| Abstract [eng] |
The aim of this work is to research an automatic method for detecting Age-related Macular Degeneration (AMD) lesions in RGB eye fundus images. For this, we align invasively obtained eye fundus contrast images (the “golden standard“ diagnostic) to the RGB ones and use them to hand-annotate the lesions. Using the data, we tain and test five convolutional neural networks: ResNet50, ResNet101, MobileNetV3, UData for segmentation and a custom convolutional neural network of 5 layers for classification task. After analysis of each one neural network we concluded that the best neural network for semantic segmentation was MobileNetV3 with 62,99 % sensitivity, 97,86 % specificity, 93,55 % accuracy and Dice coefficient of 67,76 %. |
