4D hyperspectral imaging for intraoperative tissue classification

This study examines the use of 4D hyperspectral imaging (4D HSI) for visualizing clinically relevant targets. The approach relies on intrinsic optical properties of the tissue without the need for exogenous dyes or contrast agents. We compare autofluorescence-based 4D HSI to a traditional 3D HSI approach for its ability to distinguish between nerve and ligament bundles, isolated from adult cow trotters. Datacubes are captured within the 420–720 nm interval, under illumination wavelengths ranging from 300 to 400 nm. The unmixing outcomes of individual 3D HSI cubes prove inferior to those from multiple, even as few as three, 3D cubes sequentially combined into one. The data show an increase in the ratio of correctly identified pixels with the increased number of HSI cubes taken at different excitation wavelengths. Combining multiple 3D HSI cubes into a single 4D dataset improves target identification and reveals additional features, such as damaged nerve with torn epineurium. The 4D HSI approach merges excitation–emission matrices with traditional HSI and delivers higher accuracy in distinguishing different tissue structures. Furthermore, more advanced analysis techniques, such as deep neural networks, can potentially utilize 4D HSI data further to distinguish surgical targets. The data suggest that 4D HSI has significant potential for various clinical applications, including surgical procedures, dental applications, and cancerous lesion identification, leveraging small spectral changes for improved visualization and clinical outcomes.