You may have seen or perused about a paper published before the end of last year in the Diary of the American Therapeutic Association, describing the development and approval of a profound learning calculation for recognition of diabetic retinopathy (DR) from fundus photographs.1 In the study described in that paper, a Google artificial intelligence (AI) calculation was ready to translate and grade fundus photographs with various stages of DR at any rate as precisely as an accomplice of ophthalmologists, based on machine learning technology.

The calculation’s success at diagnosing referable (direct or above) DR was contrasted and the larger part decisions of no less than seven board-ensured ophthalmologists evaluating more than 11,000 shading fundus photos. In two picture sets, the calculation accomplished sensitivity of 97.5% and 96.1% and specificity of 93.4% and 93.9%. Assuming a 8% pervasiveness of referable DR, these results yield a negative prescient estimation of 99.6% to 99.8%.

We were not authors on the Diary of the American Medicinal Association paper, however as physician consultants who worked with the Google team making this technology, we need to address some of the regular questions about this subject we have gotten notification from our eye care colleagues.


The Google profound learning calculation used in this study of mechanized DR analysis was a progressed artificial neural network loosely modeled after the human mind. Artificial neural networks are figuring system composed of numerous simple, exceptionally interconnected processors. The numerous nodes or processors inside the system each make simple calculations that are weighted and included to produce the last yield.

READ THIS TOO:   Indepth View On How Machine Learning Is Used In Optimizing Robot Learning And Artificial Intelligence

For this study, the Google system was at first trained by using around 120,000 shading fundus photos, each named with a diagnosis by ophthalmologists. In the training phase, the system made a diagnostic “guess” on each picture. It at that point contrasted its answer with the ophthalmologists’ named answer and adjusted the weights of every hub, learning how to figure with the lowest possible diagnostic mistake. It did this over and over, hundreds of thousands of times. The system can’t simply retain the diagnosis for each picture, yet rather is compelled to learn wide rules that will probably sum up to future, unseen images. At the point when this calculation was approved in the study, 11,000 at no other time seen images were shown to the calculation, and the results from the network’s analysis were contrasted and those from load up confirmed ophthalmologists.

In what manner WILL THE TECHNOLOGY Influence PATIENTS?

In underserved populations with poor access to social insurance, there are extraordinary potential benefits to the use of machine-based mechanized diagnosis, including lessened costs and increased access to care. To this end, keeping up high specificity alongside high sensitivity is basic. High sensitivity helps us to not miss patients with disease, but rather, in under-resourced clinics, high specificity is also essential to diminish congestion of clinics with patients without genuine disease. Google’s propelled calculation is the first such system to perform well on the two fronts.

In created social insurance settings, there is also a place for robotized diagnosis. Keep in mind that expansive populations in the Assembled States are not satisfactorily screened for DR. This means an extraordinary failure cost, exceptionally productive screening system could contact individuals who are as of now not being screened. One may envision screening kiosks in pharmacies and facility lobbies. This kind of service could prompt more patients accessing eye care. Also, if the nature of the system is sufficient, it may in the long run serve as a diagnostic guide to eye care professionals, improving the effectiveness of eye care conveyance.

READ THIS TOO:   Elon Musk States Why Artificial Intelligence Is A Much Bigger Threat To Civilization Than North Korea


Eye care providers may respond to the possibility of Artificial Intelligence in medication with skepticism or dread. Some experts we have spoken with expressed clear worries that these types of technology may diminish the general level of patient care, decrease eye care to “kiosk prescription,” or even turn into a danger to the livelihoods of providers who have invested so vigorously in their therapeutic training.

We trust that ophthalmologists and optometrists should see this technology with careful positivity. On one hand, it carries awesome potential. The Google technology has been produced to work synergistically with eye care providers. The potential benefits of a computerized DR screening program incorporate increased proficiency and scope of screening (ie, algorithms are programmed to withstand dull picture processing without weakness), access to screening in areas without eye care scope, prior discovery of referable diabetic eye disease, and likely lessening of general social insurance costs through prior location and mediation—also diminishment of vision loss. As a result, eye care providers may access more patients who require our one of a kind skill sets, leaving the screening to more proficient technologies.

READ THIS TOO:   General Overview of Artificial Intelligence and the Role of Natural Language Processing in Big Data + (the Yes' and Nos')

That being said, it is likely that this sort of technology, in its last version, will cause changes in clinical focus. As the “guardians of vision,” ophthalmologists and optometrists must take driving roles in deciding how best to incorporate these advances to improve patient care. As with most new technologies, early adopters will probably assume a part in the joining of AI. Thus lies the open door: to help shape this technology to be the absolute best for our patients.


Google has shown that its calculation can diagnose and grade one disease in a study setting. We still need to see the calculation’s certifiable execution. Also, there are numerous different diseases to take a shot at, as well as the basic aspect of finding the dangerous issues sometimes seen in screening images, such as visual melanoma. We are confident that clinical use of this sort of technology is close. Now, the goal for the technology is to improve access to minimal effort, fantastic eye screening technology, with a focus on underserved populations to decrease the weight of vision loss in the creating scene.

Previously, the question was, “Can a machine diagnose as well as a physician?” We are one step closer to knowing the answer to that question. We must ask now ourselves, as skilled specialists accused of preserving our patients’ vision, how would we use this technology to provide the best care for our patients?


Please enter your comment!
Please enter your name here