Diabetes Detected Early: How AI is Beating the Clock

How would you feel if you had to constantly monitor your blood sugar everyday, maintain a strict diet, through unpredictable waves of fatigue and numbness? Every year millions of people suffer because of Type 2 diabetes, a chronic condition where the body does not produce enough insulin or resists the effects of insulin to maintain normal glucose levels. This often leads to other health problems and complications in the future such as stroke and heart ache. Most people develop Type 2 Diabetes and experience the symptoms without even knowing it. Type 2 Diabetes is usually diagnosed through traditional routine blood tests run by doctors. These tests often can only detect someone becoming diabetic once they already have Type 2 diabetes. Whereas the AI model analyzes data from blood glucose monitors to track any unusual fluctuations and shifts in patterns that could reveal insulin resistance, a common cause of diabetes. These models help provide a nuanced view of data that help reveal who is at highest risk of progressing from healthy to prediabetic, or from prediabetic to full-blown diabetes. Compared to traditional tests, these AI models analyze constant data to help detect Type 2 Diabetes.

Type 2 Diabetes isn’t just a singular condition as thought by researchers. Beneath the surface it consists of many subtypes. In a Stanford Medicine study, researchers developed an artificial intelligence-based algorithm that uses data from continuous blood glucose monitors to parse three most common Type 2 diabetes subtypes, each reflected by a different physiologic factor. It includes insulin deficiency where the pancreas fails to make enough insulin, insulin resistance where cells don’t respond to insulin, and by a defect in the production of incretin, a hormone stimulating insulin secretion. The algorithm was used in a study of 54 participants, 21 of them prediabetic, and 33 of them healthy. The participants used continuous glucose monitors to produce data by measuring the rise and fall of blood sugars constantly. The algorithm was then applied to the data and used to identify any patterns within the data that corresponded to certain subtypes. The results from the algorithm were then compared with clinical data and researchers found out that the algorithm predicted the subtypes with greater accuracy than the traditional metabolic tests. The algorithm was able to correctly identify and detect the subtypes about 90% of the time. With continuous data from glucose monitors the algorithm was able to correctly predict and identify the subtypes more accurately compared to traditional tests.

Generally when trying to diagnose someone as diabetic or pre-diabetic, doctors used a HbA1c test. This test captures someone’s average glucose levels over the past few months to see their average blood sugar levels allowing them to diagnose someone as being pre-diabetic or diabetic. The problem with it though is that it can’t predict if someone will progress from healthy to pre-diabetic or pre-diabetic to diabetic. But at Scripps Research, scientists have created an AI model that uses a combination of data such as from continuous glucose monitors and clinical information to provide a more nuanced view of diabetes risk in people. The Ai is able to flag and predict early signs of diabetes risk unlike the HbA1c test. Usually in healthy individuals, blood sugar falls and rises smoothly whereas in people at risk for diabetes, their blood sugar spikes more often and takes longer to fall. This can happen before lab tests like HbA1c even notice it  but with the AI’s day to day tracking, a more detailed view of their health can be shown helping doctors prevent someone from becoming diabetic in advance. In a study of more than 1,000 people across the US, participants that were diagnosed as healthy, pre-diabetic, and diabetic were included. For 10 days in this remote clinical trial, participants wore Dexcom G6CM’s, tracked their exercise and meals, as well as sending blood and saliva samples to the researchers. With the data, a model was used to distinguish people who were healthy and those who had type 2 diabetes. With the model, researchers discovered that those with a higher resting heart rate were linked to diabetes and those with more diverse gut microbiomes had better glucose control. The Ai model also showed them that some prediabetic individuals were metabolically similar to those with diabetes and some similar to those who were healthy. By using data from continuous glucose monitors and other clinical information, the AI model was able to predict, detect, and identify any signs of diabetes risk which could help clinicians in the future create personalized treatment for patients. 

AI hasn’t just been predicting type 2 diabetes, it has been transforming diabetes care significantly. By analyzing large data sets and constant data from continuous glucose monitors, AI has been able to predict risks, make diagnoses, as well as helping enhance patient care. This can help clinicians and doctors create personalized treatment plans for patients in Diabetes care. A researcher from ScienceDirect found that AI was key in transforming a few specific domains in diabetes care, diabetes management and treatment, developing predictive models, public health interventions, lifestyle and dietary management, and etc.  By analyzing multiple sets of data and clinical information, AI has been able to help doctors diagnose patients by providing them a nuanced view of data showing any unusual patterns that could result in diabetes. Specifically, by using data from continuous glucose monitors over several months, AI can predict if someone will become diabetic allowing doctors to create personalized treatment plans in order to prevent it.

AI has completely transformed Diabetes care. Compared to traditional tests, AI has been able to analyze data and detect any risk of Diabetes much earlier allowing for prevention plans to be made much earlier. These AI models will completely change the future allowing for much more advanced treatment and care for diabetics. By allowing for much earlier prevention and care, it will help save millions of lives, possibly even yours.