Retinal image analysis aimed at support of early neural-layer deterioration diagnosis

The contribution aims at designing and testing an automatic method to estimate the status of the retinal neural fibre layer (NFL) based on analysing the output of the most common ophthalmological imaging modality - fundus camera images. As the neural layer manifests itself in these images rather faintly and is often hardly visible, the method has to utilise subtle features discoverable, with a reasonable reliability, only by combining different texture analysis approaches. The texture analysis must be preceded by detecting and removing the eye vessel structure overlaid on the neural layer thus hindering the proper NFL detection. The results of both analytic phases - the vessel structure segmentation and the neural layer detection - were evaluated using sets of both clinically obtained and international standard retinal image databases.