Brain Disease Analysis And Microservice Research Based On Image Semantic Segmentation

Brain tumor,ischemic stroke and cerebral hemorrhage are three common serious diseases of the nervous system.They pose a serious threat to human health and quality of life.With the flourishing development of medical imaging technology,medical image analysis has gradually become an important means to assist doctors in clinical disease diagnosis and research.Computed Tomography（CT）is broadly used in the diagnosis of cerebral hemorrhage due to its advantages of fast imaging,low price and high sensitivity to hemoglobin.Magnetic resonance imaging（MRI）technology,its non-invasiveness and excellent resolution of brain structure make it generally used in the diagnosis of ischemic stroke and brain tumors.The lesion segmentation of these three brain diseases helps doctors to make early diagnosis of patients,supportive treatment of symptoms and related prognosis evaluation.Nevertheless,manual segmentation is time-consuming and laborious,and the accuracy of manual analysis mode is relatively limited.Therefore,it is of great magnitude to develop an accurate computer aided diagnosis technology for brain disease.With the rapid development of current deep learning technology,it benefits from the important contribution of Convolutional Neural Networks（CNN）in the field of medical image segmentation.Computer-aided diagnosis technology primarily based on deep learning has made top notch development.But a considerable part of the current research still focuses on using a single model to solve the diagnosis of a single condition.The lack of a universal model solution makes it necessary to use multiple models in turn when diagnosing multiple diseases,which increases the consumption of computer resources in hospitals.In addition,the use of multiple models to diagnose the disease will have contradictory diagnosis problems,which will easily confuse the disease,and further lead to misdiagnosis.Therefore,in view of the aforementioned issues such as misdiagnosis of diseases and high resource consumption.This paper innovatively combines brain CT analysis and MRI multimodal pathological analysis as well as the cutting-edge image semantic segmentation algorithm in current computer vision research.In the follow-up research work,a semantic segmentation model of multi-source brain imaging images based on se Res Ne Xt50-UNet++ is proposed.The model adopts the current advanced activation function,loss function optimization method and optimized backbone network.On this basis,combined with multi-source medical data sets,the segmentation and reasoning of three different pathological contours of brain tumor,ischemic stroke and cerebral hemorrhage can be realized by a single model,which makes it a reality to analyze multiple diseases with a single model.The model effect has also been improved.Moreover,this paper uses Docker containerization technology to develop elastic nodes on the basis of the proposed model.A brain medical image analysis system based on microservice is implemented.The maximum system load is 10 QPS.