Sweet Light: Can Spectroscopy Finally Kill the Finger-Prick for Blood Glucose?

Abstract

More than 500 million people worldwide manage diabetes, and nearly all of them still draw blood to do it. For three decades, spectroscopists have promised a way out: shine light through skin, read the faint signature of glucose molecules, and skip the lancet entirely. That promise has proven maddeningly difficult to keep. This review surveys the optical arsenal now converging on the problem, near-infrared and mid-infrared absorption, Raman scattering, and photoacoustic detection, paired with the machine-learning models that make weak, noisy signals usable. Drawing on recent Spectroscopy Online reporting and the primary literature, it traces where each technique stands, what is stalling regulatory approval, and why Apple, MIT, and university labs across three continents believe the finger-prick's days may finally be numbered.

Introduction

Diabetes management runs on a single, unglamorous ritual: puncture the skin, squeeze out a drop of blood, and feed it to a test strip. It works, but it hurts, it costs money over a lifetime of daily testing, and it discourages the frequent monitoring that best controls blood sugar swings. The dream of a needle-free alternative is not new. Researchers have chased noninvasive glucose sensing since the early 1990s, betting that light itself could substitute for the lancet.¹ The rationale is simple: glucose absorbs and scatters light in characteristic, physics-defined ways, and if those signatures can be teased out of skin, blood, or interstitial fluid, a laser or an LED could replace the needle altogether.

The physics, unfortunately, is anything but simple. Glucose is a weak optical absorber that hides behind water, hemoglobin, lipids, and collagen, all of which vastly outweigh it in tissue and swamp its signal.² Decades of effort and, by some estimates, billions of dollars in investment have produced no fully noninvasive glucose monitor with FDA clearance for clinical use.² Yet the pace of publications has accelerated sharply in just the past five years, driven by better lasers, faster detectors, and machine-learning models capable of pulling a coherent glucose signal out of what once looked like noise.³ This review surveys that progress across the major spectroscopic modalities, highlights the engineering breakthroughs reported in the trade and peer-reviewed literature, and asks candidly why “almost there” has lasted so long.

Near-Infrared Spectroscopy: The Front-Runner That Won't Quite Cross the Line

Near-infrared (NIR) spectroscopy, spanning roughly 780 to 2500 nm, remains the most heavily pursued noninvasive glucose technology, largely because NIR light penetrates several millimeters into tissue while still carrying overtone and combination bands from glucose's C-H, O-H, and N-H bonds.³ A miniaturized NIR system reported through fingertip transmission has demonstrated that compact, low-cost optics can approximate laboratory-grade accuracy, a milestone for portability.⁴ Similar infrared devices aimed at fingertip or forearm measurement continue to appear, each promising to replace the lancet with a few seconds of illumination.⁵

The obstacle has never been detecting a signal, it has been calibrating one reliably across different people, skin tones, hydration states, and temperatures. A molar absorptivity modeling approach recently reported improves on that front by building physically grounded calibration curves rather than relying purely on empirical regression, an important step toward devices that generalize beyond the lab that built them.⁶ Complementary work using science-based calibration strategies has shown that incorporating a priori knowledge of the glucose absorption signal, rather than trusting statistics alone to find it, meaningfully improves multivariate model robustness in skin-reflectance measurements.⁷ Even so, NIR's Achilles' heel persists: the glucose signal is a small perturbation on a large, variable water background, and no calibration trick fully escapes that constraint.²

Raman Spectroscopy: Reading Glucose's Molecular Fingerprint

Where NIR absorption is broad and overlapping, Raman spectroscopy offers something sharper, a vibrational fingerprint specific enough to distinguish glucose from the surrounding biochemical clutter.⁸ Early transcutaneous Raman work using 785 nm excitation demonstrated strong correlation between spectral intensity ratios and blood glucose concentration in live animal models, validating the basic premise that Raman could see glucose through intact skin.⁸ A landmark 2020 study went further, directly observing glucose's Raman fingerprint in living tissue for the first time and confirming that signal changes tracked reference glucose concentrations across controlled clamping experiments in swine.⁹

More recent work has pushed the method toward clinical usability. A 2026 study demonstrated an integrated in vivo Raman platform validated during oral glucose tolerance tests, an important test of real-world reliability rather than static single-point accuracy.¹⁰ Perhaps the most consequential recent advance comes from a Massachusetts Institute of Technology (MIT) research group, which has developed a band-pass Raman architecture that eliminates the bulky diffraction gratings and CCD detectors standard Raman systems require, replacing them with off-axis 830 nm illumination and an intraspectrum reference scheme, shrinking a benchtop instrument toward something resembling a compact, point-of-care device.¹¹ Raman's chemical specificity and its minimal interference from tissue water make it, arguably, the most scientifically elegant of the noninvasive candidates, even if its instrumentation has historically been the least portable.

