605.0525

📌 Disclaimer

This Project was promised to teach and help through guidance and mentorship under Mahdy Research Academy’s private thesis program after taking approximately 70,000 BDT, (shared among 6 participants for Part A and Part B); however, The code implementation and quantum machine learning components were self-learned from Qiskit repositories and tutorials and were not guided or assisted by the academy. All coding and writing were carried out independently by the authors (Md. Shakhawat Hossain and Md. Mehedi Hasan), based on publicly available resources including this video tutorial: https://www.youtube.com/watch?v=IohyKm9c4_Q and relevant documents https://qiskit-community.github.io/qiskit-machine-learning/tutorials/11_quantum_convolutional_neural_networks.html, https://files.batistalab.com/teaching/attachments/chem584/vic_hqcnn.pdf; not by the Academy; 

This project builds upon the Quantum Convolutional Neural Network (QCNN) framework by Isaac Cong et al., with implementations adapted from IBM’s Qiskit. Parts of the code are modified from original Qiskit resources and retain required Apache License 2.0 notices and attributions. This work is part of a Qiskit-based project and complies with IBM licensing policies. The extended codebase is released under the MIT License, allowing free use, modification, and distribution.