Author: Daniel Ting, MD, PhD

The real-world application of artificial intelligence (AI), machine learning (ML) and deep learning (DL) has generated significant interest among the computer science and medical communities in recent years. DL has revolutionized the computer vision field and achieved substantial jumps in diagnostic performance for image recognition, speech recognition, and natural language processing. In Ophthalmology, this technique has shown promising diagnostic performance in detection of diabetic retinopathy (DR), glaucoma and age-related macular degeneration from fundus photographs and OCTs. Given the surge in the number of publications in the literature, it is, therefore, important to understand the clinical and technical considerations in building a DL-based AI system. This paper will focus on 2 aspects – 1) clinical aspect: the unmet needs and; 2) technical aspect the concepts of convolutional neural networks (CNN), the data distribution and characteristics for training, validation and testing, reference standards, performance metrics and the methods to explain diagnosis.