【摘要】 通过任务转换实验(task-switching task,TST)与机器学习相结合的方法对精神分裂症患者进行客观辨别。本实验中,45例精神分裂症患者和55例正常被试参与了实验,原始特征为反应时间、准确率、转换代价、干扰、年龄和性别等17个特征,利用Relief算法进行特征选择,并利用支持向量机(SVM)对不同的特征组合进行分类。研究结果表明短提示一致有转换、短提示一致无转换、短提示不一致无转换、长提示一致有转换、长提示一致无转换、转换代价、剩余转换代价和反应转换代价8个特征创建的SVM分类模型达到了90%的分类准确率,ROC曲线下面积(AUC)大小为0.89.该方法能够客观、有效地鉴别精神分裂症患者,能运用于精神分裂症的辅助诊断,以提高医生诊断结果的准确率。更多还原

【Abstract】 Schizophrenia is a chronic and complex heavy psychosis.Its barriers on thought,emotion,behavior and so on seriously affect the patient’s life.In addition to the assessment of the clinical scale,there is still a lack of an objective and convenient method for discriminating patients with Schizophrenia.Therefore,the research of auxiliary diagnostic techniques for schizophrenia has become a hot spot.This study objectively identified Schizophrenia patients by combining task switching(task-switching task,TST)and machine learning.In this study,45 patients with Schizophrenia and 55 normal subjects were involved in the experiment.The original features included 17 characteristics of reaction time,accuracy,switch cost,interference,age and gender.Relief algorithm was used to make feature selection and the support vector machine was used to classify different feature combinations.Research results show that the SVM classification model based on the 8 features,that is,short cue congruent switch,short cue congruent without switch,short cue incongruent without switch,long cue congruent switch,long cue congruent without switch,switch cost,residual switch cost,and response switch,reached a classification accuracy rate of 90% .The area under ROC curve(AUC)was 0.89.The experimental results show that this method can objectively and effectively identify Schizophrenia patients,and it can be applied to the auxiliary diagnosis of Schizophrenia,so as to improve the accuracy of diagnosis.更多还原