Toward a Robust Non-Invasive Tool Assessing Endothelial Function Using a Thermal Camera

Antoine Collet1, Doina Gangur2, François Thierry M Marelli3, Stephane Carlier4, Thierry Dutoit1
1University of Mons, 2Universite Mons, 3UMONS, 4University of Mons (UMons)


Abstract

Endothelial dysfunction is an early marker of cardiovascular disease and plays a central role in atherosclerosis physiopathology. Flow-mediated dilation is the reference non-invasive method, but it is operator-dependent and limited in accessibility. Alternative approaches, such as thermography, have been proposed to assess microvascular responses during reactive hyperemia, but remain indirect and lack standardization, with limited data on reproducibility and optimal occlusion protocols.

This study aims to improve the reproducibility of thermal video-based parameters of reactive hyperemia and to evaluate a novel occlusion protocol.

This study includes 18 healthy subjects to date. Participants fast ≥ 3 hours before measurements. Two successive brachial cuff occlusions were applied in a seated position (200 mmHg, 1 min and 3 min), separated by an 8-minute recovery period. Vascular response is assessed by echography and Doppler ultrasound of the radial artery as a reference measure, with simultaneous thermal video recording of the hands. Thermal signals are processed using a Python-based pipeline to extract thermal parameters over time, including temperature rebound and area under the temperature curve.

Preliminary analysis on 6 subjects (6M, 0F) shows intra-subject reproducibility with coefficients of variation of 15% for temperature rebound. The most reproducible metrics were observed in the palm region and for the 3-minute occlusion, particularly when analyzing absolute temperature changes rather than relative variations. A subset of 12 subjects (5M, 7F) was identified as "cold-flagged", characterized by low baseline fingertip temperature (<27°C), and showed altered thermal response despite controlled room conditions (22-26°C). Ongoing analyses aim to determine whether robust and reproducible parameters can still be extracted.

Future work will include patients with increased risk of endothelial dysfunction, notably women with breast cancer receiving anthracycline therapy. The resulting dataset will be made publicly available and may represent, to our knowledge, the first of its kind in this field.