A pioneering noninvasive radiofrequency sensor, coupled with machine learning technology, has demonstrated the potential to accurately measure glucose levels in individuals with diabetes, based on presentations at the AACE Annual Meeting.
The innovative device, developed by Know Labs, employs a radiofrequency dielectric sensor that swiftly scans various frequencies using dielectric spectroscopy. This process entails detecting voltage values at each frequency to enable real-time monitoring of glucose levels. Dominic Klyve, PhD, from Central Washington University and Know Labs Data Science & Engineering, highlighted the device’s role in addressing the economic costs, discomfort, and medical waste associated with continuous glucose monitoring (CGM).
In a clinical trial involving over 30 participants with prediabetes and type 2 diabetes, the radiofrequency sensor was utilized to continuously scan participants’ forearms during a glucose tolerance test while collecting venous blood samples every 5 minutes for comparison. By training a machine learning model using 520 paired values obtained in the study, researchers achieved a Mean Absolute Relative Difference (MARD) of 11.1%, signifying the model’s accuracy in estimating glucose levels when compared to blood samples.
Furthermore, the machine learning model demonstrated consistent accuracy across normoglycemic and hyperglycemic ranges, with a MARD of 9.5% for glucose values below 70 mg/dL in the hypoglycemic range. The model’s performance was evaluated using a surveillance error grid, wherein 82.3% of measurements were classified in the lowest risk grade, and no measurements fell into higher risk categories.
Dr. Klyve emphasized the potential of this noninvasive sensor in minimizing waste, reducing costs, and offering painless glucose measurements. Know Labs intends to conduct extensive external clinical studies to assess the sensor’s performance under various real-world conditions and glycemic ranges, ultimately contributing to the development of a wearable CGM device.
Overall, the integration of a noninvasive radiofrequency sensor with machine learning showcases significant promise for accurate glucose monitoring, heralding a potential breakthrough in diabetes management by providing a more accessible, cost-effective, and needle-free alternative to conventional CGM methods.
