Overcoming the challenges of non-invasive diabetes management through wearable devices

Diabetes is an increasingly prevalent disease, currently affecting more than 500 million adults worldwide. There is still no cure for type 1 or type 2 diabetes, so patients should monitor their BGL regularly to check their BGL. BGL measurement devices that rely on painful finger pricks have been the gold standard for decades, but modern technology is slowly opening the door to better alternatives.

Many researchers have proposed non-invasive methods to monitor BGL using widely available wearable devices such as smartwatches. For example, by applying LEDs and photodetectors in certain smartwatches to the skin, it is possible to measure oxyhemoglobin and hemoglobin pulse signals and calculate metabolic indices from which BGL can be estimated. However, due to the small size and limited power of smartwatches and similar wearable devices, the data quality of the measured signals tends to be very low. Furthermore, since these devices are attached to the limb, daily movements introduce measurement errors. These issues limit the accuracy and clinical application of such wearables in diabetes management.

A team at Japan's Hamamatsu Photonics Co., Ltd. has been actively researching this problem in search of effective solutions. In a recent study led by R&D engineer Tomoya Nakazawa, Biomedical Optics Journal(JBO), who conducted a detailed theoretical analysis of the sources of error in metabolic index-based methods. Based on this analysis, we implemented a new signal quality index to filter out low-quality data as a preprocessing step, thereby improving the accuracy of the estimated BGL.

”Smartwatches have been widely adopted by different regions and age groups, and with the global increase in diabetes patients, the need to meet the growing global demand for non-invasive smartwatches due to individual and individual differences. A method to improve signal quality that is easy to implement and apply is essential. blood sugar monitoring device” explains Nakazawa, explaining the motivation behind the research.

First, the researchers showed mathematically that the mismatch between two types of phase delays in oxyhemoglobin and hemoglobin pulse signals calculated using different methods provides a good measure of the influence of noise. Next, they considered two main sources of phase error: background noise level and estimation error introduced by sampling at discrete intervals. After formalizing these sources of error, we calculated their impact on the estimated metabolic indices.

The proposed screening method includes implementation of phase estimation and metabolic index error thresholds. Data chunks that exceed the set threshold are discarded, and missing values ​​are approximated using other means based on the remaining data.

To test this strategy, the researchers conducted a long-term experiment using sensors in commercially available smartwatches to monitor BGL during an “oral challenge” in healthy people. In each of the 30 tests conducted over a four-month period, subjects fasted for two hours before consuming high-glucose foods. Their BGL was measured using a smart watch and a commercially available continuous glucose monitoring sensor, the latter of which was used to obtain baseline values.

In particular, preprocessing the data with the proposed screening method significantly improved the accuracy. Using the Parkes error grid technique to classify measurement errors, a substantially higher proportion of data points fell into zone A when screening was applied. This refers to clinically accurate values ​​that lead to correct treatment decisions. “By employing a screening process, we improved the BGL estimation accuracy in our smartwatch-based prototype,” said Nakazawa. “Our technology has the potential to facilitate the integration of wearable continuous BGL monitoring into devices such as smartwatches and smart rings, which typically have size limitations.”,” he added, emphasizing the impact of research activities.

The research team also noted some of the current limitations of smartwatches that lead to inferior performance compared to smartphone camera-based technology. Although the proposed method could certainly help improve the performance of the former, improvements in the hardware of photodetectors and amplifier circuits would make wearable electronics more attractive and clinical for monitoring BGL. could go a long way toward making it an acceptable option.

Further research in this area will help provide people with diabetes with powerful tools to better manage their condition and lead to a higher quality of life.