Mid-Infrared and Photoacoustic Spectroscopy: Chasing Specificity at a Cost

Mid-infrared (MIR) light, covering the molecular fingerprint region beyond 2500 nm, offers even sharper glucose-specific absorption bands than NIR, but at the cost of shallow tissue penetration, MIR light barely reaches the epidermis before it is absorbed.² Photoacoustic detection sidesteps that limitation elegantly: rather than measuring transmitted or reflected light directly, it measures the acoustic pressure wave generated when absorbed light causes localized, pulsed heating, a signal that can be read from deeper tissue layers than direct optical detection allows.

An early proof of concept used a quantum cascade laser tuned to glucose's fingerprint region between 1000 and 1220 cm⁻¹, coupled to a photoacoustic gas cell, to quantify epidermal glucose concentrations across the full clinically relevant range.¹² More recent designs have paired dual quantum cascade lasers with machine-learning classifiers, reporting prediction accuracies above 96% even in the presence of interfering blood components.¹³ A simplified single-wavelength near-infrared photoacoustic variant operating at the 1625 nm glucose overtone has likewise shown that photoacoustic systems need not always require complex multi-laser architectures to remain accurate.¹⁴ Photoacoustic spectroscopy's central appeal is that it partially decouples chemical specificity from penetration depth, though building compact, affordable acoustic detectors remains a nontrivial engineering task in its own right.

Machine Learning and the Quiet Revolution in Signal Processing

If there is a single thread tying together the recent surge in noninvasive glucose publications, it is not a new light source, it is better mathematics. A widely reported 2024 study combining multi-pass Fourier-transform infrared (FT-IR) measurement, a quantum cascade laser source, two-dimensional correlation spectroscopy, and machine-learning classification achieved a reported 98.8% accuracy in blood-glucose testing, a figure that, if it replicates broadly, would approach clinically actionable performance.¹⁵ The peer-reviewed version of that work details how replacing a conventional infrared lamp with a quantum cascade laser and adding multiple internal reflections improved signal-to-noise by several-fold before the machine-learning layer was ever applied, a reminder that algorithmic gains and hardware gains tend to compound rather than substitute for one another.¹⁶

This pattern, hardware improvements paired with increasingly sophisticated chemometrics, appears across nearly every modality discussed here, from NIR calibration models to Raman ensemble learning to photoacoustic classifiers. It suggests that the ceiling on noninvasive glucose accuracy has as much to do with data science as with optics.

Clinical Validation and the Road to Regulatory Approval

Publishing an accurate laboratory result is not the same as clearing a regulatory bar. A 2025 clinical validation study evaluated a midinfrared, photothermal-detection glucose system in real patients and reported accuracy comparable to early-generation minimally invasive glucometers, a meaningful signal that some optical methods are approaching, rather than merely gesturing toward, clinical-grade performance.¹⁷ A broader 2025 review of non-invasive sensor technologies, spanning optical and antenna-based microwave approaches, likewise concluded that while individual demonstrations increasingly meet accuracy benchmarks, cross-population reproducibility, device miniaturization, and long-term stability remain the unresolved gatekeepers to commercialization.¹⁸

It is worth noting that optical methods are not competing in a vacuum. Wearable electrochemical glucose sensors, which sample interstitial fluid through microneedles or sweat rather than using light at all, have matured considerably and now represent the dominant minimally invasive alternative on the market.¹⁹ Their success is a useful benchmark: it shows that patients and regulators will embrace needle-adjacent technology once it clears an accuracy and reliability threshold, which is precisely the bar fully noninvasive optical methods have yet to clear at scale.

From Lab Bench to Wrist: The Consumer Electronics Wild Card

No development has generated more public anticipation than Apple's long-running effort to add noninvasive glucose sensing to the Apple Watch. Reporting indicates the company has spent more than fifteen years developing a silicon photonics chip using optical absorption spectroscopy, shining laser light beneath the skin to estimate interstitial glucose concentration.²⁰ As of mid-2026, the project has reportedly reached functional prototype status but remains too large to fit inside a watch case, and leadership of the effort has recently changed hands inside the company, a sign of continued investment but not of imminent release.²⁰ Whatever Apple ultimately ships, or does not, the sheer scale of consumer attention on the idea underscores how much demand exists for a mainstream, needle-free option, and how much commercial pressure now sits behind the underlying spectroscopic research.

