Automated Analysis of the Distributions and Geometries of Blood Vessels on Retinal Fundus Images

We have developed a computer-aided diagnosis system to detect the abnormalities on retinal fundus images. In Japan, ophthalmologists usually detect hypertensive changes by identifying narrowing arteriolae with a focus on an irregularity. The purpose of this study is to develop an automated method for detecting narrowing arteriolae with a focus on an irregularity on retinal images. The blood vessel candidates were detected by the density analysis method. In blood vessel tracking, a local detection function was used to go along the centerline of the blood vessel. A direction comparison function using three vectors was designed to provide an optimal estimation of the next possible location of a blood vessel. After the connectivity of vessel segments was adjusted based on the recognized intersections, the true treelike structure of the retinal blood vessels was established. The abnormal blood vessels were finally detected by measuring their diameters. The comparison between the results obtained using our system and the diagnostic results of physicians showed that our proposed method automatically detected an irregularity in diameter in 75% of all 24 narrowing arteries with a focus on an irregularity on 70 retinal fundus images. Approximately 2.88 normal vessel segments per image were determined to be abnormal, a number which must be reduced at the next stage. The automated detection of narrowing arteriolae with a focus on an irregularity could help ophthalmologists in diagnosing ocular diseases.