Analiza imaginilor de retină folosind algoritmi de învățare profundă / Retinal image analysis using deep learning algorithms

The paper presents the automatic detection of glaucoma from retinal images using deep learning algorithms. This thesis relies on the utilization of deep neural networks for the analysis and interpretation of retinal images, aiming to identify the characteristic signs of glaucoma. Deep learning algorithms, particularly Convolutional Neural Networks (CNNs), are trained using proprietary datasets of retinal images manually labeled to indicate the presence or absence of glaucoma. During training, these algorithms learn to recognize specific patterns and features associated with glaucoma, such as optic nerve changes, intraocular pressure, and vascular aspects. The focus of the work is on training and employing models like ResNet50, DenseNet121, and DenseNet201 for the analysis and interpretation of retinal images to identify distinctive signs associated with glaucoma.However, there are challenges and considerations in implementing and using these algorithms in medical practice. Precise validation of the performance of these models is crucial, and ensuring their reliability and accuracy is essential to avoid misdiagnosis in identifying glaucoma. Thus, the continuous importance of research is emphasized for the improvement and refinement of these algorithms for a more precise and efficient diagnosis of glaucoma