Research led by the National Eye Institute in Bethesda shows that incorporating artificial intelligence (AI) into retinal imaging significantly speeds up the imaging process, as well as improving image contrast.
The development of adaptive optics optical coherence tomography (AO-OCT) has allowed more detailed and 3D images of the retina to be taken and used to help diagnosis of a range of different eye diseases.
However, the images taken using AO-OCT have a lot of “noise” that makes it difficult to visualize retinal cells from a single scan as random sections of the image can be obscured. This means as many as 120 images normally need to be taken so an average can be estimated in order to improve image contrast.
“This high contrast, complex intensity distribution of speckle noise can mask cells and limit the visibility of cellular structures. In particular, the retinal pigment epithelial cells, which are essential for maintaining visual function have low intrinsic contrast compared to speckle noise and therefore are challenging to image directly,” write the researchers in the journal Communications Medicine.
“This lengthens the overall acquisition time, not only due to the additional volumes required, but also, because of the possibility for artifacts or registration errors across the sequentially acquired volumes due to constant involuntary eye movements that translate and distort the cellular visualization obtained from the microscopic imaging field of view.”
Johnny Tam, who leads the Clinical and Translational Imaging Section at the NIH’s National Eye Institute, and team used a parallel discriminator generative adversarial network (P-GAN), a type of AI, to reduce speckle on the images and improve contrast.
The addition of AI to the technology sped up the imaging time by 99-fold and also improved the contrast of the images 3.5-fold. One AI aided image was equivalent to around 120 taken using current methods.
“Artificial intelligence helps overcome a key limitation of imaging cells in the retina, which is time,” said Johnny Tam, who leads the Clinical and Translational Imaging Section at the NIH’s National Eye Institute, in a press statement.
“Our P-GAN artificial intelligence will make adaptive optics imaging more accessible for routine clinical applications and for studies aimed at understanding the structure, function, and pathophysiology of blinding retinal diseases. Thinking about AI as a part of the overall imaging system, as opposed to a tool that is only applied after images have been captured, is a paradigm shift for the field of AI.”
