Dalia Zykutė¹, doc. Vilma Jūratė Balčiūnienė², Danielius Visockas³, Jokūbas Liutkus¹

¹Lithuanian University of Health Sciences, Medical Academy, Faculty of Medicine, Kaunas, Lithuania; ² Lithuanian University of Health Sciences, Ophthalmology clinic, Kaunas, Lithuania; ³Vilnius Gediminas Technical University, Vilnius, Lithuania.

Abstract:

Age-related macular degeneration is a leading cause of blindness in the western world. With improved imaging modalities and new advances in treatment with vascular endothelial growth factor inhibitors, it is possible to delay or even prevent vision loss in people with the neovascular form of the disease. However, meticulous surveillance with optical coherence tomography (OCT) is needed in order to achieve improvement of outcomes. This approach generates significant amounts of medical images, thus increasing the workload for retinal specialists. Implementation of artificial intelligence in telemedicine could improve work flow and health care quality. In fact, medical specialties that have imaging-based diagnostic such as dermatology, radiology, pathology and ophthalmology are already implementing artificial intelligence-based diagnostic algorithms into clinical practice. At the forefront of artificial intelligence in computer vision, are convolutional neural networks (CNNs) that are based on deep layers of artificial inter-connected neurons that are able to achieve and surpass the human-level classification of images. The development of such algorithms for specific clinical tasks requires close collaboration between developers and clinicians, with the later requiring deep knowledge of the stages and features of CNN development. Therefore, in this review article we discuss current strategies on the creation and evaluation of artificial intelligence models based on CNNs, as well as specific strategies that are used to classify OCT images using biomarkers of age-related macular degeneration.

Keywords: Age-related macular degeneration, convolutional neural networks, optical coherent tomography.