Smart Medical Information Management System Based On The Validity Of Small Intestine CT Images

At present,the technology of computer-aided diagnosis is becoming more and more mature,and it is necessary to input valid data sets when performing auxiliary diagnosis.Therefore,it is particularly necessary to check the validity of the data before uploading the data to the system.Based on this,this paper focuses on the small intestine CT image data The following studies were conducted on the legality check issue:(1)Establishment and preprocessing of small intestine CT image data set.For the study of small intestine,due to the problem of data privacy protection in medical images,there is no public small intestine CT image data set at present.Therefore,our team cooperated with a medical department to solve the problem of data privacy.After solving the problem of data privacy,we collected abdominal tomographic CT data of 100 patients and evaluated the quality of the data.We selected some slice data with a stable state and clear outline with a resolution of 512 times 512.After all data screening,sorting,data denoising and data enhancement,we completed the establishment of small intestine CT image data set.(2)Constructed an improved lightweight network Mobile Net V2 based on small intestine legitimacy inspection.Based on the data set established in this paper,we use networks such as Alex Net,VGG-16,Res Net-50,Mobile Net V1 and Mobile Net V2 for training.After comparing the experimental results,we finally chose the lightweight convolutional neural network Mobile Net V2 with the least amount of parameters,the fastest running speed,and the second highest accuracy rate as the basic network of this paper;Due to the irregular features of the small intestine and the scattered distribution in the image,the problem of feature loss occurs when using traditional convolution to extract features.Therefore,this paper chooses deformable convolution that can adaptively receptive fields and extract irregular features to replace the deep convolution in the inverted residual block structure of the original Mobile Net V2 network.And because of the small amount of data,the migration learning training method is adopted,and the final accuracy rate reaches 94.08%,which is an increase of 2.13% compared with the original network.Finally,the classification performance has been significantly improved.(3)Build a medical information management system based on Java Web.This paper integrates the small intestine legitimacy check algorithm into the system,and realizes the legitimacy screening function of the small intestine data by calling the pre-trained model inside the server.Finally,a smart medical information management system integrating registration and login module,user information management module,announcement release and browsing module,data browsing module,data uploading and model calling module,doctor-patient message and interaction module is completed.