Korpinen L, Pietilä T, Peltola J, Nissilä M, Keränen T, Touvinen T, Falck B, Petránek E S, Frey H
Department of Neurology, University of Tampere, Finland.
Comput Methods Programs Biomed. 1994 Nov;45(3):223-31. doi: 10.1016/0169-2607(94)90206-2.
Epilepsy Expert is a decision support system based on the International Classification of Epilepsies and Epileptic Syndromes (1989). The aim of this study was to evaluate the Epilepsy Expert. First the diagnostic performance was validated. This was done in 3 stages: collection of the patient cases, determination of the 'correct diagnoses' and testing the system. How the users perceived the functionality of the system was studied by using an inquiry. Three physicians, experts of epilepsy, from different hospitals were asked to choose 10 patients. In the patient description was a short history, a detailed description of the seizure, EEG findings and their own diagnosis. Next, each expert made a diagnosis of the cases supplied by other experts by using the International Classification of Epilepsies and Epileptic Syndromes. The 'correct diagnosis' (so-called majority agreement) was the diagnosis given by the majority of the experts. The diagnosis of each expert was compared with the 'correct diagnosis'. The diagnoses obtained by the Epilepsy Expert were then compared with the 'correct diagnoses'. In the evaluation the expert physicians agreed on 37% of cases and all 3 disagreed on 17%. A majority agreed on 25 cases, which were used in the evaluation. In these 25 cases the experts' (A,B,C) diagnoses were correct or partly correct in 100, 64, 80% of cases, respectively. The program's diagnoses were correct or partly correct in 80% (module I) and 76% (modules IV and V) of cases. In the evaluation Epilepsy Expert was found to be only partly successful. The main reason for this was the weakness of the international classification. However, the program seems to be very close to the level of the experts. According to this limited inquiry Epilepsy Expert is not suitable for clinical use, because it is, for example, too simple and does not contain enough information.
癫痫专家系统是一个基于《国际癫痫及癫痫综合征分类(1989年)》的决策支持系统。本研究旨在评估癫痫专家系统。首先对其诊断性能进行验证。这分三个阶段进行:收集患者病例、确定“正确诊断”并测试该系统。通过问卷调查研究用户对系统功能的看法。邀请了三位来自不同医院的癫痫专家医生,让他们每人挑选10例患者。患者描述包括简短病史、发作的详细描述、脑电图检查结果以及他们自己的诊断。接下来,每位专家依据《国际癫痫及癫痫综合征分类》对其他专家提供的病例进行诊断。“正确诊断”(即所谓的多数人一致意见)是大多数专家给出的诊断。将每位专家的诊断与“正确诊断”进行比较。然后将癫痫专家系统得出的诊断与“正确诊断”进行比较。在评估中,专家医生在37%的病例上达成一致,在17%的病例上三人意见全部不同。有25例得到多数人认同,用于评估。在这25例病例中,专家A、B、C的诊断分别在100%、64%、80%的病例中正确或部分正确。该程序的诊断在80%(模块I)和76%(模块IV和V)的病例中正确或部分正确。在评估中发现癫痫专家系统仅部分成功。主要原因是国际分类存在缺陷。然而,该程序似乎非常接近专家水平。根据这项有限的调查,癫痫专家系统不适合临床使用,因为它例如过于简单且包含信息不足。