Quantum neural network-assisted learning for small medical datasets: a case study in emphysema detection

Advancement in AI and deep learning has transformed medical image diagnosis; however, approaches to disease diagnosis, such as the detection of emphysema, one of the severest forms of COPD, stand to benefit from these technologies more. The main problem with the detection of emphysema from CT scans is the lack of very large, annotated datasets to train deep learning models. Classic models, like CNNs, are usually underfitting to small datasets and hence have very poor diagnostic accuracy. To address this challenge, we consider an integrated hybrid quantum-classical neural network model by combining the quantum variational circuits with CNNs. This new approach uses the power of quantum computing to identify subtle patterns in small datasets and could solve one of the key problems of deep learning. The model is pre-trained on large chest X-ray datasets and fine-tuned on a smaller emphysema dataset, which allows it to generalize more when data is limited. The experimental results confirm that the proposed approach is effective; namely, the quantum-assisted model reaches an accuracy of 0.5690 and F1-score of 0.5990, outperforming the traditional CNN models. This work points to the novelty of quantum computing in diagnosis with limited amounts of data, a very important challenge in this area of medical AI. Given that our research will conform to the limitation and work on small datasets, this work opens a new frontier in medical image analysis and shows ways in which QNN can substantially outperform traditional methods in detecting subtle markers of diseases; this indeed contributes to the growing body of knowledge in quantum-enhanced AI and opens up new frontiers toward some potential applications in the field of diagnosis of rare diseases and health diagnostics.