Design And Implementation Of Medical Image Recognition System Based On U-net

The dramatically improvement in image and information technology field not only affects people's lives and industry management modes in some common areas such as information management,but also significantly improves the efficiency of all aspects of our life.In the medical field,due to the current high disparity in the proportion of doctors and patients,more and more doctors need to pay more attention on patients' medical tasks,in order to reduce the doctor's detection workload for CT and ultrasound images,and remote communication and interaction between patients and doctors.The paper designed and implemented an medical image recognition system based on deep learning semantic segmentation of CT images and ultrasound screening images.In this paper,we design and implement a U-net model based medical image recognition system according to the requirements of medical workers and patients.In this system,the U-net network structure with mature effects in semantic segmentation was used,and the Inception V3 structure and the residual block structure were integrated.In the U-net network,the ELU activation function is used instead of the previously used ReLU activation function to implement an improved image model,and the average accuracy of the training results is improved.In the medical image recognition system,this algorithm model is an important module.The system is built based on Python as the development language,and the MySQL 5.7 as the system database,and we using the more advanced and fast Django framework version 1.9.0 as the Web development tool.The framework realizes the main functions of hospital information management,predictive model management and problem consultation.This system has achieved good interaction with users.After the final test,the algorithm part of the system meets the accuracy requirements of the demand.For the average accuracy of the cervical vertebra ultrasound image and the lumbar CT image,the accuracy is improved from the original 69.39% and 67.92% to 71.88% and 70.96%.The medical image recognition system based on this semantic segmentation algorithm can stably handle various aspects of medical workers and patients,and can perform rapid image segmentation processing within the scope of requirements.The test version of system has been used by relevant user groups,which can improve the communication efficiency of doctors and patients and improve the relevant work efficiency of medical workers,and achieve design requirements.