Skip links

AI and Diabetes: Promising Research Calls For Jubilation

How Technology Is Beginning to Help Us Address a Major Disease


As incidence of diabetes increases around the world, it is clear that medicine could use assistance in its diagnosis and maintenance. While there are over 425 million people living globally with the disease, several million more are yet to be identified. Information from the WHO has shown that diagnosing and managing the disease burden is extremely capital intensive.

One of artificial intelligence’s major strengths lies in its ability to utilize large chunks of data as it hones and carries out activities, thus granting users fact-based results that pave the way toward better analysis. In this article, we shall discuss how AI will help us recognize, manage, and treat diabetes.

Artificial Intelligence in Action

While there are several existing and potential uses of AI when it comes to the diabetes, we will focus on four important applications, as identified by Dankwa-Mullan et al. (2019):

Attacking Diabetic Retinopathy Via Automated Retinal Screening

Automated Retinal Screening is the use of AI-based and machine learning tools like multiple perception, convolutional connected neural networks (CNN), random forest, support vector machine (SVM), and other methods for the detection of diabetic retinopathy, maculopathy, exudates, and other abnormalities from normal findings.

An Example of a banner with a button

Get the book for free

Diabetic retinopathy (DR) is an eye disease that causes moderate to severe vision loss. It has been identified as the leading cause of blindness among working-age people suffering from long-standing diabetes.  

Perhaps the biggest danger associated with DR is that its symptoms do not reveal themselves until the disease reaches an advanced stage, which increases our global burden. For some time, however, scientists and engineers have utilized computer-based analysis of fundus images in screening. Now, as they incorporate the use of AI-based techniques while doing so, the chance of avoiding blindness from DR has increased by a phenomenal 90%.

AI can reduce the chances of developing permanent blindness from diabetic retinopathy by 90%.

-Padhy et al. (2019)

The AI-based system utilizes machine learning algorithms powered by a convolutional neural network (CNN) and a massive-training artificial neural network (MTANN) to decide whether a patient would benefit from a referral. The machine is cheaper than seeking an ophthalmologist to conduct an in-person screening. In line with this data, in 2018, the FDA approved the use of an AI algorithm developed by IDx that employs a Topcon fundus camera for DR identification.

Clinical Decision Support and Combating Diabetic Neuropathy

The earliest AI algorithms were geared toward supporting physicians in managing clinical and patient data. Among many other resourceful uses all throughout medicine, recent AI has focused on the detection and monitoring of diabetes and comorbidities such as neuropathy and wounds.

Occurring when high blood sugar (glucose) injures the body’s nerves, diabetic neuropathy is a form of nerve damage. The condition tends to produce a wide range of symptoms that range from numbness and pain in the legs and feet to issues with the digestive system, blood vessels, heart, and urinary tract.

In November 2019, a group of scientists published a paper about an artificial intelligence-based deep learning algorithm useful in the diagnosis of diabetic neuropathy; it used corneal confocal microscopy. This powerful algorithm leveraged a convolutional neural network with data augmentation to automate the quantification of the corneal sub-basal nerve plexus, which supported diagnosis. The researchers trained the algorithm using a high-end graphics processor unit on 1698 corneal confocal microscopy images, and it was tested on 2137 images.

Ultimately, the exciting AI-powered technique provided a rapid and precise localization performance for the quantification of corneal nerve biomarkers. It presented as a potential clinical screening program for diabetic neuropathy.

But Can AI Help in Predicting Population Risk?

Yes, it seems so!

Artificial intelligence has been used in the identification of diabetic subpopulations at higher risk for complications, hospitalizations, and readmissions. Machine learning algorithms point to ten top variables that can accurately predict the risk of heart failure in patients suffering from diabetes. Some of those predictors are weight, age, hypertension, and diabetes control.

In one research study, scientists and engineers developed and validated a machine-learning-based model that readily integrated clinical, laboratory, and electrocardiographic variables for the prediction of heart failure among outpatients with type-2 diabetes mellitus. This algorithm could go a long way in determining which patient is more likely to experience heart failure and why, which will assist physicians as they plan next steps.

AI and Self-Management

There is widespread utilization of patient self-management tools like the AI-improved glucose sensors, activity and dietary trackers, and artificial pancreases. Let’s take a closer look:

a.        AI-improved glucose sensors:

These glucometers use AI and biomedical engineering principles to achieve accurate and real-time monitoring of blood glucose levels.

b.         Activity and dietary tracking devices:

IBM Watson, in partnership with Medtronic, has created an AI technology called the Sugar IQ diabetes assistant. This device was manufactured for investigational use in the Guardian Connect Continuous Glucose Monitor and was unveiled at ADA. Sugar IQ is an interactive app that employs cognitive computing and analytic technology for delivering personalized, simplified daily diabetes feedback to each of its individual users. It features a combination of AI, advanced analytics, and diabetes technology to constantly evaluate and update patients about their glucose levels so they can make the necessary adjustments to their food intake, daily routines, and insulin dosages.

c.        Artificial pancreases:

An artificial pancreas releases insulin in response to fluctuating blood glucose levels. It works similarly to a human pancreas, only it’s controlled by AI-based algorithms. Following the findings of a research published in December 2019, the FDA recently approved the Control-IQ artificial pancreas system for the automatic monitoring and regulation of blood glucose levels in patients suffering from type 1 diabetes. Capable of tracking glucose levels with a continuous glucose monitor (Dexcom G6 CGM), the pancreas is manufactured by Tandem Diabetes Care. For certain people, the machine can cut out the necessity for frequent fingerstick blood glucose testing and multiple insulin injections.


With the recent advances in AI, it is clear that the way we diagnose and manage those with diabetes is evolving rapidly. Surely, the near future holds many more exciting innovations.

Thoughts? Questions? Ideas? We would love to hear from you below, and don’t forget to share this post with a friend or colleague.

Leave a comment

This website uses cookies to improve your web experience.