Abstract
The natural fluorescence (autofluorescence) of tissues has been noted as a biomarker for cancer for several decades. Autofluorescence contrast between tumors and healthy tissues has been of significant interest in endoscopy, leading to development of autofluorescence endoscopes capable of visualizing 2-3 fluorescence emission wavelengths to achieve maximal contrast. However, tumor detection with autofluorescence endoscopes is hindered by low fluorescence signal and limited quantitative information, resulting in prolonged endoscopic procedures, prohibitive acquisition times, and reduced specificity of detection. Our lab has designed a novel excitation-scanning hyperspectral imaging system with high fluorescence signal detection, low acquisition time, and enhanced spectral discrimination. In this study, we surveyed a comprehensive set of excised tissues to assess the feasibility of detecting tissue-specific pathologies using excitation-scanning. Fresh, untreated tissue specimens were imaged from 360 to 550 nm on an inverted fluorescence microscope equipped with a set of thin-film tunable filters (Semrock, A Unit of IDEX). Images were subdivided into training and test sets. Automated endmember extraction (ENVI 5.1, Exelis) with PCA identified endmembers within training images of autofluorescence. A spectral library was created from 9 endmembers. The library was used for identification of endmembers in test images. Our results suggest (1) spectral differentiation of multiple tissue types is possible using excitation scanning; (2) shared spectra between tissue types; and (3) the ability to identify unique morphological features in disparate tissues from shared autofluorescent components. Future work will focus on isolating specific molecular signatures present in tissue spectra, and elucidating the contribution of these signatures in pathologies.