眼底病灶的计算机图像识别及临床应用可靠性测试

Computer assisted Identification of Fundus Lesions and a Study of Its Reliability

设计眼底病灶识别的计算机程序，并进行可靠性测试。采用“区域生长法”，在ＶｉｓｕａｌＣ＋ ＋ 环境下设计程序；应用该程序测量糖尿病棉絮状渗出灶面积（１５ 只眼），计算变异系数以判断稳定性。结果表明渗出灶面积测量的变异系数为３．２２％ ～９．４６％ 。应用此方法识别并定量测量病灶简便迅速。

This study aimed at the use of computer for identifying fundus lesions quickly and reliably. We adopted the ‘area filling’ method to program in Visula C ++ , used the program to mesure the retina cotton wool patches in diabetics (15 eyes), and then calculated the coefficient of variation to evaluate its reproducibility. The results showed that the coefficient of variation of the measurements ranged from 3.22% to 9.46%. In conclusion, this method is convient, quick and reliable.