Near-Infrared Spectroscopy
Meets Deep Intelligence

The science of light-tissue interaction, reimagined through AI — delivering biomarker measurement at the molecular level without a single needle.

How NIR Spectroscopy Works

Near-infrared light in the 700–2500 nm wavelength range penetrates skin and subcutaneous tissue without ionizing damage. Different biological molecules — glucose, hemoglobin, proteins, electrolytes — absorb specific NIR wavelengths at unique intensities determined by their molecular structure.

iGlutek's 8-channel photodetector array captures the full diffuse reflectance spectrum in milliseconds. Our proprietary algorithm stack then deconvolves overlapping absorption signals to isolate individual analyte concentrations with clinical-grade precision.

700–2500nm Range
8-Channel Array
Diffuse Reflectance
NIR LED Epidermis Dermis + Blood G H O N 8ch Detector Absorption Spectrum → Analyte Concentration

A Five-Layer
Signal-to-Insight Pipeline

L1

Raw Spectral Acquisition

8-channel photodetector captures full NIR diffuse reflectance at 50Hz. Hardware-level noise filtering via lock-in amplification removes ambient light interference.

50 Hz
L2

Pre-Processing & Normalization

Savitzky-Golay smoothing, multiplicative scatter correction (MSC), and standard normal variate (SNV) transformation eliminate motion artifacts and skin-tone variation.

MSC + SNV
L3

Feature Extraction (1D-CNN)

A 12-layer 1D convolutional neural network extracts spectral features across wavelength neighborhoods, learning absorption fingerprints specific to each analyte class.

12-Layer CNN
L4

Personalization Layer (LSTM)

A recurrent LSTM module maintains a rolling 30-day physiological baseline per user, compensating for circadian rhythms, hydration state, and individual absorption coefficients.

LSTM + Baseline
L5

Output & Uncertainty Quantification

Bayesian confidence intervals accompany every measurement. Predictions outside the personalized confidence band trigger clinical-grade alerts with physician-ready event logs.

Bayesian CI

How iGlutek Compares

Benchmarked against current clinical standards and consumer wearables in controlled trials.

Parameter iGlutek CGM (Dexcom G7) Fingerstick (Accu-Chek) Smartwatch (Apple Watch)
Needles Required✓ None✗ Subcutaneous✗ Fingerstick✓ None
Glucose Monitoring✓ Yes✓ Yes✓ Yes✗ No
Consumables✓ Zero✗ Sensor €90/mo✗ Strips €80/mo✓ None
Biomarkers Monitored11+1 (Glucose)1 (Glucose)4 (SpO₂, HR, Temp, ECG)
Predictive AI✓ 48h aheadBasic trend✗ NoneLimited
Clinical Accuracy (MARD)<9% target8.2%6.8%N/A (glucose)
Monthly Cost✓ €0 after purchase€90–120€60–100N/A

* iGlutek accuracy figures represent development targets pending clinical trial completion. Independent validation in progress.

A Deep Moat of
Defensible Innovation

3 Patent Applications

Filed on NIR sensor array geometry, multi-analyte spectral deconvolution, and personalized calibration LSTM architecture.

Proprietary Dataset

2M+ paired NIR spectra and lab-confirmed biomarker measurements — the largest non-invasive spectral training corpus in the field.

Edge AI Runtime

Compressed model runs entirely on the device's embedded MCU — full inference in <80ms without cloud dependency.