Format

Send to

Choose Destination
J Med Imaging (Bellingham). 2016 Apr;3(2):024502. doi: 10.1117/1.JMI.3.2.024502. Epub 2016 Jun 3.

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

Author information

1
University of Arkansas , Department of Biomedical Engineering, 1 University Boulevard, Fayetteville, Arkansas 72701, United States.
2
University of Arkansas for Medical Sciences , Department of Pathology, 4301 West Markham Street, Little Rock, Arkansas 72205, United States.
3
University of Arkansas for Medical Sciences , Department of Gastrointestinal Surgery, 4301 West Markham Street, Little Rock, Arkansas 72205, United States.

Abstract

Qualitative screening for colorectal polyps via fiber bundle microendoscopy imaging has shown promising results, with studies reporting high rates of sensitivity and specificity, as well as low interobserver variability with trained clinicians. A quantitative image quality control and image feature extraction algorithm (QFEA) was designed to lessen the burden of training and provide objective data for improved clinical efficacy of this method. After a quantitative image quality control step, QFEA extracts field-of-view area, crypt area, crypt circularity, and crypt number per image. To develop and validate this QFEA, a training set of microendoscopy images was collected from freshly resected porcine colon epithelium. The algorithm was then further validated on ex vivo image data collected from eight human subjects, selected from clinically normal appearing regions distant from grossly visible tumor in surgically resected colorectal tissue. QFEA has proven flexible in application to both mosaics and individual images, and its automated crypt detection sensitivity ranges from 71 to 94% despite intensity and contrast variation within the field of view. It also demonstrates the ability to detect and quantify differences in grossly normal regions among different subjects, suggesting the potential efficacy of this approach in detecting occult regions of dysplasia.

KEYWORDS:

computer-aided diagnosis; gastrointestinal dysplasia; image analysis; image quality control; microendoscopy

Supplemental Content

Full text links

Icon for PubMed Central
Loading ...
Support Center