Glasgow, UK: Inherited retinal diseases (IRDs), single-gene disorders that affect the retina, are rare and very difficult to diagnose because they involve alterations in one of many candidate genes. Outside of specialty centers, few specialists have sufficient knowledge of these diseases, making it difficult for patients to get proper tests and diagnoses. But now, British and German researchers have developed a system that could use artificial intelligence (AI) to offer a wider range of tests and improve efficiency.
Dr. Nicholas Pontikos, group leader at the UCL Institute of Eye Medicine and Moorfields Eye Hospital, London, UK, announced today (Saturday, June 10) at the European Society of Human Genetics annual conference that his team could develop an AI system I plan to talk about Eye2Gene, which is. A study to identify the genetic cause of IRD from retinal scans. “Identifying the causative gene from retinal scans is considered very difficult even for experts. But AI can achieve this with greater accuracy than most human experts.” says Dr Ponticos.
Researchers had access to a vast database of information on IRDs at Moorefields Hospital, covering over 30 years of research. At Moorfields, he has more than 4,000 patients undergoing genetic diagnosis and detailed retinal imaging, making it the largest single center patient dataset with both retinal and genetic data.
Identification of genes involved in retinal diseases is often performed using patient phenotypes defined using the Human Phenotype Ontology (HPO). HPO uses standardized medical terminology and structured descriptions of patient phenotypes (personal observable characteristics resulting from gene expression) to enable scientists and physicians to communicate more effectively. It is included. “However, HPO terminology is often an incomplete description of retinal image phenotypes, and the promise of Eye2Gene is that operating directly from retinal images can provide a much richer source of information than HPO terminology alone.” says Dr. Pontikos.
The research team benchmarked Eye2Gene on 130 IRD cases for which whole exome/genome, retinal scans, and detailed HPO descriptions were available and had a known genetic diagnosis, and compared their HPO gene scores with Eye2Gene gene scores. compared. They found that Eye2Gene provided a rank of the correct gene in more than 70% of his cases with his HPO-only score or better.
In the future, Eye2Gene could easily be incorporated into standard retinal examinations, initially as a specialist hospital assistant to obtain a second opinion, and ultimately as a “general It will be available as an “expert”. “Ideally, the Eye2Gene software would be embedded in a retinal imaging device,” says Dr. Pontikos.
The system must receive regulatory approval to demonstrate safety and efficacy before its use becomes more widespread. This future use of AI has the potential to be a more effective, less invasive, and more widely accessible approach to diagnose patients and improve management and treatment. “Further evaluation of Eye2Gene is required to assess its performance on different types of IRD patients in different types of settings, including different ethnicities, different types of imaging devices, e.g. first-line and second-line therapy. A clinical trial is needed to bring our system into the clinic as medical device software,” says Dr. Pontikos.
“We all know that seeing is believing, so we had some hope that AI-interpreted retinal scans could outperform HPO conditions alone. We were pleasantly surprised to find that Eye2Gene still performed as well or better than the HPO-only approach, even when very specific HPO terminology was used. “We hope that AI will help patients and their families by making professional care more efficient, accessible and equitable,” he concluded.
The chair of the conference, Professor Alexandre Raymond, said: “Real experts are essential, but the use of AI will help reduce stigma and, in the future, make diagnoses for everyone.”
