眼底攝影之動靜脈分析及測量

Abstract

視網膜血管是人體唯一可以直接且用非侵入式的方法看到的血管結構，視網膜血管反應系統微血管的狀況像是腦部、心臟和腎臟，之前的研究指出血管的管徑和系統心血管疾病有關聯性。在這篇論文中，提出在視網膜圖片上測量血管管徑和分類動靜脈的自動量化分析，為了得到更準確的血管管徑和分類動靜脈，根據動態螢光眼底影片(dynamic fluorescein angiography)和螢光眼底圖片(fluorescein angiography)的量化測量方式和傳統的彩色眼底圖片(color fundus photography)的方法比較，利用視神經盤和血管的切割的結果來偵測血管，接著，利用顏色強度的特徵分類動靜脈。在動態螢光眼底影片中，在主幹血管上的血管分類偵測率和正確率可以達到100%，相較於傳統的彩色眼底圖片血管分類偵測率97.83%和正確率86.67%，動態螢光眼底影片可以得到較準確的結果。總結來說，在視網膜血管管徑的計算和分類動靜脈上提出量化的方法，使其提供準確的測量結果，對於未來的分析動靜脈管徑的比值和系統心血管疾病的關聯性上有很大的幫助。

Blood vessels of retina are the only vascular structure in the body that can be seen directly and non-invasively. Retinal vessels reflect the conditions of systemic capillaries such as in brain, heart, and kidney. Previous studies have reported the association between vascular diameter and systemic vascular disease. In this study, an automatic quantitative analysis which measured the vascular diameter in retina images and classified the vessels into arteries and veins was proposed. In order to get more accurate vessel width and vessel classification, the measurement based on dynamic fluorescein angiography (DFAG) and fluorescein angiography (FAG) were quantified and were compared with that of the conventional color fundus photography (CFP). Optic disc and vessel segmentation were used to detect the number of vessels. Then, intensity features were extracted for vessel classification. In the DFAG, the detection rate and accuracy of vessel classification achieved 100% for the trunk vessels that were better than those in the CFP (97.83% and 86.67%). Summarily, the proposed quantitative method for retinal vessel width calculation and vessel classification provides an accurate measurement for the future analysis in the correlation between arteriovenous ratio and the systemic vascular diseases.