Research On Medical Image Classification Method Based On Quantum Hybrid Neural Network

In recent years,the Novel Coronavirus 2019(COVID-19)has caused increasing medical pressure due to its high infectivity,and the widely used COVID-19 nucleic acid test and antibody test also have certain false negatives.Chest CT scan images and chest X-ray images are the most widely used in the COVID-19 epidemic.Accurate detection of such images can effectively assist in the diagnosis and treatment of whether people are infected with COVID-19.Therefore,this thesis mainly proposes a medical image classification method based on quantum hybrid neural network for chest CT scan images and chest X-ray images.The main work is as follows:Firstly,the distribution of pixel values of chest CT scan images and chest X-ray images is analyzed to determine whether the image needs to be preprocessed.If preprocessing is required,histogram equalization is used to make the image pixels evenly distributed.The experimental results show that compared with the original chest CT scan image,the accuracy of the proposed two quantum hybrid neural network models is improved by 6%and7.8%respectively,which can effectively classify medical images.Secondly,a medical image classification method based on Hybrid Quantum-Classical Computing(HQCC)model is proposed.The feature extraction part of the classification method is composed of a self-created Pre-Residual Convolutional Network(PRCN),which is mainly composed of cascading convolutional blocks and residual blocks,which can combine the shallow features of the image with the deep features,so as to learn richer image information.At the same time,a quantum layer with quantum neurons is introduced to enhance the flexibility and versatility of quantum computing by parameterizing quantum circuits,and it is combined with PRCN to act as a classifier.The HQCC model formed by it can smooth the convergence trend of training and reduce the compilation loss,which improves the classification performance of the model.Furthermore,a medical image classification method based on a Classical-to-Quantum Ensemble(Ensemble(CQ))model of transfer learning is proposed.The introduction of more complex quantum circuits,namely Dressed Quantum Circuits(DQCs),can effectively obtain quantum advantages and improve model classification performance.Use Res Net50,VGG16 and Alex Net three pre-trained models,respectively cascaded with DQCs.At the same time,in order to improve the robustness of the model,the three single models cascaded with DQCs are integrated with DQCs to form a multi-model ensemble.On this model,the classification accuracy of chest CT scan images and chest X-ray images is as high as 80.8% and 96%,respectively.Finally,through the verification of breast ultrasound images and brain tumor images,it is proved that the proposed quantum hybrid neural network can effectively classify medical images.At the same time,it also shows that the introduction of quantum circuits in the research of medical image classification methods based on deep learning has certain value and significance.