AI helps identify rare hereditary retinal condition – 06/18/2023 – Health

An AI (artificial intelligence) tool was able to detect more than 75% of hereditary retinal dystrophy (IRD) cases by scanning the eye.

Hereditary retinal dystrophies are a set of rare diseases, inherited from parents to children, with different patterns of genetic inheritance, and can cause degeneration of cones and rods, light receptor cells present in the organ, causing progressive visual loss.

There are several manifestations of IRDs, but the detection of the different types (phenotypes) depends on the evaluation of an expert ophthalmologist, often difficult to find in most medical services.

With eye examination, the technique can automatically detect the presence of types of hereditary dystrophies.

The tool, called Eye2Gene, was developed by researchers from the Institute of Ophthalmology at University College London (UCL), the Moorfields Eye Hospital Research Fund and other institutions in the United Kingdom, Germany and Belgium, and presented on the last day 10 at the European Congress of Human Genetics in Glasgow (Scotland).

To create the AI ​​model, the scientists used data from 130 patients with hereditary retinal dystrophy for whom there was a known genetic diagnosis. As the physical manifestations of IRDs are variable, it was also necessary that each of the patients had at least three scores (criteria used by specialists) specific to the disease.

Scores for inherited retinal diseases are defined by an international protocol called HPO (Human Ontological Phenotype). The patterns (or phenotypes) found are described from a series of characteristics that allow the identification of the different associated genes.

The problem is that the criteria that define the different eye diseases are difficult to determine and time consuming. As in some cases the manifestation patterns are similar, the determination of which gene causes each type of condition is crucial for the correct diagnosis.

In the study, the Eye2Gene tool hit rate was more than 75%, equal to or greater than the correct diagnosis when evaluated only by an expert.

For Nikolas Pontikos, principal investigator at the UCL Institute of Ophthalmology and director of data at the AI ​​laboratory at Moorfields Hospital and first author of the study, the tool may be more appropriate in some cases for detecting IRDs.

“Especially in cases where the defining criteria used to describe the condition is just a phenotype [manifestação clínica]the tool manages, through direct retinal scans, to process the image using AI in order to identify characteristic phenotypes of certain retinal diseases”, he said in an interview with Sheet.

According to him, the tool can be very useful for interpreting variants and incorporating detection criteria in an automated and objective way.

“Eye2Gene has data so far for the most common IRDs, which include the genes responsible for Stargardt disease [impossibilita a distinção de cores e de pequenos detalhes]retinitis pigmentosa [degeneração progressiva da retina] and achromatopsia [caracterizada pela ausência de visão de cores], among others. We look forward to being able to conduct larger clinical studies, including a larger number of subjects, in order to evaluate the safety and efficacy of Eye2Gene before it can be fully incorporated into clinical practice,” he said.

In Brazil, the estimated incidence of rare diseases ranges from 1 to 2,000 individuals. Some of the diseases are more common than others. For most of them, genetic testing is the only one capable of diagnosing the condition with certainty.

There is no cure for hereditary retinal diseases, but some treatments are in the research phase, since the main effects can be progressive visual loss or, in some cases, combine with damage to other organs, as in Usher syndrome, also causing hearing loss.

According to Retina Brasil, a non-profit association created to inform and help families and patients with retinal diseases, the only treatment available is Luxturna gene therapy, indicated for a type of disease caused by the RPE-65 gene, and not yet available. in Brazil.