A team of researchers from Harvard and Tufts Medical Center recently found artificial intelligence (AI) could be helpful in documenting and quantifying neuropathic corneal nerve pain. These findings are being presented this week at ARVO.
The investigators reported that AI was both quick and accurate in analyzing in vivo confocal microscopy images of microneuromas in the sub-basal nerve plexus.
Since neuropathic corneal pain is an under-diagnosed ocular surface disease with a lack of clinical signs to help explain a patient’s symptoms, researchers sought to prove that a form of AI known as deep neural networks—commonly used for pattern recognition in large image databases—could be used in the analysis of corneal sub-basal nerve alterations. Investigators also hoped to identify microneuromas associated with neuropathic corneal pain.
The study included 15 patients with dry eye disease and neuropathic corneal pain, and a total of 12,212 images. Two ophthalmologists identified and confirmed the presence or absence of microneuromas. Then, a deep neural network with more than 24 million parameters was used to analyze the images for microneuromas, finding them in 407 images.
Researchers concluded the deep neural network was highly accurate in identifying microneuromas associated with neuropathic corneal pain, opening the door for the possible standardization of in vivo confocal microscopy image analysis, the researchers said in the presentation abstract.
"This exciting application of AI using a neural network to detect neuropathic corneal pain by analyzing IVCM images shows promise," says Joseph Shovlin, OD, of Scranton, PA. “By looking at corneal sub-basal nerve alterations, clinicians may someday have an additional tool to accurately and rapidly detect the microneuromas associated with neuropathic pain.”
|Koseoglu N, Beam A, Hamrah P, et al. The utilization of artificial intelligence for corneal nerve analyses of in vivo confocal microscopy images for the diagnosis of neuropathic corneal pain. ARVO 2018. Abstract 3440.