A Multiclass Age-Related Macular Degeneration Classification

Implementing artificial intelligence in the image classification process can accelerate the diagnosis and monitoring of eye diseases, lowering the cost and time burdens on both the patient and the treatment staff. This paper focuses on distinguishing three eye conditions including normal eyes, and two eye-related diseases; Dry Age-Related Macular Degeneration (AMD) and Wet AMD. To reach this goal, Optical Coherence Tomography images of these conditions will be used based on the data collected from Negah Hospital. Features of the images are extracted after preprocessing as sequences and then classified by Recurrent Neural Networks. Pre-trained convolutional neural networks including VGG۱۹, InceptionV۳, and Xception have been used for modeling. The trained models in this project achieved ۹۲ percent accuracy. The results present a clear look at the challenging nature of diagnosing certain conditions from each other. Challenges of medical data collection and integration of artificial intelligence in the diagnosis process will also be discussed.

کلیدواژه ها:

Transfer Learning ، Medical image classification ، Age-Related Macular Degeneration ، Optical Coherence Tomography ، Convolutional Neural Networks