Dysplasia in Colonic Polyps: Discrimination through Digital Image Analysis
David R Martin, Alton B Farris III. Emory University School of Medicine, Atlanta, GA
Background: Determining the level of dysplasia within colon polyps is often difficult. This distinction is clinically important, as adenomas with high-grade dysplasia (HGD) are considered advanced lesions clinically and may warrant more stringent screening. Given apparent limitations in correlation between pathologists for the determination of low-grade vs high-grade dysplasia, we investigated polyp nuclear size using digital image analysis.
Design: Archived cases representing normal colonic mucosa (n=11), hyperplastic polyps (HPs) (n=13), conventional tubular adenomas (TAs) (n=12), and adenomas with HGD (n=12) were digitally scanned with a whole slide scanner. Representative nuclei from areas exhibiting the diagnostic pathologic change for each case were then circumscribed at 40x-magnification. Parameters for the discriminate analysis between groups included: nuclear circumference (um), area (um2), and 13 different nuclear staining measurements from a positive pixel count (PPC) algorithm.
Results: Average nuclear area and circumference progressed in the following order: normal colonic mucosa, HPs, TAs, and adenomas with HGD (Figure 1). Differences between groups were found to be statistically significant using a Student's t test (p<.0001 between all groups except HPs vs controls). Pleomorphism was also substantially greater in adenomas with HGD than tubular adenomas (standard deviation of 54.9 um2 vs 15.2 um2). These findings were largely mirrored using the PPC algorithm, with statistical significance demonstrated between groups when evaluated for: number of strongly staining pixels, total staining pixels, and staining intensity. Hierarchal clustering showed that polyp types clustered into distinct groups based on their nuclear size and PPC algorithm measurements (Figure 1).
Conclusions: Our results confirm that quantitative image analysis is an effective measure that reflects dysplastic change in colon polyps. Nuclear area, circumference, and findings seen with the PPC algorithm distinguished polyp types from one another with statistical significance. Ultimately, precise quantification of cytologic nuclear size and staining intensity measurements may prove to be a useful diagnostic tool in determining the degree of dysplasia in colon polyps with equivocal or controversial histology.
Tuesday, March 5, 2013 1:00 PM
Poster Session IV # 134, Tuesday Afternoon