Hyperspectral imaging for evaluating the degree of mesenteric ischemia

Abstract

Hyperspectral imaging has emerged as a global trend in addressing various applied medical challenges, offering non-invasive solutions and significant potential for mathematical processing. The high spectral resolution of modern hyperspectral imaging systems allows for the development of accurate classification models based on classical segmentation methods and advanced neural network approaches.
In this paper, we propose a hyperspectral imaging method that incorporates a developed laboratory system and mathematical algorithms for processing hyperspectral data using machine learning techniques. To achieve this, we created a hyperspectral imaging system featuring a broadband radiation source with a fiber-ring illuminator and a hyperspectral camera covering a spectral range of 400 to 1000 nm. We studied the intestinal tissue of 9 animals (Wistar rats) after applying ligatures to the arcade vessels to simulate the ischemia durations of 0, 1, 6, and 12 hours. Taking advantage of the different light absorption characteristics of unbound and oxygen-bound forms of hemoglobin, we calculated the tissue saturation parameter using a two-wave approach in the near-infrared region. A machine learning method was used to evaluate the extent of ischemic damage, resulting in a cumulative accuracy metric of AUC = 0.99 ± 0.01 for the XGBoost model.
The results indicate that this method exhibits high sensitivity and effectiveness in diagnosing early-stage mesenteric ischemia and is capable of detecting reversible changes caused by blood flow disturbances. This study underscores the importance of hyperspectral imaging in medicine and lays the foundation for further research in this field.

Speaker

Ilya Goryunov
Orel State University named after I.S. Turgenev
Orel

Discussion

Ask question