- Retinopathy of prematurity is an eye disease that affects preterm infants and can lead to vision loss and blindness if not detected and treated in the early stages of the disease.
- Regular screening of preterm infants can help prevent these adverse outcomes, but there is a shortage of pediatric ophthalmologists, especially in low- and middle-income countries.
- A recent study showed that an artificial intelligence (AI) model can analyze retinal images and accurately diagnose retinopathy of prematurity in preterm infants..
- The AI models used in this study did not require coding experience and could potentially be deployed in resource-constrained environments.
terrible
Screening programs can help prevent the progression of retinopathy of prematurity, but there is concern that there will be a shortage of pediatric ophthalmologists capable of performing these screenings, especially in resource-limited settings.
in the investigation,
a
Researchers said the AI model could diagnose severe retinopathy of prematurity using images acquired with devices other than the one used to develop the model, albeit with reduced accuracy.
Although further validation is needed, the researchers said their results show that a code-free AI model may be able to accurately diagnose retinopathy of prematurity in resource-limited settings.
“In sub-Saharan Africa, 30% of newborns have some degree of retinopathy of prematurity, and treatments are now readily available,” said study author Konstantinos Barakas, PhD. “It can lead to blindness if not detected and treated early.” Associate Professor, University College London. “This is often due to a shortage of ophthalmologists, but given that it is detectable and treatable, no child should be blinded by retinopathy of prematurity.”
“As the disease becomes more common, many communities lack adequately trained ophthalmologists to screen all at-risk children,” said Dr. Balascus. medical news today. “Our technology to automate the diagnosis of retinopathy of prematurity will improve access to care in underserved communities and prevent blindness for thousands of newborns worldwide. I hope.”
Retinopathy of prematurity is an eye disease that affects the retina, which forms the lining of the eye and is responsible for converting light into nerve impulses.
Retinopathy of prematurity is commonly seen in infants born before 31 weeks of gestation or weighing less than 3 pounds.
This eye disease is caused by abnormal growth of blood vessels in the retina. In mild retinopathy of prematurity, retinal vascular changes resolve spontaneously. In contrast, abnormal growth of blood vessels in severe retinopathy of prematurity can cause retinal detachment and lead to blindness.
Severe retinopathy of prematurity is characterized by structural changes with enlargement and tortuosity of the retinal vessels and is termed plus disease. The presence of plus disease is considered a marker of retinopathy that requires treatment.
Current guidelines recommend regular screening by a pediatric ophthalmologist for preterm and low birth weight infants. Advances in technology and increased screening have greatly improved premature infant survival, but the lack of adequate numbers of pediatric ophthalmologists is an obstacle to the sustainability of this effort.
The shortage of pediatric ophthalmologists is even more acute in low- and middle-income countries. Over the past decade, artificial intelligence applications have shown promise in solving this problem, but there are several obstacles to using this innovative approach for screening.
Ophthalmologists use images of the retina to visualize blood vessels and diagnose plus disease.Over the past decade, artificial intelligence applications have
Specifically, these applications are based on deep learning, a type of artificial intelligence that simulates the learning processes that occur in the brain. Deep learning models are trained using image datasets that have been annotated or labeled by medical professionals before being deployed in disease diagnosis. For retinopathy of prematurity, this includes using images previously identified by an ophthalmologist as with healthy or diseased.
However, there are several obstacles to the direct introduction of these models to the diagnosis of plus disease in clinical settings, especially in low- and middle-income countries. For example, most of these deep learning models are optimized using data from North America and Asia.
These data are expected to underestimate ethnic groups and people from lower socioeconomic backgrounds. It has been suggested that these models may not be generalizable because the development of retinopathy of prematurity is influenced by ethnicity.
Additionally, the research group trained most of these AI models for positive disease detection using data acquired with a specific imaging device called Retcam. Imaging devices such as Retcam tend to be expensive, and other devices are commonly used in low- and middle-income countries.
However, the accuracy of these models has not yet been evaluated on datasets acquired using other imaging devices. AI algorithms often show reduced accuracy when deployed to analyze image data acquired using a different device than the one used for model development, and external datasets are used prior to actual deployment. emphasizes the need to validate these models.
Deployment of these AI models is also limited by the need for expensive computer hardware and the expertise of data scientists. These resources may be inaccessible even to individual clinicians and research groups, especially in low- and middle-income countries.
These obstacles associated with customized deep learning models can be circumvented with code-free deep learning applications that require no coding expertise and have easy-to-use interfaces. Additionally, code-free deep learning programs are often cloud-based, so they don’t require expensive hardware. These code-free deep learning platforms require annotated datasets but can be used by clinicians without coding experience.
In this study, researchers compared the performance of experienced clinicians and bespoke code-free deep learning models in diagnosing diseases, based on the analysis of image data acquired using Retcam from different countries. compared.
In addition, they tested the ability of these models developed using Retcam to accurately identify plus disease using images acquired with different devices.
Researchers first developed a bespoke code-free deep learning model using Retcam images acquired from ethnically and socioeconomically diverse newborns in UK hospitals. Specifically, a bespoke code-free deep learning model was first trained on a subset of images from these neonates and then evaluated for accuracy on the remaining images of this dataset.
A custom code-free deep learning model showed similar accuracy to senior ophthalmologists in detecting infants with and without plus disease or with pre-plus disease. Preplus disease refers to vascular abnormalities similar to those seen in Plath disease, but not severe enough to be diagnosed as Plath disease. Detection of Preplus disease can help initiate early treatment of retinopathy of prematurity.
The two models also showed similar high diagnostic accuracy when analyzing Retcam image datasets from the United States and two low- and middle-income countries (Brazil and Egypt). However, code-free deep learning models were shown to be less accurate than bespoke models in detecting cases of pre-plus disease.
The researchers also evaluated the model’s performance using another Egyptian dataset acquired using a different imaging device called 3nethra. Both models showed lower diagnostic accuracy than training or validation datasets during analysis of this dataset obtained using 3nethra.
These results raise the potential of code-free deep learning models in diagnosing plus disease in low- and middle-income countries, where pediatric ophthalmologist shortages and limited resources may hinder regular screening of preterm infants. I am emphasizing it.
“This is an enlightening study that demonstrates the potential for very useful applications of artificial intelligence. We have shown that it performs as well as senior ophthalmologists in the field,” said Deepak Butt, Ph.D., MPH, Director of Mount Sinai Hart in New York.
“Machine learning and AI have moved from science fiction to potential in the clinical setting,” Batt said. medical news today. “This study is a good example of that. We need more studies like this in diverse populations.”
