Industry News
AI Could Soon Assist Ophthalmologist Appointments
Three of the most common chronic eye diseases require regular check-ups and injections into the eye by retinal specialists to avert impending blindness. A study by the University of Bern and the Inselspital, in cooperation with a startup for artificial intelligence (AI) applications in ophthalmology, shows that patients' ideal individual examination frequency can be predicted quite precisely using machine learning.
Age-related macular degeneration (AMD) is the leading cause of vision loss in people over the age of 50. This chronic disease affects up to 12 percent of people over 80. Retinal vein occlusion (RVO) - caused by a retinal vein thrombosis - affects an estimated 16.4 million adults worldwide. It is the second most common cause of blindness due to retinal vascular diseases after diabetic retinopathy (DR). On the other hand, DR is the main cause of blindness in industrialized nations and affects approximately four out of five diabetics after 20 years. A DR can lead to diabetic macular edema (DME) with partial or complete loss of vision.
All three diseases are treated with injections of a so-called anti-vascular, endothelial growth factor (anti-VEGF) into the eye at regular intervals to slow the progression of the disease and prevent blindness. Since a central human sense is at risk with eyesight, patients want to be sure that they are treated often enough to avoid rapid deterioration. And doctors want to make sure that they see everyone affected often enough not to miss any important changes.
As society ages, cases of AMD, RVO, or DME are increasing worldwide, making it difficult for specialist eye clinics to keep up with the growing demand for regular check-ups. “As doctors, we want to give all patients the attention and frequency of treatment they need,” says Sebastian Wolf, head of the Inselspital's eye clinic, which currently carries out almost 6000 treatments a year for patients with AMD, RVO and DR. "It is an organizational challenge to do justice to all patients and to collect all relevant imaging data in a short period of time to assess the individual course of the disease and make treatment decisions."
Optical coherence tomography (OCT), an imaging technique that generates extremely high-resolution 3D images of the eye, is usually used to monitor the progression of chronic eye diseases. In cooperation with the ARTORG Center for Biomedical Engineering Research at the University of Bern, Inselspital has developed AI-based, automated OCT analysis systems that can support ophthalmologists in assessing an entire patient OCT set in a few seconds. Together with RetinAI, a startup specializing in AI-based technologies for ophthalmology, they have now conducted a retrospective study to assess how well AI can predict anti-VEGF treatment needs from the start.
The study examined OCT data from 340 patients with AMD and 285 with RVO or DME who were treated with anti-VEGF at Inselspital between 2014 and 2018. Based on morphological features, which were automatically extracted from the OCT volumes at the beginning and after two consecutive visits, as well as demographic information of the patients, two AI models based on machine learning processes, were trained to determine the likelihood of long-term treatment needs Predict a new patient (one for AMD and one for RVO and DME).
Based on the first three visits, it was possible to predict with similarly high accuracy whether patients had a low or high treatment need for both the AMD and the RVO & DME group. Most importantly, the study showed that it is possible, on the first visit, and even before the first injection, to make a relatively good prediction of whether a patient will need injections less often.
"We have shown that machine learning classifiers can predict the need for treatment when a patient is diagnosed with a chronic eye disease for the first time," says Mathias Gallardo, postdoc at the ARTORG AI in Medical Imaging (AIMI) Lab and member of the new Center for Artificial Intelligence in Medicine (CAIM) entered the study results.
"These results show that artificial intelligence could soon help to create patient-specific treatment plans for the most common chronic eye diseases," adds Raphael Sznitman, head of AIMI and director of the new Center for Artificial Intelligence in Medicine (CAIM).
Planning the ideal treatment frequency for each patient has several advantages. First, patients can be assured that their disease is being treated in the best possible way without having to undergo frequent visits and uncomfortable injections in the eye. Secondly, individualized planning can help clinics deal with constantly growing patient numbers and make the best possible use of specialized medical skills and infrastructures. Third, objective planning helps to avoid oversupply, improve cost efficiency and reduce overall expenditure.
According to the researchers, this study illustrates the proven collaboration between clinicians from Inselspital and data scientists from the ARTORG Center, who develop technology solutions suitable for everyday use directly from clinical needs. Another important building block for the clinical implementation of such a technology was the startup RetinAI. “We are very pleased that we can use the EU funding that we have received to develop patient-oriented solutions in ophthalmology. In doing so, we ensure that the technology can be transformed into products from which patients can really benefit and treatment can be improved on a large scale,” says Carlos Ciller, CEO of RetinAI.
