We often say imaging devices such as SD-OCT help doctors identify disease. But what if those gadgets actually did the identifying themselves? Thanks to deep learning technology, they’re starting do just that. Take, for instance, a study published in Ophthalmology.
Researchers fed SD-OCTs equipped with deep learning capabilities thousands of images of eyes with and without glaucomatous visual field damage. After being “trained” to identify disease using a combination of datapoints, including retinal nerve fiber layer (RNFL) thickness maps, RNFL enface images and confocal scanning laser ophthalmoscopy (CSLO) images, the deep learning models outperformed standard RNFL thickness measurements in predicting all quantitative visual field metrics. It also showed high accuracy in measuring visual field sectoral pattern deviation in the inferior nasal and superior nasal sectors, as well as moderate accuracy in inferior and superior sectors, but lower accuracy in central and temporal sectors.
Accurately predicting the severity of glaucomatous visual field damage from SD-OCT imaging can help clinicians better tailor visual field testing frequency, the investigators concluded.
|Christopher M, Bowd C, Belghith A, et al. Deep learning approaches predict glaucomatous visual field damage from optical coherence tomography optic nerve head enface images and retinal nerve fiber layer thickness maps. Ophthalmology. September 23, 2019. [Epub ahead of print].|