Method Hand-Driven Used for Features Extraction in OCT B-Scan Images Processed

AMD macular degeneration and glaucoma stand out among the different types of pathologies that exist due to vision loss. The objective of this work is to use a dataset of images of patients with AMD pathology, which were semantically segmented, and to be able to perform the extraction of characteristics of the RNFL layer of the image with artificial vision and statistical methods, to carry out an experimental predictor 168 semantically segmented photos of patients with AMD were used. Through artificial vision, the fibers of the retina, known as RNFL, were highlighted. The Hand-Driven method was applied, which extracts an image’s one-dimensional and two-dimensional characteristics; the Information was stored in a matrix, and discriminatory variables were formed. An experiment with a multilayer perceptron was proposed using the matrix of discriminating variables as input. Two outputs were classified; this experiment achieved a 93.94% effectiveness, and the feature extraction's utility was verified. It was concluded that the Hand-Driven method is robust with great functionality.