Identification of Near-Infrared Characteristic Bands of Small Intestine Necrosis Based on Least Trimmed Squares With Regularization

Abstract

Hyperspectral imaging is a promising tool for identifying ischemic necrotic small intestine. To analyze the causes of small bowel necrosis, studying characteristic bands is crucial. However, differences in samples and spectral acquisition devices limit the availability of all bands for analysis, posing challenges in selecting bands adapted to individual variations. This study proposed a method based on the least trimmed squares algorithm, enhanced with regularization, to identify characteristic bands. The method successfully differentiated normal and necrotic tissue and analyzed necrosis causes, which originated from the same rabbit, different rabbits, and different necrosis durations. It identified 763 nm as the characteristic band, corresponding to the deoxyhemoglobin absorption peak. This approach offers accurate, automated band identification while addressing sample and device discrepancies, enabling the selection of more suitable characteristic bands.