AI Brings Vision to Sight Impaired

Image credit: Leben Care

Researchers are tapping artificial intelligence to improve eyesight in people with impaired vision. New technologies can quickly and easily identify retinal conditions in diabetic patients, allowing healthcare workers to begin treatment sooner. AI-based technologies are also being used to augment eyesight in people who have partial vision loss.

On the diagnostic side, Singapore-based Leben Care has developed Netra.AI, a retina risk assessment software-as-a-service platform. Using AI and machine learning, the software is used as a screening tool to determine the health of a retina in diabetic patients. The cloud-based AI solution is powered by the Intel® Xeon® Scalable processor platform and uses the Intel® Deep Learning Boost (Intel® DL Boost) and Intel® Advanced Vector Extension 512 acceleration for fast analysis.

Internal images of the patient’s eyes are uploaded anonymously to the Netra.AI platform through a web portal or API. The software uses a four-step deep convolutional neural network (DCNN) to check the images for indications of diabetic retinopathy, glaucoma, macular degeneration, and other retinal anomalies. Netra.AI enables healthcare workers to bring swift treatment to patients to prevent further damage or vision loss. Reports are generated within two minutes, and Leben Care claims the system is as accurate or better than human diagnosis.

Leben Care is working with Sankara Eye Foundation to deploy the technology in India, where an estimated 77 million people have diabetes and trained retinal specialists are lacking. As of March 2021, more than 3,000 patients in India have been screened using Netra.AI, and nearly 750 were flagged as at-risk. Because diabetic retinopathy can cause blindness if not treated, healthcare workers can use Netra.AI to identify patients and start treatment early.

In addition to the Netra screening component, Netra Track maintains and archives the eye images, so medical staff can monitor degradation overtime. The Netra API allows solutions providers and OEMs to integrate the Netra.AI technology with existing healthcare software systems and devices.

Image of eye retina and it's red veins

Image credit: Leben Care

An Eye on the Future

Beyond diagnosing sight impairments, AI is being used to enhance sight for people who have profound or partial vision loss. Foregoing the premise of restoring “natural” vision, which is proving to be too difficult, researchers at UC Santa Barbara are using AI to augment impaired vision in a manner that is useful and practical, providing cues to the visually impaired.

This AI-based bionic eye operates similarly to a cochlear implant. A retinal prosthesis is surgically implanted and sends electrical stimulation to the existing retinal cells. Rather than try to recreate an entire image, researchers are using computer vision algorithms, deep learning, and image processing to simplify an image or scene.

For example, a sighted person might see a red car driving down a street with a pedestrian running on the sidewalk next to a graffiti-covered wall. The technology would isolate the shapes of the person and the car in white, allowing a vision impaired person to “see” those objects without concern for the other images, which could fade to black. The goal is to use the high contrast to provide basic orientation and enable mobility tasks.

In the future, the UCSB researchers claim the prosthetic could be integrated with other technologies, such as GPS to get directions or identify obstacles along the path. Night vision capabilities could be possible through the addition of an infrared sensor that extends the range of visible light. The UCSB research team is testing the bionic eye technology on sighted patients using virtual reality headsets.