Medical Image Annotation And Analysis System Based On Crowdsourcing

With the rapid development of software and hardware research and development in the 21 st century,various medical imaging equipment has been introduced by major hospitals.The medical images that medical patients have come into contact with in the treatment process provide important basis for clinicians to diagnose diseases and determine treatment plans.Common medical image assistant detection systems such as CT(electronic computer fault scanning)Pet(positron emission computed tomography),MRI(MRI),US(ultrasonic imaging)and so on,have significantly improved the efficiency of clinicians’ treatment and saved the cost of diagnosis and treatment time.However,the rapid development of medical imaging field also brings many problems.Firstly,the summary of medical image observation requires a lot of experience and professional knowledge of the imaging doctors,which undoubtedly increases the learning cost of clinicians.Secondly,even for doctors with rich experience and knowledge,it is very time-consuming to consult,diagnose and report large numbers of medical images,which also increases the time cost of clinicians.In view of the above problems,combined with the research and application of the increasingly active group intelligence field,this paper proposes and implements a medical image annotation and analysis system based on crowdsourcing idea and principle.The main work contents are as follows:(1)A set of medical image annotation scheme based on crowdsourcing mechanism is achieved.This system integrates the crowdsourcing task mechanism and the annotation and diagnosis characteristics of medical images,and builds an application system platform which supports the functions of digital acquisition,slice storage and crowdsourcing.(2)The mechanism of role-based authority management is established.The permission control of the system is based on the role of the user.The access permission package of resources is associated with the role.The user has a role,and the corresponding permissions associated with the role,which makes the user,role and system authority decoupled.The permission configuration and role configuration are reusable.(3)The paper explores the application of tensorflow or pytorch export model in the field of machine learning to carry out intelligent diagnosis of medical images,and supports task results export and offline machine learning model training.