论文部分内容阅读
为解决目前临床上对早期糖尿病视网膜病(diabetic retinopathy,DR)诊断困难的问题,利用多焦视网膜电流图(multifocal electroretinogram,mfERG)和模式分类的方法,对DR进行早期诊断和自动分类的研究。对31例正常样本和58例早期DR样本,提取mfERG信号的平均一阶响应(K1成分)峰值和模板相关系数两个关键特征。使用线性分类器进行分类,并用“留一法”统计结果,正常人和早期DR病例的分类错误率为21.35%。该文还用聚类方法分析了早期DR样本的分布特点。结果显示,使用平均一阶响应的N2峰的幅值和位置作为特征对早期DR的诊断效果较好;早期DR样本的聚类结果具有一定的临床价值。
To address the current clinical difficulties in the diagnosis of early diabetic retinopathy (DR), the study of early diagnosis and automatic classification of DR was conducted using multifocal electroretinogram (mfERG) and pattern classification. For 31 normal samples and 58 early DR samples, two key features of the average first-order response (K1 component) and template correlation coefficient of mfERG signal were extracted. Using the linear classifier to classify and use the “leave one method” statistical results, the classification error rate of normal and early DR cases was 21.35%. The paper also uses the clustering method to analyze the distribution characteristics of early DR samples. The results show that the average first-order response of the peak amplitude and location of N2 as a feature of early DR diagnosis is better; early DR clustering results have some clinical value.