Lumos: An Open-Source Device for Wearable Spectroscopy Research

Co-authors: Amanda Watson, Anush Lingamoorthy

Spectroscopy and its use in wearable devices

When we think of wearable health devices of the future, we envision technology capable of tracking and analyzing intricate physiological changes in real time that are currently not possible, such as blood glucose and blood alcohol. One possible way to detect these physiological changes is through the use of spectroscopy.

Spectroscopy is the study of the interaction between electromagnetic radiation and matter, and is already widely used in medicine through devices like X-rays, CT scans, and MRI. By measuring the interaction of this radiation with the human body, we can learn a great deal about what is happening internally.

X-rays are one example of an imaging technique that uses spectroscopy, specifically X-ray radiation, to provide information on skeletal muscle. Photo by Umanoide on Unsplash.

Thus far, spectroscopy has been an underused modality limited to providing transient snapshots of data. Wearable devices offer the promise of continuous monitoring, which contrary to the aforementioned imaging technologies, can passively collect data throughout the day, creating longitudinal datasets that provide more insight into the monitored parameter.

Current wearable devices on the market (Apple Watch, FitBit, etc) use some form of spectroscopy already for physiological monitoring, such as red and IR wavelengths for pulse rate and SpO2. However, these devices lack the spatial resolution to analyze physiological interactions with wavelengths across the visible spectrum, which has previously limited research in the area.

Currently, commercial wearable devices collect spectroscopy data in green (~550nm), red (~700nm), and/or infrared (~900nm) wavelengths only, leaving a lot of room for exploration in other regions of the visible spectrum.

Lumos, a next-gen wearable spectroscopy device

Enter Lumos. We have created a device that utilizes LEDs and photodiodes across the visible spectrum to enable on-body spectroscopy research. The device can be fit into both smartwatch and finger clamp form-factors and is intended to facilitate clinical research into previously unresearched spectroscopic interactions within the human body, with potential applications to glucose, alcohol, and blood pressure monitoring.

Lumos is a wearable optical spectrometer that enables non-invasive health monitoring in the real-world. The device can be fit into both smartwatch and finger clamp form factors.

We verified the performance of Lumos by measuring the spectral response against mediums of known wavelengths; the device demonstrated a mean absolute error of 13nm with a standard deviation of 8nm, indicating strong performance to appropriately measure and characterize measured mediums.

Lumos spectral responses against mediums of known wavelengths (blue and green in this example). The device demonstrated a mean absolute error of 13nm with a standard deviation of 8nm.

Glucose Pilot Study

We also conducted a pilot study to analyze the performance of Lumos at measuring blood glucose non-invasively. Glucose values were measured on one subject over the course of an hour and compared with a commercially available glucometer. We found high statistical significance ( p < 0.05) and high Pearson correlations (0.843 and -0.927 respectively) for 470nm and 515nm readings. Other readings exhibited lower correlations (0.359 for 680nm for example) that may not be appropriate for a single point comparison with glucose, but may be used along with other wavelengths to provide information about internal blood glucose dynamics.

Performance of 475nm, 515nm, and 680nm wavelengths relative to a glucometer glucose reading in one subject over an hour study.

The device has been made completely open-source and is available on GitHub. The full peer-reviewed article on Lumos can be found here.

SpectraVue — rapid data visualization and analysis for Lumos

Due to the density of data Lumos measures, we found it can be hard to rapidly analyze and visualize its response, especially for those with a non-technical background. To minimize barriers for clinical research we developed SpectraVue, a lightweight, easy-to-use platform that processes user-uploaded Lumos data and visualizes it for further inspection. The platform was built using Plotly Dash.

SpectraVue: an interactive, minimal platform used for visualization and analysis of 3D spectroscopy signals from the Lumos device.

The application has three modes: static mode, which returns a graph with the mean counts over the entire provided dataset, animation mode, which returns an animation with the spectral responses at each respective time-point, and biomarker mode, which returns an animation of the spectral responses as well as a graph of an associated biomarker. Biomarker mode is especially useful as it provides an analysis of specific spectral regions and their correlation with a specific biomarker. An example of the biomarker mode output is shown below.

Biomarker output of SpectraVue: an animation of the Lumos spectral response is shown on the left, with the associated clinical biomarker on the right.

We intend to continue to improve SpectraVue based on user feedback; future features of the platform include additional data parsers, data processing interactivity, additional data analysis tools (such as correlation analysis), and mobile device adaptation for real-time viewing.

As with Lumos, SpectraVue has been made completely open-source. The full peer-reviewed article is currently being published and will be shared soon!