The Design And Implementation Of Medical Image Augmentation System Based On Mutation Analysis

With the increasing accuracy and efficiency requirements of medical image diagnosis,as well as the rapid development of hardware computing power and algorithm model,artificial intelligence technology has been applied more and more widely in the field of medical image.Meanwhile,the number and quality of datasets are the key factors that affect the performance of the intelligent medical diagnostic model.Insufficient and poor quality datasets can negatively impact the training and testing results of the model,which affect the effectiveness of the deep learning model,especially in securitycritical areas such as medical field.It may cause serious accidents and losses,so the demand for data augmentation is even more compelling.However,due to data standardization and privacy restriction,the acquisition and sharing of medical image data have been hindered.Therefore,how to effectively generate a large number of reliable medical image dataset has been an urgent problem to be solved.This thesis implements the medical image augmentation system,which mainly provides the intelligent augmentation solutions and multidimensional quality evaluation for medical image.Because of the overall similarity and partial diversity of medical images,the system designs and implements medical image augmentation technology based on image feature mutation and domain semantic mutation by combining lesion signs and mutation analysis,so as to obtain its potential information and generate rich and diverse medical image datasets.At the same time,we propose a multi-dimensional evaluation system of image augmentation quality.The evaluation indexes are used to analyze and evaluate the augmentation effect of the system from multiple dimensions,and the visualization results are generated by using the Echarts library.Based on the Spring Boot framework,the system provides users with an interactive interface to manage datasets and augmentation tasks,while calling image augmentation service and quality evaluation service through Docker.At present,the system implements 21 augmentation methods and provides users with differentiated custom augmentation schemes.Based on the Deep Lesion dataset and Res Net-50 image classification model,experiments are carried out in the four dimensions of data volume,method setting,augmentation effect and augmentation quality.The results indicate that the dataset generated by the intelligent medical image augmentation system can alleviate the difficulty of data collection on the one hand,and the quality of the augmented dataset can be guaranteed to a certain extent.On the other hand,it can improve the effect of the medical image assisted diagnostic model.In all experiments,the accuracy of the model has an average improvement of about 5%.