Summary and Conclusions

Five years of accelerating publication activity have not yet produced an FDA-cleared, fully noninvasive glucose monitor, but they have narrowed the gap considerably. Near-infrared spectroscopy remains the most mature and miniaturized approach, constrained chiefly by calibration robustness across diverse populations.³⁻⁷ Raman spectroscopy offers superior chemical specificity and has moved from bulky benchtop demonstrations toward compact, point-of-care architectures.⁸⁻¹¹ Mid-infrared and photoacoustic methods deliver the sharpest glucose-specific signatures at the cost of more demanding instrumentation, though recent single-wavelength and dual-laser designs are simplifying that trade-off.¹²⁻¹⁴ Across every modality, the largest recent accuracy gains have come not from new light sources but from pairing improved hardware with machine-learning-based signal extraction, a trend exemplified by FT-IR systems now claiming accuracy above 98%.¹⁵,¹⁶ Clinical validation studies are beginning to report performance comparable to early minimally invasive glucometers, a genuinely new development rather than an incremental one.¹⁷,¹⁸ Comprehensive comparative reviews of the broader noninvasive sensing landscape reinforce that optical methods, whatever their individual promise, must still clear reproducibility and miniaturization hurdles that have stalled prior generations of devices.²¹

Future Outlook

The next five years will likely be decided less by spectroscopic novelty than by systems engineering: shrinking lasers and detectors into wearable form factors, standardizing calibration across skin types and hydration states, and generating the large, diverse clinical datasets regulators require. Consumer electronics investment, most visibly Apple's, provides financial and engineering momentum the academic field alone has rarely enjoyed, and could accelerate miniaturization even if Apple's own product timeline continues to slip.²⁰ Meanwhile, hybrid approaches that combine multiple optical modalities, or pair spectroscopy with electrochemical or microwave sensing, may prove more clinically robust than any single method pursued in isolation.¹⁸ If the past five years are any indication, the finger-prick's replacement will not arrive as a single breakthrough but as the slow convergence of better lasers, better detectors, and increasingly capable algorithms, each closing a little more of the gap that has defined this field for thirty years.

References

1. Workman, J., Jr. Noninvasive Glucose Monitoring Using Spectroscopic Methods. Spectroscopy 2025, September 23. https://www.spectroscopyonline.com/view/noninvasive-glucose-monitoring-using-spectroscopic-methods (accessed Aug 3, 2026).

2. Delbeck, S.; Heise, H. M. Evaluation of Opportunities and Limitations of Mid-Infrared Skin Spectroscopy for Noninvasive Blood Glucose Monitoring. J. Diabetes Sci. Technol. 2021, 15 (1), 19–27. DOI: https://doi.org/10.1177/1932296820936224.

3. Chen, Q.; Zhang, W.; Wu, W. VIS/NIR/FIR Spectroscopy for Blood Glucose Monitoring: A Review. AIP Adv. 2025, 15 (3), No. 030703. DOI: https://doi.org/10.1063/5.0239187.

4. Workman, J., Jr. New Mini-NIR Device Shows Promise for Non-Invasive Blood Glucose Monitoring. Spectroscopy 2025, February 4. https://www.spectroscopyonline.com/view/new-mini-nir-device-shows-promise-for-non-invasive-blood-glucose-monitoring (accessed Aug 3, 2026).

5. Workman, J., Jr. New Infrared Device Measures Blood Sugar Without a Prick. Spectroscopy 2025, September 11. https://www.spectroscopyonline.com/view/new-infrared-device-measures-blood-sugar-without-a-prick (accessed Aug 3, 2026).

6. Workman, J., Jr. Molar Absorptivity Model Powers Near-Infrared Glucose Testing. Spectroscopy 2025, September 10. https://www.spectroscopyonline.com/view/molar-absorptivity-model-powers-near-infrared-glucose-testing (accessed Aug 3, 2026).

7. Heise, H. M.; Delbeck, S.; Marbach, R. Noninvasive Monitoring of Glucose Using Near-Infrared Reflection Spectroscopy of Skin—Constraints and Effective Novel Strategy in Multivariate Calibration. Biosensors 2021, 11 (3), No. 64. DOI: https://doi.org/10.3390/bios11030064.

8. Workman, J., Jr. Glucose Monitoring for Diabetes Using Non-Invasive Raman Spectroscopy. Spectroscopy 2025, February 5. https://www.spectroscopyonline.com/view/glucose-monitoring-for-diabetes-using-non-invasive-raman-spectroscopy (accessed Aug 3, 2026).

