Quantitative analysis of ex vivo colorectal epithelium using an automated feature extraction algorithm for...

Area of Science:

• Gastroenterology
• Medical Imaging
• Computational Pathology

Background:

• Fiber bundle microendoscopy shows promise for qualitative colorectal polyp screening.
• High sensitivity and specificity are reported, but require trained clinicians, leading to interobserver variability.

Purpose of the Study:

• To develop and validate a quantitative image quality control and image feature extraction algorithm (QFEA).
• To reduce training burden and provide objective data for improved clinical efficacy of microendoscopy.

Main Methods:

• QFEA was developed using microendoscopy images of porcine colon epithelium.
• The algorithm underwent validation on ex vivo human colorectal tissue from clinically normal regions.
• QFEA performs quantitative image quality control and extracts features like crypt area and circularity.

Main Results:

• Automated crypt detection sensitivity ranged from 71% to 94%, robust to image variations.
• QFEA demonstrated flexibility with both mosaic and individual images.
• The algorithm detected and quantified differences in grossly normal regions between subjects.

Conclusions:

• QFEA offers a quantitative, objective approach to microendoscopy image analysis.
• This algorithm has potential for detecting occult dysplasia in colorectal tissue.
• QFEA may enhance the clinical efficacy and accessibility of microendoscopy for polyp screening.