Classification of Normal and Lesional Colon Tissue Using Fluorescence Excitation-Scanning Hyperspectral Imaging as A Method for Early Diagnosis of Colon Cancer

Malvika Lall, Joshua Deal, Shante Hill, Paul Rider, Carole Boudreaux, Thomas Rich, Silas Leavesley


Colon cancer is the second leading cause of cancer death in the United States. The purpose of colorectal screening exams is the detection and diagnosis of lesions to help reduce the associated morbidity and mortality by identifying lesions early and accurately prior to advancement into cancer or tissue invasion. The objective of this study is to classify colon tissue into normal and lesional tissue by measuring spectral changes using Fluorescence Excitation-Scanning Hyperspectral Imaging. Fresh normal and colon cancer surgically resected tissue specimens were obtained in collaboration with the University of South Alabama Departments of Surgery and Pathology and prepared for scanning.  All procedures were carried out in accordance with Institutional Review Board protocol # 13-120. Normal and cancerous tissue types were confirmed by histologic evaluation of H&E permanent sections and scanned by excitation scanning hyperspectral imaging using a novel microscope constructed at the University of South Alabama. At least three fields of view (FOV) were located on each specimen. MATLAB and ENVI were used for spectral correction and to extract the spectra from each region of interest. Spectral graphs for each region of interest for both lesional and normal tissue were generated and compared. Results comparing average spectra obtained from normal colon tissue showed homogeneity as demonstrated by spectral images with similar peak wavelengths and shapes in all regions of interest within a FOV. However, in comparison the extracted spectra from colon cancer showed high heterogeneity with spectral images of varying peak wavelengths and shapes. Results from an automated classification using a Maximum Likelihood algorithm yielded high sensitivity (97.91) and high specificity (92.75) values, showing that there was a high probability that a cancerous pixel was correctly classified as cancerous and a high probability that a normal pixel was correctly classified as normal respectively. A high overall accuracy (94.99%) showed that most pixels were correctly classified. We conclude that fluorescence excitation-scanning hyperspectral imaging could detect differences in the spectral patterns of normal and lesional colon tissue which would allow for the classification of colon tissue into normal and lesional tissue types. This information could be used in the development of new methods for early diagnosis of colon cancer by translating the technology into a real-time endoscopic platform to classify tissues based on the spectral changes.


Colon cancer; HIFEX; Colorectal screening; Spectral changes

Full Text: PDF


  • There are currently no refbacks.