Researchers at the National Institutes of Health used artificial intelligence (AI) to enhance a technique for creating high-resolution images of eye cells. They found that AI made the imaging process 100 times faster and increased the image contrast by 3.5 times. This advancement will give researchers a more effective tool for studying age-related macular degeneration (AMD) and other retinal diseases.The experts claim that advancements in artificial intelligence will offer researchers a better way to assess age-related macular degeneration (AMD) and other retinal diseases.
“Artificial intelligence is helping to overcome a major limitation in imaging cells in the retina, which is time,” stated Johnny Tam, Ph.D., who leads the Clinical and Translational Imaging Section at NIH’s National Eye Institute.
Tam is working on a technology called adaptive optics (AO) to enhance imaging devices based on optical coherence tomography (OCT). Similar to ultrasound, OCT is noninvasive, quick, painless, and standard equipment in most eye clinics.
Imaging RPE celThe use of AO-OCT presents new challenges, such as dealing with a phenomenon known as speckle. Speckle can disrupt the clarity of AO-OCT images, similar to how clouds can hinder aerial photography. To address this issue, researchers continuously capture images of cells over an extended period, allowing the speckle to change and reveal different parts of the cells over time. This process requires the scientists to meticulously combine numerous images to produce a clear, speckle-free image of the RPE cells. Additionally, Tam and his team have created an innovative AI-based method to assist in this process.A deep learning algorithm known as parallel discriminator generative adverbial network (P-GAN) was developed by the researchers. They used this method to train the network by inputting nearly 6,000 manually analyzed AO-OCT-acquired images of human RPE, along with their corresponding speckled originals. The goal was to teach the network to recognize and restore speckle-obscured cellular features.
In tests using new images, P-GAN effectively removed the speckles from the RPE images, revealing cellular details. It was able to produce results comparable to the manual method, which involved capturing and averaging 120 images. The team used various performance metrics to evaluate the algorithm’s effectiveness.
In terms of cell shape and structure, P-GAN performed better than other AI methods. Vineeta Das, Ph.D., a postdoctoral fellow at the NEI’s Clinical and Translational Imaging Section, believes that P-GAN reduced imaging acquisition and processing time by approximately 100 times. P-GAN also provided a higher contrast, about 3.5 times greater than before.
“Adaptive optics takes OCT-based imaging to the next level,” stated Tam. “It’s akin to transitioning from a balcony seat to a front row seat for imaging the retina. With AO, we can uncover 3D retinal structures at cellular-scale resolution, allowing us to focus on very early signs of disease.”
While adding AO to OCT offers a better view of cells compared to OCT, but processing AO-OCT images after they have been taken is more time-consuming. Tam’s latest research focuses on the retinal pigment epithelium (RPE), which is a layer of tissue located behind the light-sensing retina. The RPE supports the metabolically active retinal neurons, including the photoreceptors. The retina is responsible for capturing, processing, and converting light into signals that are transmitted to the brain through the optic nerve. Scientists are particularly interested in studying the RPE because many retinal diseases occur in this area.
As the RPE deteriorates, Tam suggests that combining AI with AO-OCT could help overcome a significant barrier to regular clinical imaging with AO-OCT, especially for diseases affecting the RPE, which has historically been challenging to image.
“Our findings indicate that AI has the potential to revolutionize the way images are obtained,” Tam stated. “Our P-GAN artificial intelligence has the potential to make AO imaging more accessible for routine clinical use and for research into the structure, function, and pathophysiology of sight-threatening retinal diseases. Considering AI as an integral part of the imaging system, rather than just a tool, will be key.”The application of AI-assisted adaptive optics optical coherence tomography, which reveals living human retinal cells obscured by speckle, is a significant shift for the AI field.