Artificial Intelligence Could Stop You From Going Blind?

To give some background information, Diabetic retinopathy, a somewhat common complication of diabetes with varying severity, is the result of damage caused to the retinal blood vessels at the back of eye due to high blood sugar content. Stemming from both type 1 and type 2 diabetes, the disease progresses from nil symptoms to mild vision defects and eventually leads to serious defects (such as blindness) if actions are not taken. Over time, the sugar content of your blood could lead to blockages in miniscule retinal blood vessels, and prevents the blood supplication, and as a result, new blood vessels are formed to overcome this defect, but these are very prone to being impaired and leaking rather severely. The retina itself, is a layer of light-sensitive cells, at the back of the eye that detects light and expresses it via electrical signals, which are sent off the brain and formed as an image. Since this is effectively going on all the time, the retina requires a constant blood supply, which is the issue caused by diabetic retinopathy as it can cut off this supply as new blood vessels leak and old blood vessels are blocked. There are three main stages to the development of this condition:

  • background retinopathy (or mild non-proliferative retinopathy)– micro-bulges (microaneurysms) form on your blood vessels, which rarely bleed and rarely cause vision defects.

  • pre-proliferative retinopathy (or moderate non-proliferative retinopathy which stems into severe non-proliferative retinopathy)– more significant bleeding in your eye occurs, due to a greater severity of changes to the blood vessels. Larger retinal vessels may begin to dilate, and become inconsistent in diameter, as well as potential swelling of the nerve fibers or macula (central region of the retina) can occur in the background and pre-proliferative phase, but with a greater extent in the latter phase.

  • proliferative retinopathy – loss of vision could occur, as scar tissue and new blood vessels form, the latter which is prone to bleed rather easily. The damaged blood vessels close off, hence new abnormal blood vessels form but leak into the vitreous (jelly-like substance that films the centre of your eye). The scar tissues can cause the retina to detach from the back of the eye. In addition to this, the optic nerve (carries images to your brain from the eye) can be damaged, which is called glaucoma, this will occur if the new blood vessels affect the normal flow of fluid out of the eye, leading to pressure build up.

 
A retinal image of diabetic retinopathy, you can see soft and hard exudates on the right hand side of the image (pale scatters), with various haemorrhages and microaneurysms Credit: Medical News Today  (annotated example  here  via mdpi.com)

A retinal image of diabetic retinopathy, you can see soft and hard exudates on the right hand side of the image (pale scatters), with various haemorrhages and microaneurysms Credit: Medical News Today

(annotated example here via mdpi.com)

 

For those who are interested in other features of the condition, such as risk factors and complications (e.g. Vitreous haemorrhage and Retinal detachment as we mentioned in the former section), I highly recommend the NHS website and Mayo Clinic;

Let’s get into the news!

According to findings published in Diabetes Technology & Therapeutics, an AI screening device matched rival human screening fellows in the identification of potential cases of diabetic retinopathy. Dr Malavika Bhaskaranand, director of product development at Eyenuk Inc, alongside colleagues, used the EyeArt™ technology produced by Eyenuk Inc to analyse 850,908 fundus retinal images collected in a 21 month period commencing from January 2014, via 404 primary care clinics which were obtained through 107,000 unique enquiries from patients with diabetic complications. The results of this test, i.e. the accuracy of the system, was directly compared with optometrists and ophthalmologists to assess the usefulness of such a system in a modern clinical environment. Dr Malavika Bhaskaranand commented “The currently used manual screening setups cannot scale up to effectively triage the ever-increasing population of people with diabetes at risk for vision loss, and the limited number of ophthalmologists, creating a large unmet need for screening,”. As for the results, the system accurately diagnosed 15,177 cases of pre-proliferative diabetic retinopathy, 2,819 cases of proliferative diabetic retinopathy and another 2,625 cases of severe diabetic retinopathy, which amounts to a grand total of 20,261 referable cases from the data set. The reseaerchers said that 98.5% of the potentially treatable cases (which were 5,446 of these referrals) were correctly referred by the system. The system also identified 8,816 cases of background diabetic retinopathy and 72,189 cases of no diabetic retinopathy at all, with only 5,373 cases that were unable to be deciphered by the system. With this, the researchers claimed that the screening system showed a sensitivity of 91.3% whereas clinical screenings from medical providers only revealed a sensitivity of 91.1%, showing a marginal difference. These are promising results, and could very well be the foundation of revolution in ophthalmology and diabetics, with AI furthering its input into clinical healthcare, very promising indeed!

