Integrating Quantum Principles with Classical Transfer Learning Approaches for Pneumonia Detection from Chest Radiographs

Pneumonia is a severe and deadly lung disease, that kills millions of people around the world annually. If detected early, these death cases can be avoided by appropriate medical care. Yet, the regions with poor access to healthcare might find it difficult for early detection. The application with a pre-programmed chest X-ray pneumonia detection method is proper which can easily meet these issues and give reliable, accurate diagnosis. Automated pneumonia detection using chest X-rays is an active field of study. Taking together, the results collected from the experiments in this study underscore the effective combination of traditional machine learning(ML) with quantum machine learning(QML) for chest X-ray pneumonia diagnosis. Initially, few transfer learning approaches like ResNet-18, AlexNet, Inception-v3 and VGG-16 were taken to create the classical binary image classification models. Secondly, the aggregated classification model architecture was implemented based on the union of classical and quantum networks engineered utilizing the same blueprint based entirely on traditional transfer learning techniques. Compared to classical models of similar designs, the hybrid models trained and tested in this experiment display greater classification proficiency. The AlexNet model, when incorporated with the hybrid transfer learning classifier achieved an effective classification accuracy of 97.79 %, yielding a total improvement of 2.37 % over the models based solely on the classical approach of similar architecture. Optimal results were achieved by this hybrid model where AUC is 0.99 and other performance metrics like F1-score, recall and precision generated a value of 0.97.