Maestría en Analítica Aplicada · 2026

Cone density estimation in AOSLO images using image processing and deep learning

Resumen

In ophthalmology, early detection of degenerative diseases in the eye is crucial, however, many times conventional clinical cameras do not allow quantification of retinal cell loss and, in addition, manual analysis of these images is inefficient and poorly automated. To address this problem, techniques using deep learning models for image processing and machine learning are proposed to provide an accurate and automated estimation of cone density to improve the analysis of Adaptive Optics Scanning Laser Ophthalmoscopy (AOSLO) images, which enables detailed visualization of the fundus and individual cells without invasive procedures. This approach enables earlier and more accurate diagnosis of genetic eye diseases such as retinitis pigmentosa and Stargardt’s disease. Index Terms—AOSLO, Cone Density Estimation, Deep Learn ing, U-Net, Lightweight Architecture, Medical Image Analysis, Retinitis Pigmentosa, Stargardt Disease