Hyperspectral VNIR and NIR Sensors for the Analysis of Human Normal Brain and Tumor Tissue

Abstract

Hyperspectral (HS) imaging (HSI) is arising as a novel imaging technique to delineate brain tumor tissue in surgical-time. The accurate identification of the boundaries between tumor and normal tissue determines prolonged survival. In this work, a preliminary spectral analysis of different brain tissue was performed to identify features between different classes. A HSI dataset of in-vivo brain tissue was acquired by two HS cameras, covering the VNIR (Visual and Near-Infrared) [400–1000 nm] and NIR (Near-Infrared) [900–1700 nm] spectral ranges. Both HS images were registered using feature-based techniques with different geometric transformations to perform a spectral analysis of a certain pixel. Reflectance and absorbance spectral signatures were analyzed identifying spectral absorbance peaks related with hemoglobin and water. Finally, a statistical analysis was performed, where normal tissue (NT), tumor tissue (TT), and hypervascularized tissue (HT) were compared, obtaining highly statistically significance between HT-NT and HT-TT in both VNIR and NIR spectral ranges analyzed, and some no statistically significant differences between TT and NT in certain spectral ranges.