ESPE Yearbook of Paediatric Endocrinology

Nature Digital Medicine 2018;1:39.

DOI: 10.1038/s41746-018-0040-6

Summary: This prospective observational study evaluated the performance of a diabetic retinopathy diagnostic system (IDx-DR) compared to the gold standard diagnostic for diabetic retinopathy. Nine hundred individuals with diabetes but without a history of diabetic retinopathy were examined. Retinal images of the patients were obtained using a robotic camera, and a clinical diagnosis was made in 20 seconds by an artificial intelligence (AI) diagnostic system and compared to images read by three experienced and validated readers. The AI system correctly identified 173 of the 198 individuals with more than mild diabetic retinopathy, (a sensitivity of 87%), and 556 of the 621 disease-free individuals (a specificity of 90%).

Comment: This breakthrough study describes the first FDA authorized autonomous AI diagnostic system in any field of medicine. AI is the simulation of human intelligence processes by computer systems. These processes include learning (the acquisition of information and rules for using the information), reasoning (using rules to reach approximate or definite conclusions) and self-correction. One of the techniques used in AI is referred to as “deep learning”.¹ These methods have dramatically improved the state-of-the-art in visual object recognition, object detection speech recognition, language translation and many other domains such as drug discovery, genomics, robotics and even self-driving cars.

In medicine, the impact of AI is categorized into three branches²: 1) for clinicians, 2) for health systems and 3) for patients. For clinicians, deep learning is helpful in interpreting medical scans, pathology slides, skin lesions, electrocardiograms, endoscopy and faces. For health systems that use electronic record data, deep-learning algorithms enable predicting several health outcome parameters and improving workflow algorithms. For patients, deep-learning algorithms enable accessing their own data, such as smartwatch algorithms to detect atrial fibrillation and continuous sensing of blood-glucose. Despite the controversy involving AI, this article attests to the beginning of a new and exciting era.

References: 1. LeCun Y, Bengio Y, Hinton G. Deep learning. Nature. 2015; 521(7553): 436–44.

2. Topol EJ. High-performance medicine: the convergence of human and artificial intelligence. Nature Medicine. 2019; 25(1): 44–56.