Researchers have developed a small near-infrared spectrometer designed to identify substances, measure chemical concentrations, and support noninvasive biomarker monitoring.
The device, described in Nature Photonics, was developed by researchers from the University of Cambridge and GlitterinTech Limited. Known as ConvSpec, it uses a mathematical principle called the convolution theorem to convert optical signals into spectra. The approach is intended to reduce the size, cost, and computational demands associated with conventional spectrometers.
The fully packaged instrument measured 4.8 × 6.2 × 0.6 cm, weighed 16.5 g, and had an estimated component cost of approximately $10. It covered a near-infrared wavelength range of about 1,200 to 1,700 nm and collected and processed each spectrum in less than 0.4 seconds.
In laboratory testing, the device distinguished spectral peaks separated by 5.8 cm−1 and recovered more complex spectral patterns with low relative error. Tests conducted at temperatures from −20 °C to 80 °C indicated that temperature-related changes could be corrected using the system’s software.
The researchers evaluated the instrument using plastics, pharmaceuticals, coffee, flour, and tea. Each sample was measured 80 times, and the resulting data were divided into training and testing sets. Machine learning models correctly classified all samples in the internal test sets.
The system also measured concentrations of ethanol, lactate, glucose, and ethylene glycol solutions ranging from 0.1 to 40 percent. Mean absolute errors ranged from 0.006 to 0.017 percent. In glucose solution experiments, ConvSpec predicted concentrations more accurately than the commercial dispersive and Fourier-transform benchtop spectrometers tested.
To examine possible medical applications, the team used a wrist-worn probe to collect approximately 6,000 spectra across 126 participant sessions. The device measured light reflected from the skin, while established methods provided reference values.
The reported mean absolute errors were 2.49 percent for skin moisture, 1.70 mg/dL for blood alcohol, 0.81 mmol/L for lactate, and 0.36 mmol/L for glucose. These findings suggest that a compact spectrometer could eventually support portable or wearable monitoring, although the study did not establish the device as a clinical diagnostic test.
The human testing was exploratory and conducted under controlled conditions in Asian adults. Although the glucose models performed well on the study’s test data, they could not reliably predict glucose levels in new participants, whose individual physiological differences affected the measured spectra.
Newsletters
Receive the latest analytical science news, personalities, education, and career development – weekly to your inbox.
