Introduction:

The high prevalence of Diabetes Mellitus type 2 in Mexico has positioned diabetic retinopathy as the main cause of blindness in adults of productive age in Mexico. Therefore, the timely detection of this disease is a priority task for the public health system. This article studies the efficiency of a new algorithm for determining the shape of the Major Temporal Arcade of the retina, using image segmentation techniques and numerical modeling of curves.

Method:

The proposed methodology uses Gaussian Matched Filters that enhance the geometry of the blood vessels. Subsequently, the vascular structure is segmented by global thresholding of the enhanced image. Said segmentation is used as input to build a numerical model of the Superior and Inferior Temporal Arcades, using Spline functions.

Results:

The performance evaluation was carried out using 136 images of $6000\times 4000$ pixels. The automatic retinal vein segmentation algorithm using the GMF method obtained an Accuracy of 0.9852; the numerical modeling algorithm gave a result of 6.01 pixels for the metric Mean Distance to the Closest Point (MDCP). Another previous study reported 12.33 pixels. Regarding time, the method reported an average time of 10.65 seconds per image.

Discussion:

The proposed method was able to carry out the numerical modeling of temporal arches in fundus images efficiently. The results show that this method is an useful computational tool for the diagnosis of alterations in the anatomy of the eye.

        · resumo em Espanhol     · texto em Espanhol     · Espanhol ( pdf )