Artificial Intelligence with Adaptive Optics Aids in the Detection and Treatment of Glaucoma, Eye Disease

Article: AI Spots Individual Neurons in the Eye Better than Human Experts
Source: Duke University, via ScienceDaily  and NEI
Published: May 4, 2021

Biomedical engineers recently made progress in ophthalmic imaging through combining deep-learning artificial intelligence with optical coherence tomography (OCT) and adaptive optics to enable better diagnosis and monitoring of neuron-affecting diseases such as glaucoma. Traditional OCT is able to scan the thickness of retinal cell layers, but cannot visualize individual retinal ganglion cells. The new technique improves upon imaging by tracking changes in the number and shape of the eye's retinal ganglion cells. The axons of retinal ganglion cells form the optic nerve that relays visual information to the brain; it is also these axons that are damaged in neurodegenerative diseases such as glaucoma. With greater sensitivity of detecting changes to individual neurons conferred by adaptive optics, AO-OCT imaging can detect disease at earlier stages and monitor disease progression more rapidly. However, the higher resolution also generates a large amount of data that causes an image analysis bottleneck, which the research team solved with the addition of deep-learning algorithms, dubbed WeakGCSeg. As one of the researchers states, “Our experimental results showed that WeakGCSeg is actually superior to human experts, and it’s superior to other state-of-the-art networks that can process volumetric biomedical images.” A potential application of this technology is for use in clinical trials. Because the technique can detect disease at earlier stages and at shorter time spans, it can more precisely and more quickly monitor differences in a progressive disease that otherwise would take the death of hundreds or thousands of cells and months or years to manifest. The research team plans to expand their technique to the visualization of other cell types, such as photoreceptors, and other neurodegenerative diseases.

My rating of this study: ⭐⭐⭐

Soltanian-Zadeh S, Kurokawa K, Liu Z, et al. "Weakly supervised individual ganglion cell segmentation from adaptive optics OCT images for glaucomatous damage assessment." Optica.  8(5):642-651. 2021. https://doi.org/10.1364/OPTICA.418274