Science & Enterprise subscription

Follow us on Twitter

  • This is nuts. Manufacturing, distribution, and public health infrastructure here are more important than who made t… https://t.co/AOHVdFBlrX
    about 12 hours ago
  • Don't let the perfect be the enemy of the good ... https://t.co/nHtOdudliO
    about 1 day ago
  • Among the tech giants, Apple is the most aggressive in acquiring start-ups developing A.I. applications over the pa… https://t.co/RjpknHBKAR
    about 2 days ago
  • New post on Science and Enterprise: Infographic – Apple Leads in A.I. Acquistions https://t.co/WTof485yka #Science #Business
    about 2 days ago
  • HBCUs still hand out more bachelor's degrees in STEM fields to Blacks than traditionally white institutions, but no… https://t.co/xW2ICnR09b
    about 3 days ago

Please share Science & Enterprise

A.I.-Aided Method Predicts Glaucoma Progression

DARC retinal scan

Retinal scan image made with DARC showing degenerating nerve cells as white spots (University College London, Western Eye Hospital)

4 May 2020. Clinical trial results show a technique visualizing cells in the retina and analyzed with an algorithm accurately predicts the progression of glaucoma. Findings from the trial conducted in the U.K. are reported in yesterday’s issue of the journal Expert Review of Molecular Diagnostics (paid subscription required).

Glaucoma is the name given to a collection of eye conditions resulting in damage to the optic nerve that in advanced stages can lead to vision loss. In most cases of glaucoma, abnormally high intraocular pressure in the eye results in the optic nerve damage The Glaucoma Research Foundation cites reports from World Health Organization showing glaucoma is the second leading cause of blindness in the world, affecting some 60 million people.

Researchers from University College London, led by ophthalmology professor Francesca Cordeiro, developed a technique to detect vision loss from degeneration of the retinal ganglion cell, the nerve cell in the retina that transfers visual information from the eye to the brain. The technique, called detection of apoptosing retinal cells, or DARC, finds apoptosis, or programmed cell death in retinal ganglion cells, an early sign of glaucoma. Most of today’s diagnostics for glaucoma detect the disorder only well after damage to the nerve occurs.

With the DARC technique, patients are given an intravenous injection of a fluorescent dye called annexin 5, which travels to the eye. There, the dye attaches to cells in the retina, where cells illuminate when undergoing apoptosis. In an earlier clinical trial, healthy volunteers showed no serious adverse effects from the dye now code-named ANX776. Retinal scan images can then show the illuminating cells enabling physicians to make their diagnosis.

A continuing problem, say the authors, is disagreements among specialists interpreting the retinal scan images. To provide more speed and certainty to DARC, the University College team developed an algorithm trained by retinal image scans to interpret DARC images and provide more guidance to ophthalmologists. The algorithm uses a convolutional neural network that combines image analysis and machine learning to dissect an image by layers for understanding features in the image. Different aspects of each layer discovered and analyzed by the algorithm are translated into data that the algorithm then uses to train its understanding of the problem being solved, in this case interpreting retinal image scans.

A team from University College London, Imperial College London, and Western Eye Hospital in London conducted the mid-stage clinical trial enrolling 40 healthy adults and 20 individuals with glaucoma, matched in age to the healthy participants. All participants received a single injection of ANX776, with laser retinal scans taken four hours later.

The computer images were then analyzed with the DARC interpretation algorithm. The study team followed up with glaucoma participants 18 months later to detect progression of the disease, using optical coherence tomography, a non-invasive imaging technique to measure deterioration of retinal ganglion cells.

The findings show the algorithm returns accuracy, sensitivity (true positive), and specificity (true negative) results of 91 to 97 percent detecting glaucoma in individuals with the disease, compared to healthy participants. After 18 months, the algorithm accurately predicts progression in participants’ glaucoma with 86 percent sensitivity and 92 percent specificity.

“These results are very promising as they show DARC could be used as a biomarker when combined with the AI-aided algorithm,” says Cordeiro in a University College London statement released through EurekAlert. “What is really exciting, and actually unusual when looking at biological markers, is that there was a clear DARC count threshold above which all glaucoma eyes went on to progress.”

University College London holds a patent on the DARC algorithm, with Cordeiro and co-author John Maddison listed as inventors. In February 2020, Cordeiro co-founded the company Novai Ltd. in London that licenses the technology for commercialization.

More from Science & Enterprise:

*     *     *

Please share Science & Enterprise ...

Comments are closed.