Assessment of autofluorescent signatures in multiple tissue types with novel excitationscanning hyperspectral imaging

Autofluorescence imaging has gained prominence as a potential technology for real‐time biopsies of cancer in the last two decades. Cancerous and healthy tissues can often be differentiated based on their peak fluorescence emission wavelengths using autofluorescence endoscopes. However, comparative studies of autofluorescence endoscopes have indicated only marginal improvements in tumor detection efficiency over white‐light or narrow‐band endoscopy. These results are due to high interoperater variability, which in part is due to limited contrast between newly‐developed neoplasms and healthy tissues. Recently, hyperspectral endoscopes have been developed to provide spectral discrimination and improve contrast between cancerous and healthy tissues, but these systems are often stymied by prolonged image acquisition times that are prohibitive for real‐time endoscopic procedures. Our laboratory has developed a novel hyperspectral approach that provides significantly shorter image acquisition times (10–100× that of current hyperspectral approaches) coupled with high spectral discriminatory power. This approach, excitation‐scanning, filters the fluorescence excitation at discrete wavelengths to provide a contiguous excitation spectrum. In this work, we assessed the presence of multiple tissue‐specific autofluorescent signatures using excitation‐scanning to image fresh, unlabeled ex vivo tissues. Samples were cut to reveal distal portions of each tissue. Excitation scanning was performed using an array of thin‐film tunable filters placed after a 300W Xe arc lamp. Hyperspectral images were acquired from 360 to 550 nm in 5 nm increments. An automated endmember extraction technique was used to mitigate interoperator characterization of unique spectral signatures in tissues. Our results suggest spectral commonalities between disparate tissue types. We identified common peak excitation wavelengths at 360 nm, 375 nm, and 450 nm. These peaks varied in intensity between each tissue type, suggesting varying concentrations of specific autofluorescent components within tissue types. Post‐processing spectral unmixing techniques revealed morphologically distinct regions in each field‐of‐view, corresponding to the heterogeneic nature of autofluorescent components within each tissue. These results suggest the presence of shared autofluorescent components. These components have unique spectral signatures that may be elucidated further to discriminate tissues based on isolated autofluorescent components rather than overall spectral trends or peak wavelengths alone. Additionally, the automated endmember extraction technique provided a method of mitigating user bias in spectral extraction. Further application of the excitation‐scanning approach is easily implementable for endoscopy, and would provide suitably fast image acquisition speeds and high spectral discriminatory for real‐time optical biopsy of cancer. Support or Funding Information This work is supported by P01HL066299 and the Abraham Mitchell Cancer Research Fund.