A new AI model looks to diagnose type 2 diabetes with just 12 hours of glucose profile data. If proven to work at scale, it could assist medical professionals in making quicker diagnoses of the disease before impaired glucose tolerance. Worldwide, type 2 diabetes affects around 463 million people. The disease can damage organs, impair their functions, or worse if left untreated. This includes damage to the eyes, kidneys, central nervous system, and heart. Blood tests are used to measure a person’s average blood sugar level over a three-month period to discover if someone is developing the disease. This model cuts the time down to just 12 hours.
Speaking with Medical News Today, Jouhyun Clare Jeon, Ph.D., principal investigator at Klick Applied Sciences and lead author of the study speaks on the benefits to both patients and medical professionals. “I believe our method offers a lot of potential to be used as a novel tool to aid healthcare providers in their own decision-making, especially for remote or virtual care of patients. For the general public, our method could not only be used for monitoring and early screening but alerting a patient of their risk of developing diabetes.”
For this study, the researchers used a data set from 436 participants in India who wore a CGM device. These devices help people with diabetes to easily and regularly monitor their blood sugar levels. Over the course of 12 days, they wore the devices which monitored and tracked blood sugar levels. Other data points collected also included their sex, age, and BMI or Body Mass Index. Within that group, 172 were already diagnosed with type 2 diabetes, 87 had pre-diabetes, and the last 177 were healthy adults. The researchers created AI prediction models based on different blood glucose level time durations from this data. The windows of data included 12, 24, 72, 168, & 288 hours. After analyzing the data, they found that the CGM data was 1.21, 1.34, and 1.17 times more accurate than simple demographic data in identifying type 2 diabetes, pre-diabetes, and non-diabetic individuals.
They found that their 12-hour model was as effective as the other four model windows used, even though it was the shortest duration. After optimizing the 12-hour model, they were able to identify over 80% of those with diabetes, pre-diabetes, and non-diabetic individuals. But, this isn’t a silver bullet for quick diabetic diagnoses just yet. The study’s lead author notes that their data size of around 400 patients was too small and to continue this research they will need to continue with greater data sets. “Our findings are developed based on about 400 patients’ CGM signals. Further evaluation is required using an independent larger cohort and bigger population data to generalize our method. However, we are encouraged by the results and look forward to our continued work in this area.”
Patients who wear the CGM can learn the results while at home thanks to this AI model. By reducing the lag time between data and results, treatment and preventative medicine can engage with the patient much faster, and in the long term reduce the chance of complications related to non-treatment.
Read more data science news here.