人工智能辅助超声危急值临床诊断的初步应用研究

Preliminary application study of artificial intelligence-assisted ultrasound in critical value clinical diagnosis

目的研究一款智能超声危急值管理系统并探讨其在临床应用的价值。方法运用H5技术研发一款基于微信平台的超声危急值智能监测软件,建立超声危急值数据库。通过智能识别超声危急值,进行危急值预警。收集就诊疑似腹部超声危急值的82例患者,同时进行人工智能(AI)诊断(AI组)及专业超声医生诊断(医生组)。以手术病理或临床医师最终的综合诊断为金标准,比较人工报告与软件智能识别2种诊断结果的灵敏度、特异性、阳性预测值以及诊断准确率。结果医生组诊断准确率93.90%、阳性预测值97.37%、灵敏度96.10%,AI组诊断准确率90.24%、阳性预测值96.00%、灵敏度93.51%,2组比较,差异无统计学意义(均P>0.05)。结论AI组与医生组的诊断超声危急值具备足够的准确度,可协助医生判读超声报告,提高诊断的准确性,有助于危急重症患者的救治。

Objective To study an intelligent ultrasound emergency value management system and explore its value in clinical application.Methods Use H5 technology to develop an intelligent monitoring software based on WeChat platform and establish ultrasonic critical value database.Through the intelligent identification of ultrasonic emergency value,emergency value warning.A total of 82 cases to be diagnosed as common abdominal ultrasound critical value were collected,and AI diagnosis(AI group)and professional sononologist diagnosis(doctor group)were performed simultaneously.Overall final diagnosis of surgical pathology or clinician was taken as the gold standard.To compare the sensitivity,specificity,positive predictive value and diagnostic accuracy of manual report with software intelligence.Results Diagnosis accuracy rate was 93.90%,positive prediction value 97.37%,and sensitivity 96.10%.Group AI,diagnostic accuracy was 90.24%,positive predictive value 96.00%,and sensitivity 93.51%.There was no significant difference between the two groups(P both>0.05).Conclusion The diagnosis of AI group and doctors has sufficient accuracy to assist doctors to interpret the ultrasound report,improve the diagnostic accuracy,and facilitate the treatment of critical patients.