Do We Trust Image Measurements? Variability, Accuracy and Traceability of Image Features
Mylene H. Simon, Joe Chalfoun, Peter Bajcsy, Mary C. Brady
The paper addresses the problem of traceability and variability of image features used in a plethora of image-based scientific studies. Image features are extracted using software packages that vary in terms of programming languages, theoretical formulas for the same image feature, algorithmic implementations, input parameters, units of measurements, and definitions of image regions of interest. Our motivation is to quantify feature numerical variability across software packages, determine image accuracy with respect to reference images, and address the traceability by designing a client-server system for sharing and reproducing image feature values. We report the results for 117 image features extracted using ImageJ/Fiji, Matlab, Python, and in-house Java software packages with 43 duplicate features across the four packages. Using the normalized difference as metric, we identified 10 out of 43 features to differ over 1% in value. The sources of these numerical differences are discussed. In order to extract traceable image features, we designed a web system with (1) interfaces to load images and extract image features while utilizing distributed computational resources, (2) access to the four software packages as plugins, (3) provenance information gathering mechanism, and (4) feature values hyperlinked with all computational provenance artifacts.
Proceedings of the 2016 IEEE International Conference on Big Data