Non-Invasive Optical Glucose Sensing Technologies

Introduction to Non-Invasive Optical Glucose Sensing Technologies

Non-invasive optical glucose sensing tries to read glucose by shining light into the skin and measuring what comes back, with no needle and no fingerstick [1]. The physics is the hard part. Glucose makes up only about 0.07% to 0.1% of plasma, and the water, pigments and proteins around it interact with the same light [2]. That problem, called signal specificity, runs through every method below.

Principles and Technologies

Each optical method leans on a property that belongs to glucose itself: how strongly it absorbs certain wavelengths, how it rotates polarized light, or the particular shift it gives to scattered light [1]. Three approaches get the most attention:

• Near-Infrared Spectroscopy (NIRS): NIRS measures how much near-infrared light the tissue absorbs. Water is the obstacle. In the near-infrared it has two strong absorption peaks, at about 1350 to 1520 nm and 1790 to 2000 nm, and tissue holds far more water than glucose [2].
• Raman Spectroscopy: Raman methods look for the small wavelength shift glucose gives to scattered light, a molecular "fingerprint" whose strongest peak sits at 1125 cm⁻¹ (wavenumbers, the unit for that shift) [1]. That fingerprint is specific, but faint. Almost all scattered light comes back unshifted, so a Raman setup needs an intense light source and a very sensitive detector, and background fluorescence from proteins can be as strong as the glucose signal or stronger [1].
• Photoacoustic Spectroscopy (PAS): PAS sends pulses of light into the tissue. Light that glucose absorbs turns into a tiny heat pulse, and that heat produces a pressure wave an acoustic detector can pick up [1]. Because only absorbed light makes sound, scattered light doesn't distort the reading the way it does in purely optical methods [1]. Timing can also tell the detector how deep a signal started: one sensor presented at a 2025 workshop uses time-gating to target blood-rich skin layers, and had reached a first pilot study in healthy volunteers [3].

Innovations and Market Developments

Two efforts show where the field has tried to go next: shrinking the optics onto a chip, and moving to mid-infrared light.

• Silicon Photonics: Silicon photonics builds lasers, light guides and detectors onto silicon chips. In May 2021, Rockley Photonics described a program to make chip-size, laser-based infrared sensors small enough for a smartwatch, designed to measure glucose, alcohol and oxygen levels, and said Apple was the largest of six customers it had contracts or engagements with [4]. As of October 2026, the FDA has not authorized any smartwatch that measures blood glucose on its own [5]. Rockley filed for Chapter 11 bankruptcy protection in early 2023 and emerged that June, saying its glucose monitoring program was continuing [6]. We found no update from the company after 2023, so its current status is unclear.
• Photothermal Deflection: DiaMonTech's device uses a quantum cascade laser, a compact semiconductor laser, to send mid-infrared light about 100 µm into the skin. Glucose in the fluid between skin cells absorbs that light and warms the tissue slightly; a second laser beam bends as it passes through the warm spot, and the size of the bend tracks the glucose level [7]. In a 2025 study of the tabletop D-Base prototype in 35 adults, the best algorithm reached a MARD (mean absolute relative difference, the average percentage gap between a device's readings and a reference measurement) of 20.7% and 19.6% in two sessions [7]. All participants were white, and a wearable version was still at the concept stage [7].

Critical Challenges

The research and regulatory record keeps returning to three problems:

• Regulatory Approval: On February 21, 2024, the FDA warned people not to use smartwatches or smart rings that claim to measure blood glucose without piercing the skin. The agency said it has not authorized, cleared or approved any smartwatch or smart ring intended to measure or estimate blood glucose on its own, and that inaccurate readings can lead to errors in diabetes management [5]. As of October 2026 that notice was still posted with no update [5]. Which FDA-authorized meter or CGM fits your treatment is a decision for your care team.
• Interference and Variability: Changes in tissue temperature alter optical absorption, and melanin (the pigment that gives skin its color) absorbs light along with glucose [1]. In the DiaMonTech study, the researchers traced unstable signals to uneven skin, movement and skin-care products, and they listed sweat glands, scars and hair among the features they want to detect and work around [7]. A 2020 review added that a model tested on simulated tissue may not hold up across people of different ages, skin colors and genders [2].
• Accuracy and Reliability: MARD is the accuracy figure quoted most often, but neither FDA's integrated CGM (iCGM) rules nor ISO 15197 sets a MARD cutoff. FDA's iCGM special controls use percent-within limits instead; across the measuring range, for example, the lower 95% confidence bound for readings within ±20% of blood glucose must exceed 87% [8]. ISO 15197:2013, the standard for fingerstick meters, also counts results inside set bands of ±15 mg/dL or ±15% [9]. A 2026 workshop report warned that MARD figures from early studies, often measured against fingerstick meters, can mislead when set beside validated clinical systems, and that isolating glucose from other variables is still a basic physics problem in a wearable [3].

Conclusion

The physics behind near-infrared, Raman, photoacoustic and photothermal sensing is well described [1]. What's still missing is a device that pulls glucose's faint signal away from water, pigment, temperature and skin changes during everyday wear. The most complete clinical result we read, DiaMonTech's MARD of about 20%, came from a tabletop prototype rather than something you could wear [7]. Until that changes, the FDA's advice is to measure blood glucose only with FDA-authorized devices [5].