People with diabetes face a common challenge: dealing with hypoglycemia. It is a condition that occurs when blood sugar levels drop below the level necessary for humans to function properly. In fact, people with type 1 diabetes typically experience two episodes of mild hypoglycemia per week. Symptoms such as sweating, fatigue, dizziness, and tremors occur during these episodes. In this case, if you do not start consuming sugar right away, there is also a risk that your condition will worsen.
Treatment of hypoglycemia has previously been difficult for patients. However, by employing the latest artificial intelligence technologies, we are finding new ways to reduce the impact of these episodes on people’s everyday lives.
Hypoglycemic Event Detection
I lead a team at Texas AandM University in Qatar trying to address the problem of hypoglycemia in diabetes patients. In collaboration with the Qatar National Research Fund, we developed an AI-based model that shows hypoglycemic conditions.
One of the main risks of hypoglycemia is that people cannot tell what is happening to them, especially in children who feel like something is wrong but often cannot comprehend and understand the symptoms they are experiencing with diabetes. that these symptoms are from low blood sugar, so we set out to develop a machine learning framework that would effectively identify hypoglycemic events.
Smart technology can be critical to medical innovations like this one. Based on the premise that physical tremors could be a sign of hypoglycemia, our team collected data on tremor from patients using smartwatches and compared it to data collected by continuous glucose monitoring devices – small devices that many patients are among with diabetes they have implanted them directly into the skin to continuously monitor glucose levels. Next, we created an AI model that correlated the patterns between the frequency and strength of the tremors, including whether each participant had low or high blood sugar levels.
The results were compelling: physical tremors are a clear sign of hypoglycemia, and by monitoring the tremors more closely, we can help people with diabetes inefficiently manage their condition.
Episode Prediction
While detecting hypoglycemia is helpful, the real potential of technological innovation lies in improving our ability to predict when hypoglycemia will occur. If an episode is already happening, we are already too late. We can make the biggest difference in the lives of people with diabetes by enabling them to predict when they are approaching a seizure and giving them the ability to prevent it, rather than simply controlling it when it happens.
The next step in our research was therefore to develop the model of artificial intelligence from a primarily analytical to a predictive function. We have achieved this by developing a custom application that interprets the data collected from each patient and are currently able to predict episodes that occur 30 seconds to a minute earlier. Our algorithms have been shown to be 82% cure in adults and 86% cure in children.
We designed the TREMOR app to be very easy to use and we hope it will be available in app stores in the coming months to help patients manage their diabetes more proactively.
A low-cost wearable tech solution
To be effective, any technology that detects hypoglycemic events must always be able to monitor the diabetic. Our data collection was based on smartwatches and with the rapid rise of wearable technology in recent years, similar devices are a natural solution for patiently monitoring their diabetes continuously. There are downsides, however, as the cost of smartwatches poses a challenge.
In order to ensure that innovations in diabetes management are accessible to all, inexpensive alternatives must be found. Our researchers are currently making progress in developing a new portable device for patients with diabetes that will be much cheaper. Bracelet or ring that transmits shaking data to smartphones via Bluetooth.
Changing lives
The AI-based application system we developed is the first of its kind to enable hypoglycemia monitoring and prevention in real time. Designed to be non-invasive and with simple warnings warning users of dangerous glucose levels, it is designed to make diabetes management easier for anyone living with the condition. For parents in particular, such solutions offer a quick and effective way of monitoring their child’s blood sugar level, which certainly gives them security.
Before this project, significant advances had been made in predicting hypoglycemic episodes. Continuous glucose monitoring devices can currently predict episodes with reasonable accuracy up to twenty minutes in advance. However, applications like TREMOR can enable self-monitoring and self-monitoring. control their condition. without the time consuming and sometimes stressful experiences of processes like collecting blood samples. We expect the demand for this type of non-invasive technology to increase in the years to come.
However, improving warning time for non-invasive technologies will be crucial, which is the focus of our next phase of research. We are currently working on developing our algorithms so that they can predict episodes up to 15 minutes before the event, which is approaching the time reached by invasive options.
Combining the efforts of doctors
If we can help patients with diabetes cope better with their illness, we can not only improve their quality of life, but also relieve unnecessary burdens on health systems. Improper treatment of hypoglycemia can all too easily lead to more serious symptoms that require medical attention; But with the right tools, most people could avoid this and set doctors free to help patients with the most severe cases.
Similar AI-based wearable technology solutions targeting a wide variety of health conditions are expected in the months and years ahead. The difference this could make in patient lifestyle, cost savings and capacity pressures is enormous.
