QuPath for automated analysis of digital images of oral epithelial dysplasia.

Abstract

Objectives: Grading of oral epithelial dysplasia (OED) has been plagued with intra-observer and inter-observer variations. To overcome this subjectivity, a more objective digital image analysis is obligatory using computer-aided software. The use of open-source software like QuPath, which is a new bio-image evaluation software program, may fulfil this growing need in virtual pathology. This study used the QuPath software for automatic analysis of morphometric parameters in hematoxylin and eosin (H and E)-stained digital images of oral epithelial dysplasia.

Methodology: 150 H and E digital images of varying grades of OED captured under 40x magnification were processed using QuPath software for automatic analysis of cellular and nuclear parameters.

Results: The parameters that showed statistical significance included nuclear hematoxylin OD, nuclear eosin OD, cellular hematoxylin OD, cellular eosin OD, cytoplasm hematoxylin OD, and cytoplasmic eosin OD (P < 0.05), while none of the other parameters showed statistically significant differences. A prediction accuracy of 76%, 74%, and 70% for mild, moderate, and severe dysplasia was obtained, respectively.

Conclusion: The quantitative results outlined in this paper are encouraging to indicate that the use of this technique may improve the diagnostic reliability of OED. Morphometric analysis of OED using Qupath software can be fast and reproducible and can be quantitated automatically.