A high-speed hyperspectral imaging system and large-scale hyperspectral dataset for abdominal surgical applications

Modern foundation models have shown promise in classification tasks and may be used in medical imaging systems, but the current imaging modalities, such as fluorescence and narrow-band imaging, have drawbacks hindering the ability to generate the large datasets required of current computer vision (CV) models. Hyperspectral imaging (HSI) is a non-invasive and label-free modality and has shown promise in tissue classification and cancer detection. But, to leverage the rich information provided with HSI alongside modern CV architectures, larger datasets must be curated. As such, we have designed an HSI system and workflow to address this gap by allowing for high-throughput ex vivo tissue imaging. We utilize an optical configuration consisting of three hyperspectral cameras along with a custom in-house application for efficient imaging. The system covers a wavelength range of 460-960 nm, acquiring 30 hyperspectral images that are averaged into a single hypercube with 55 bands, a process completed in under 6 seconds. The system has been used to acquire 1835 hyperspectral images of tissues from four animal models including porcine, murine, galline, and bovine organs. The high-speed HSI system and the corresponding dataset can be further applied to many minimally invasive surgical applications including robotic-assisted laparoscopic surgery.