NIRS in Peripheral Artery Disease

Differentiating upper and lower limbs and correlation with the ankle-brachial index

Pilot study evaluating near-infrared spectroscopy as a non-invasive complementary tool for PAD assessment
CardioNIR
NIRS
PAD
Machine Learning
Published

August 21, 2025

Published in Microvascular Research 162, 104864 (2025) — doi:10.1016/j.mvr.2025.104864 — Open Access (CC BY 4.0)

At a Glance

0.91
AUC-ROC for limb
differentiation (95% CI 0.88–0.94)
51
PAD patients
102 lower limbs evaluated
1429 nm
Start of optimal wavelength
for ABI prediction
0.167
R² NIRS–ABI correlation
(RMSECV = 0.186)

Background

Peripheral artery disease (PAD) affects over 230 million adults worldwide, yet remains frequently underdiagnosed. The ankle-brachial index (ABI) is the standard first-line test, but has known limitations — particularly in patients with arterial calcification, diabetes, or chronic renal disease, where falsely elevated values can mask compromised flow.

Near-infrared spectroscopy (NIRS) offers a non-invasive, portable alternative that penetrates superficial tissues to assess hemoglobin oxygenation and tissue composition. Despite its established role in cerebral oximetry, its application to PAD remains underexplored. This pilot study investigated whether NIRS spectral signatures can differentiate ischemic from non-ischemic limbs and whether they carry information correlated with ABI — all using a portable spectrometer at rest in a routine outpatient setting.

Key Findings

🔴 Strong Limb Discrimination

Random Forest classification achieved an AUC of 0.91, with the most discriminative wavelengths below 1060 nm and in the 1250–1375 nm range — consistent with oxygenated hemoglobin absorption peaks.

🟢 ABI Spectral Correlation

iPLS identified a narrow band at 1429–1463 nm as the most informative for ABI prediction. This region captures O–H stretching (water), N–H stretching (proteins), and CH₂ combination bands (lipids) — pointing to vascular tissue composition signatures.

📡 Portable & At-Rest Protocol

Unlike previous NIRS–PAD studies requiring exercise testing, we demonstrated feasibility at rest in an outpatient clinic, using a compact InGaAs-based spectrometer (900–1700 nm) with no cooling requirements.

🔗 Complementary to ABI

NIRS may address ABI limitations in patients with non-compressible vessels. The modest R² was expected — the goal was not to replicate ABI but to identify whether NIRS carries independent hemodynamic information.

Spectral Windows

The most discriminative wavelengths for limb classification and ABI prediction fall in distinct regions of the NIR spectrum, reflecting different physiological information:

900 nm 1100 1300 1500 1700 nm
Classification: OxyHb absorption (900–1060 nm) ABI prediction: O–H, N–H, CH₂ bands (1429–1463 nm)

The 900–1060 nm window aligns with the absorption characteristics of oxygenated hemoglobin (typically peaking between 800–1000 nm), while the 1429–1463 nm region captures molecular signatures from water (O–H), proteins (N–H), and lipids (CH₂) — potentially reflecting vascular tissue composition and perfusion status rather than oxygenation alone.

Methods

Step 01
NIRS Acquisition

20 scans per site (right thumb, both halluces), 900–1700 nm, 225 data points. Supine, at rest.

Step 02
Pre-processing

Standard Normal Variate (SNV) transformation for scatter correction and baseline normalisation.

Step 03
Classification

Random Forest (upper vs. lower limb), 2/3–1/3 split preserving within-patient measurements.

Step 04
Regression

Interval PLS (iPLS) for wavelength selection and ABI prediction across 21 spectral intervals.

Study Design

This was an observational, cross-sectional study employing a convenience sample of 51 outpatients from the Angiology and Vascular Surgery service at Unidade de Saúde Local de São João, Porto. Patients with PAD (Rutherford 0–3, ABI ≤ 0.90 or abnormal Doppler) were recruited between November 2023 and May 2024. The study received approval from the Ethics Committee (ref. CE-382-24).

Cohort Summary

Characteristic Value
N 51 patients (102 lower limbs)
Age 65.5 ± 8.3 years
Sex 86.3% male
Rutherford 1–3 94.1%
Median ABI 0.65 (IQR 0.4)
Smoking 94.1% (37.2% current)
Hypertension 72.5%
Diabetes 27.5%
Coronary artery disease 23.5%
Claudication (limb-level) 66%
No palpable distal pulse 84.6%
Monophasic / occlusion (Doppler) 64.1%

NIRS Device

Spectroscopic measurements used a portable NIR reflectance system (NIR-S-G1) with an InGaAs detector covering 900–1700 nm. The device requires no cooling, holds FCC and CE certifications, and allows rapid spectral acquisition — making it suitable for bedside and outpatient use.

Limitations & Future Directions

This pilot study has important limitations: a relatively small convenience sample, a predominantly male cohort limited to Rutherford stages 1–3, and the absence of patients with non-compressible vessels (ABI > 1.3) — a subgroup where NIRS could have particular clinical value. The assumption that upper limbs were non-ischemic was based on clinical evaluation without formal hemodynamic confirmation.

Future work should prioritise larger and more diverse cohorts, matched healthy controls, integration of exercise or provocative protocols (e.g., transient leg elevation), and comparison with complementary modalities such as TcPO₂. Standardised guidelines for NIRS sensor placement and acquisition parameters are also essential for cross-study comparability.

Team

Jéssica Braga Amorim *†
RISE-Health, Dept. Surgery & Physiology, FMUP
Conceptualization · Investigation · Writing (original draft) · Visualization · Validation · Data curation · Methodology
Marina Dias Neto †
RISE-Health, FMUP · Angiology & Vascular Surgery, USL São João
Conceptualization · Investigation · Supervision · Writing (review & editing) · Visualization · Validation · Data curation · Methodology
Sandra Magalhães
RISE-Health, Dept. Surgery & Physiology, FMUP
Investigation · Methodology · Data curation
António S. Barros
RISE-Health, Dept. Surgery & Physiology, FMUP
Conceptualization · Formal analysis · Software · Funding acquisition · Project administration · Supervision · Resources · Writing (review & editing) · Visualization · Validation · Investigation · Methodology · Data curation

* Corresponding author   † Equal contribution

Citation

Braga Amorim J, Dias Neto M, Magalhães S, Barros AS. Pilot study on near-infrared spectroscopy in peripheral artery disease: Differentiating upper and lower limbs and its correlation with the ankle-brachial index. Microvasc Res. 2025;162:104864. doi: 10.1016/j.mvr.2025.104864

BibTeX

@article{BragaAmorim2025,
  title   = {Pilot study on near-infrared spectroscopy in peripheral
             artery disease: {Differentiating} upper and lower limbs
             and its correlation with the ankle-brachial index},
  author  = {Braga Amorim, J{\'e}ssica and Dias Neto, Marina and
             Magalh{\~a}es, Sandra and Barros, Ant{\'o}nio S.},
  journal = {Microvascular Research},
  volume  = {162},
  pages   = {104864},
  year    = {2025},
  doi     = {10.1016/j.mvr.2025.104864}
}

Funding

This research was supported by Fundação para a Ciência e Tecnologia (FCT) under the CardioNIR project — CARDIOvascular Near-InfraRed spectroscopy probing — 2021 (PTDC/EMD-EMD/3822/2021).