Segmentação e extração das características do epitélio pigmentado da retina em imagens de OCT para detecção precoce de DMRI

Abstract

The increase in life expectancy over recent decades has brought new challenges to global public health, particularly concerning diseases associated with population aging. Among these conditions, Age-related Macular Degeneration (AMD) stands out as one of the leading causes of progressive vision loss and irreversible blindness in the elderly. The diagnosis of AMD, when performed through ophthalmologic examinations such as Optical Coherence Tomography (OCT), constitutes a non-invasive imaging technique that enables the acquisition of high resolution cross-sectional images of the different retinal layers. This technique allows for a detailed assessment of retinal structures and the identification of morphological changes, such as the accumulation of drusen, which is one of the main biomarkers of the disease. Early detection of AMD is essential for clinical monitoring and for planning interventions that may slow its progression. Based on this context, the present study proposed an automatic and unsupervised algorithmic strategy for the segmentation of the retinal pigment epithelium, aiming to support the diagnosis of drusen. The method was developed using fundamental digital image processing techniques. The algorithm’s performance was evaluated through shape signature characterization and contour reconstruction using the Fourier transform. The results obtained demonstrated significantly satisfactory performance, with an accuracy of 89.8%, sensitivity of 90.4%, and specificity of 89.2%, highlighting the potential of the proposed methodology as a clinical decision support tool.