Hyperspectral imaging of squamous cell carcinoma of the head and neck – Clinical implementation and initial results of an ex vivo application

 

Introduction The aim of curative surgical therapy for head and neck cancer is complete resection with sufficient safety margin. Differentiation between tumor and healthy tissue is performed via white light endoscopy and microscopy. Hyperspectral imaging (HSI) is a non-invasive imaging technique that analyzes tissue after white light exposure with respect to the wavelengths of reflected light for tissue differentiation.

Methods The tumor specimen was examined in 13 patients immediately after resection using RGB and HSI images with a resolution of 1,600 x 1,600 pixels. Data acquisition was performed using an imec HSI-Snapscan camera from IMEC in combination with an OPMI PENTERO 900 microscope (Zeiss). In the RGB image, image regions were visually annotated as "tumor center", "tumor margin" and "healthy mucosa" as well as "musculature".

Results The preparation of an HSI image took on average 10 minutes. In the case of large-area specimens, the division into several individual images seems reasonable. A total of 24 HSI images of the 10 preparations with 109 spectral channels each were available for analysis. The annotations performed manually in the RGB images were successfully registered to the HSI images and analyzed consecutively using artificial intelligence (AI).

Conclusion HSI imaging is a seminal technique to increase the precision in tissue differentiation. It can be reliably applied in an ex vivo setting. The results of AI analysis are promising. However, the accuracy of AI depends on the amount of data available. Transition to in-vivo use is the goal.

ZEISS Group

Publication History

Article published online:
12 May 2023

Georg Thieme Verlag
Rüdigerstraße 14, 70469 Stuttgart, Germany