Abstract
This chapter addresses object measurements from 2D microscopy images. Object measurements (called image features) vary in terms of theoretical formulas for the same image feature, the physical units used to represent pixel-based measurements, the definitions of objects in images (called regions of interest or ROIs), algorithmic implementations, the number of exposed parameters to a user, and programming languages. Our motivation is to introduce readers to image-based object measurements and to quantify numerical variability of (a) image features and (b) feature-based classification outcomes. The variability is evaluated across widely-used image feature extraction libraries to highlight the sources of variability when deriving image-based scientific conclusions. By characterizing these feature variations across Python scikit-image, CellProfiler, MaZda, ImageJ, and in-house Java libraries, we concluded 15.6% of 32 intensity features, 47.9% of 71 shape features, and 88.2% of 68 textural features differ in values. While intensity feature variations had no impact on classification outcomes, shape and textural feature variations had a negative impact on classification outcomes in 52.9% and 97.1% of all single feature based classifications respectively. All numerical results are available on-line at
https://isg.nist.gov/deepzoomweb/resources/featureVariability/index.html via interactive interfaces that enable traceability to image objects. A reader can perform similar studies with his/her data using a web image processing pipeline (WIPP) system developed for this study and available at
https://isg.nist.gov/deepzoomweb/software/wipp. The reported numbers suggest that any published microscopy research should come with detailed provenance information about the image feature extraction libraries to deliver reproducible scientific work.