 
Stock image showing an optometrist in diagnostic action, a process that could very well be rapidly sped up with the advent of AI. Credit: Complete Eyecare of Medina, Plymouth Eye Doctor

Stock image showing an optometrist in diagnostic action, a process that could very well be rapidly sped up with the advent of AI. Credit: Complete Eyecare of Medina, Plymouth Eye Doctor

 

Creators of the EyeArt™ system, Eyenuk Inc, based in Los Angeles (California), is a diagnostic company that heeds machine learning tools to aid quick and accurate identification of patients who are prone, or currently suffering, from various eye diseases and chronicity at the point of care, the conditions themselves may be seriously impairing or even progressing towards blindness. By applying machine learning know-how and computer vision (how computers can acquire data from digital imaging), the company is quickly developing a product range that is developed from their proprietary retinal imaging and analysis software in tandem with deep learning techniques (that have been trained via retinal imaging databases and nurtured to find trends that may be indicative of eye disease) which can be used to track the progression of various conditions, such as diabetic retinopathy, macular degeneration, glaucoma, Alzheimer’s disease and cardiovascular risk. Their fore front product, EyeArt™, is commercially distributed in European and Canadian markets to date, and has found support from the National Eye Institute across the pond!

Promotional imaging on the company’s website, revealing EyeArt™’s steps in fully automated diabetic retinopathy analysis Credit: Eyenuk, Inc

Promotional imaging on the company’s website, revealing EyeArt™’s steps in fully automated diabetic retinopathy analysis Credit: Eyenuk, Inc

As for former successes, a publication last year (March 2018) in the ‘Eye’ journal (Springer Nature) showed a very high efficacy in detecting diabetic retinopathy or DR (95.8%) using using EyeArt™ automated screening software with tandem with the Remidio Fundus on Phone system. The effectiveness increased to 99.1% where the AI was looking out for progressions that were sight compromising, known as STDR or sight-threatening diabetic retinopathy. The specificity of these results sat at 80.2% for DR and 80.4% for STDR, and that false positives were recorded due to the presence of unrelated lesions such as drusen, retinal vein occlusions and tessellated fundus. Kaushal Solanki, Founder & CEO of Eyenuk, Inc said at the time, “We are very encouraged by this study demonstrating impressive sensitivity and specificity using EyeArt™ with images from this novel smartphone-based fundus imaging system. This is especially notable given that our AI system was never trained on images captured by the Remidio FOP system, suggesting the robustness and broad applicability of our software algorithms. We look forward to converting these impressive results into real-world clinical practice by eventually making EyeArt™ available for mass diabetic retinopathy screening in conjunction with a lower-cost portable high-quality fundus imaging device like the Remidio system.”. In addition to this, Dr. Anand Sivaraman, CEO of Remidio Innovative Solutions, who were the creators of the Remidio system used in tandem with EyeArt™, said, “The simplicity achieved by combining smartphones with our patented optics has resulted in Remidio Fundus On Phone to be priced at 1/5th to 1/10th of any traditional desktop retinal imaging system, delivering high quality relevant images repeatedly with high sensitivity and specificity exceeding 92% and 98% when clinically validated. The use of artificial intelligence like EyeArt™ in screening the images from the Remidio Fundus on Phone device, now enables large scale public health screening for eye health among the 400 million affected by Diabetes, globally.” It is great to see that the company has continued to progress with their software, and I wish both these companies the very best of luck!

The following is a summary of the EyeArt™ system.

The following is a summary of the EyeArt™ system made by Eyenuk themselves, with commentary from Dr Srinivas R. Sadda

A very big thanks to Eyenuk Inc and Remidio Innovative Solutions, Kaushal Solanki and Dr. Anand Sivaraman (the respective CEOs) for their innovative efforts, and addressing the issues of under-staffing and misdiagnosis in clinical environments. The works of these companies and gentlemen could very well be the new golden standard of healthcare in the coming future, and give us all, doctors and patients, an easier time! Thank you for contributing to my article, and to my readers, thank you for reading this article!

Article thumbnail credit: Alphabet’s DeepMind uses A.I. to detect signs of eye disease, CNBC [click here for page]. Note that this article has not contributed to any of the written content of my article, and that you click the above link at your own discretion as the page has not been checked by our team. Thank you to our readers for your continued support, and to anyone who contributes to our fantastic team to keep our services in orderly fashion.