基于深度学习方法的精神分裂症听觉稳态诱发电位分析

Auditory Steady State Responses of Schizophrenia Based On Deep Learning Algorithm

目的文本旨在使用深度学习方法对精神分裂症建立自动分类模型,为临床上精神分裂症患者的鉴别诊断提供参考。方法通过提取受试听觉稳态诱发电位的能量、相位、信噪比和微分熵作为模型输入特征,使用准确率、灵敏度、特异度和受试工作者特性曲线对深度信念网络和支持向量机构建的模型进行分析和比较。结果深度信念网络模型的准确率、灵敏度、特异度、曲线下面积分别为85.6%、88.33%、75.50%和0.88。深度信念网络模型诊断能力明显高于基于线性核、径向基函数核、sigmoid核的三种支持向量机模型。结论基于深度信念网络算法的诊断模型可以有效协助临床医生对于精神分裂症患者的诊断,达到早期发现疾病的效果。

Objective To establish an automatic classi fication model for schizophrenia using deep learning method, and to provide a reference for the diagnosis of schizophrenia in clinical practice. Methods We extracted the energy, phase, signal-to-noise ratio and differential entropy of auditory steady state responses as input features of models and compared the performance of deep belief networks (DBNs) and support vector machines (SVM) using accuracy, sensitivity, speci ficity and receiver operating characteristic curve. Results The accuracy, sensitivity, specificity and area under the curve of DBNs model were 85.6%, 88.33%, 75.50% and 0.88, respectively. The diagnostic capacity of DBNs model was significantly higher than that of three kinds of SVM models based on linear kernel, radial basis function kernel and sigmoid kernel. Conclusion The diagnostic model based on DBNs can effectively assist clinicians in the diagnosis of patients with schizophrenia and achieve the early detection of disease.