Artificial intelligence is doing fascinating things for health care. At best, doctors can be hopeful that partnering with these robo-techs will help to provide the most comprehensive picture of patients’ health with minimal invasiveness. For instance, at a recent Vision Expo meeting in Las Vegas, manufacturers boasted that deep learning software applied to diagnostic imaging devices, such as OCT and fundus cameras, can figure out a patient’s sex based from the retina image alone—and researchers don’t even know why.1 Now, new research is showing that the same type of technology is able to diagnose meibomian gland dysfunction (MGD) using in vivo laser confocal microscopy images alone.2
“Our network combining deep learning and in vivo laser confocal microscopy learned to differentiate between images of healthy meibomian glands and images of obstructive MGD with a high level of accuracy that may allow for automatic obstructive MGD diagnoses in patients in the future,” the study says.2
Investigators relied on 137 images of obstructive MGD and 84 images of normal meibomian glands to “train” the deep learning software. Researchers noted a 94.2% sensitivity and an 82.1% specificity.2
The authors mentioned that they employed the Heidelberg Retina Tomograph II-Rostock Cornea Module (HRT II-RCM, Heidelberg) with a diode laser without everting the eyelid and obtained excellent results.2
Though real-world application of such technology is likely a ways away, the foundational work is now being done and increasingly proving its aptitude for myriad diagnostic responsibilities.
1. Poplin R, Varadarajan A, Blumer K, et al. Prediction of cardiovascular risk factors from retinal fundus photographs via deep learning. Nature Biomedical Engineering. February 19, 2018.
2. Maruoka S, Tabuchi H, Nagasato D, et al. Deep neural network-based method for detecting obstructive meibomian gland dysfunction with in vivo laser confocal microscopy. Cornea. February 7, 2020. [Epub ahead of print].