Comprehensive Summary
Pathological myopia (PM) is a disorder of the eye that affects up to 3% of the general population. It is characterized by a degenerative disease slowly wearing down the nerve structures at the back of the eye, leading to long term impairments such as blindness. The disease can be treated, but diagnosing PM early is key in beneficial prognoses for patients. This article utilizes an artificial intelligence and deep learning system called ATN which stands for atrophy, traction, and neovascularization. It uses a broad and comprehensive institutional dataset of 2500+ photographs of the interior surface of the eye (the fundus) to identify markers of different stages of PM. This study outlines the strengths of machine learning in this setting, as well as identifying new tests and procedures to improve the accuracy of machine learning to provide solid results that can be used in a medical setting. Ultimately, the intelligence spans across all three of the ATN components while also tying in widely-used OCT scans to provide a comprehensive diagnosis of the eye combining the strengths of both modalities.
Outcomes and Implications
The incorporation of AI diagnostic imaging to streamline pathological myopia diagnosis is revolutionary in its promises for patient outcomes. While it does not directly provide a cure, many modern treatments for the disorder have its highest efficacies when implemented during the early stages of the disease. This technology has a lot of potential to serve as a mainstream diagnostic tool, combining two individually strong modalities, to provide accurate diagnoses especially in rural and more remote areas and areas with high prevalence of PM like East Asia. Furthermore, on the machine learning side, this model's ability to classify even minute optical markers and compile large data sets to be transferred to medical application can be very promising for the scalability of other deep learning models in healthcare. It sets a foundation for implementation and also regulation methods for artificial intelligence in medicine where the stakes are high and accurate readings are of utmost importance. And going forward, this technologically can be leveraged to reduce the diagnostic burden and circumvent human error not just in the field of ophthalmology, but in countless healthcare specialities.