摘要
目的对同一患者在不同时间拍摄的眼底图像,通过寻找合适的眼底图像对位频闪方法,以期有助于眼科医生对青光眼、糖尿病性视网膜病变等疾病的诊断。方法利用互信息对眼底图像对位,并且每隔1 s进行交替闪烁。采用2种方法对其评价:一种是通过设定固定图像来检验对位后图像的误差来评价;另一种是由临床专家操作来进行目测检验来评价。结果对50张数码眼底照片,每张重复10次,检验对位后图像的旋转角度和平移量,P>0.05,差异无统计学意义。通过100名随访患者眼底图像的临床应用,眼科专家认为该方法适合临床应用。结论互信息法是眼底图像对位频闪中有效且精度高的一种方法。该方法操作简单,重复性好,对青光眼、糖尿病性视网膜病变等疾病的诊断有重要价值。
Objective Registration flicker on eye fundus images that were screened in different time is important for ophthalmologist to study the glaucoma, diabetic retinopathy. Methods The normalized mutual information could be used to deal with eye fundus images and flicker them every one second. There are two methods to estimate it; one is based on the error of eye fundus images registration, the other is according to testing by ophthalmologist. Results 50 eye fundus images were tested and repeated ten times. Circumvolve and shift value had been analyzed. The result is P〉0.05. 100 eye fundus images had been matched. Conclusion Image registration methods based on mutual information had been accepted as the accurate and efficient method in eye fundus images registration flicker. It is easy to operate and repeat.
出处
《首都医科大学学报》
CAS
2006年第5期672-674,共3页
Journal of Capital Medical University
基金
北京市自然科学基金(3052009
3063024)资助项目
关键词
眼底图像
对位
频闪
eye fundus image
registration
flicker