Research On Consistency Detection Methods For Chest X-ray Diagnostic Reports

Chest X-ray has become the priority of patients’ chest examination.It can quickly and clearly observe the general situation of chest,including lung,heart and other organs.Chest X-ray diagnosis report has an important application value.It can not only give early warning of possible diseases,but also be an important reference for clinicians to formulate treatment plans.The core content of the diagnosis report is image description and diagnosis conclusion.These two parts are important references to assist doctors’ diagnosis and patients’ treatment.The doctor’s writing of the diagnosis report is quite subjective,which may lead to the interpretation error of the image description content due to lack of experience or fatigue,which may lead to missed diagnosis and misdiagnosis of some diseases.In addition,the description of the image in the diagnosis report is free,mostly the description language of medical convention.The complicated image description content may also affect the differential diagnosis of doctors and draw wrong diagnosis conclusions.Screening out these abnormal diagnosis reports whose image descriptions are inconsistent with the diagnosis conclusions can first reduce the misdiagnosis rate of the disease and provide more accurate and effective reference for the diagnosis and treatment of the clinicians.Secondly,it provides the basis for the establishment of standardized medical examination system and the realization of efficient and accurate medical services.Finally,enhance the management level of the hospital,supervise and inspect the technical literacy of medical workers.Considering the characteristics of the diagnosis report,the paper first maintains the original semantic information through entity recognition,and then on this basis,carries out the consistency detection of the diagnosis report.In this paper,the chest X-ray diagnosis report is taken as the research object to test the consistency of the diagnosis report.Aiming at the problem of entity recognition of diagnosis report,according to the characteristics of diagnosis report,this paper takes the word joint vector and the part of speech and suffix features of diagnosis report as input,and uses the two-way short-term memory network combined with conditional random field model to identify the entity of diagnosis report.It overcomes the problems of large amount of data,many unlisted words,no corresponding dictionary and long symptom entity.The experimental results show that the method has high accuracy and recall rate for entity recognition of diagnostic reports.Aiming at the problem of consistency detection in the diagnosis report,this paper transforms the consistency detection of the chest X-ray diagnosis report into the problem of whether the entity features described in the image can match the entity features in the diagnosis conclusion.Firstly,this paper extends the entity features of the diagnosis report,then uses the improved LDA model to judge whether the image description and the diagnosis conclusion features match in the diagnosis report,and finally determines the threshold to distinguish whether the diagnosis report is consistent,and completes the consistency detection of the diagnosis report.The experimental results show that the performance of this method is better than that of traditional methods.