2019/01/26 by Rony Gelman, Gelman, Rony · 1 citation
Computer Science · Health Professions · Medicine · #Artificial Intelligence in Healthcare #Digital Imaging for Blood Diseases #FOS: Computer and information sciences #Machine Learning (cs.LG) #Machine Learning (stat.ML) #Retinal Diseases and Treatments #Retinal Imaging and Analysis
paper · pdf · doi:10.48550/arxiv.1902.04151
openalex publication_date 2019/01/26 · openalex created_date 2022/07/30 · openalex updated_date 2026/07/28
Deep learning has been successfully applied to a variety of image\nclassification tasks. There has been keen interest to apply deep learning in\nthe medical domain, particularly specialties that heavily utilize imaging, such\nas ophthalmology. One issue that may hinder application of deep learning to the\nmedical domain is the vast amount of data necessary to train deep neural\nnetworks (DNNs). Because of regulatory and privacy issues associated with\nmedicine, and the generally proprietary nature of data in medical domains,\nobtaining large datasets to train DNNs is a challenge, particularly in the\nophthalmology domain.\n Transfer learning is a technique developed to address the issue of applying\nDNNs for domains with limited data. Prior reports on transfer learning have\nexamined custom networks to fully train or used a particular DNN for transfer\nlearning. However, to the best of my knowledge, no work has systematically\nexamined a suite of DNNs for transfer learning for classification of diabetic\nretinopathy, diabetic macular edema, and two key features of age-related\nmacular degeneration. This work attempts to investigate transfer learning for\nclassification of these ophthalmic conditions. Part I gives a condensed\noverview of neural networks and the DNNs under evaluation. Part II gives the\nreader the necessary background concerning diabetic retinopathy and prior work\non classification using retinal fundus photographs. The methodology and results\nof transfer learning for diabetic retinopathy classification are presented,\nshowing that transfer learning towards this domain is feasible, with promising\naccuracy. Part III gives an overview of diabetic macular edema, choroidal\nneovascularization and drusen (features associated with age-related macular\ndegeneration), and presents results for transfer learning evaluation using\noptical coherence tomography to classify these entities.\n