9. Kang, J. W.; Park, Y. S.; Chang, H.; Lee, W.; Singh, S. P.; Choi, W.; Galindo, L. H.; Dasari, R. R.; Nam, S. H.; Park, J.; So, P. T. C. Direct Observation of Glucose Fingerprint Using In Vivo Raman Spectroscopy. Sci. Adv. 2020, 6 (4), No. eaay5206. DOI: https://doi.org/10.1126/sciadv.aay5206.

10. Chu, J.; Liu, N.; Liu, J.; Xu, J.; Wang, S. Noninvasive Monitoring of Blood Glucose With In Vivo Raman Spectroscopy. J. Biophotonics 2026, 19 (1), No. e202500295. DOI: https://doi.org/10.1002/jbio.202500295.

11. Bresci, A.; Kim, Y.; Jue, M.; So, P. T. C.; Kang, J. W. Band-Pass Raman Spectroscopy Unlocks Compact Point-of-Care Noninvasive Continuous Glucose Monitoring. Anal. Chem. 2025, 97 (49), 27020–27026. DOI: https://doi.org/10.1021/acs.analchem.5c01146.

12. Pleitez, M.; Lieblein, T.; Bauer, A.; Hertzberg, O.; von Lilienfeld-Toal, H.; Mäntele, W. In Vivo Noninvasive Monitoring of Glucose Concentration in Human Epidermis by Mid-Infrared Pulsed Photoacoustic Spectroscopy. Anal. Chem. 2013, 85 (2), 1013–1020. DOI: https://doi.org/10.1021/ac302841f.

13. Aloraynan, A.; Rassel, S.; Kaysir, M. R.; Ban, D. Dual Quantum Cascade Lasers for Noninvasive Glucose Detection Using Photoacoustic Spectroscopy. Sci. Rep. 2023, 13, No. 7927. DOI: https://doi.org/10.1038/s41598-023-34912-3.

14. Aloraynan, A.; Chu, E.; Wang, J.; Alsaedi, D.; Ban, D. A Single-Wavelength Near-Infrared Photoacoustic Spectroscopy for Noninvasive Glucose Detection Using Machine Learning. Bioengineering 2026, 13 (4), No. 444. DOI: https://doi.org/10.3390/bioengineering13040444.

15. Workman, J., Jr. Blood-Glucose Testing: AI and FT-IR Claim Improved Accuracy to 98.8%. Spectroscopy 2025, February 3. https://www.spectroscopyonline.com/view/blood-glucose-testing-ai-and-ft-ir-claim-improved-accuracy-to-98-8- (accessed Aug 3, 2026).

16. Song, L.; Han, Z.; Shum, P.-W.; Lau, W.-M. Enhancing the Accuracy of Blood-Glucose Tests by Upgrading FTIR with Multiple-Reflections, Quantum Cascade Laser, Two-Dimensional Correlation Spectroscopy and Machine Learning. Spectrochim. Acta A 2025, 327, No. 125400. DOI: https://doi.org/10.1016/j.saa.2024.125400.

17. Kaluza, M.; Janik, S.; Lubinski, T.; Saita, M.; Lachmann, P.; Canini, L.; Lepro, V.; Mäntele, W. Clinical Validation of Noninvasive Blood Glucose Measurements by Midinfrared Spectroscopy. Commun. Med. 2025, 5, No. 509. DOI: https://doi.org/10.1038/s43856-025-01241-7.

18. Shi, J.; Fernández-García, R.; Gil, I. Sensor Technologies for Non-Invasive Blood Glucose Monitoring. Sensors 2025, 25 (12), No. 3591. DOI: https://doi.org/10.3390/s25123591.

19. Saha, T.; Del Caño, R.; Mahato, K.; De la Paz, E.; Chen, C.; Ding, S.; Yin, L.; Wang, J. Wearable Electrochemical Glucose Sensors in Diabetes Management: A Comprehensive Review. Chem. Rev. 2023, 123 (12), 7854–7889. DOI: https://doi.org/10.1021/acs.chemrev.3c00078.

20. Rossignol, J. Apple Watch for Diabetes: The Latest on Apple's Plans for Non-Invasive Blood Sugar Monitoring. MacRumors, May 26, 2026. https://www.macrumors.com/2026/05/26/apple-watch-non-invasive-blood-sugar-plans/ (accessed Aug 3, 2026).

21. Di Filippo, D.; Sunstrum, F. N.; Khan, J. U.; Welsh, A. W. Non-Invasive Glucose Sensing Technologies and Products: A Comprehensive Review for Researchers and Clinicians. Sensors 2023, 23 (22), No. 9130. DOI: https://doi.org/10.3390/s23229130.