Prediction of Covid'19 and pneumature from lung x-ray images using quantum machine learning methods

Quantum convolutional neural networks (QCNNs) expand the capabilities of CNNs by leveraging some of the potentially powerful aspects of quantum computing. It works on the input data by locally transforming the data using a series of random quantum circuits. Based on the efficiency of classical convolutional neural networks, using Quanvolutional neural networks (QNNs) data were analyzed, predictions were made and results were evaluated. Binary classification of the covid'19 data set encoded in quantum form was performed. In addition, the performance of Pennylane's "default qubit" device was investigated by taking into account different parameters. The dataset used contains 250 training and 65 test images for the model. The images provided in the dataset are real-life chest X-rays and have not been previously modified. However, due to some limitations in computational resources, the size is kept as 28x28 in this study. Model-1 classifies between two classes, 'Normal Person' and 'Covid'19/Viral Pneumonia'. Model-2 classifies between two classes, 'Covid'19' and 'Viral Pneumonia'. In Quantum Classifier 1, 11 features were used from 256 feature sized input data extracted by Fundamental Data Analysis. Here, approximately 70% accuracy has been achieved. Using the TruncatedSVD method in Quantum Classifier 2, 256 features of each image are reduced to 4. Approximately 72% accuracy (accuracy) was obtained. In Quantum Classifier 3, its data is reduced to only 2 features. Unexpectedly, this yielded an accuracy of 76%, the highest ever approached.