Online Program Home
My Program

Abstract Details

Activity Number: 516 - Case Studies of Scalar-On-Image Regression
Type: Contributed
Date/Time: Wednesday, July 31, 2019 : 10:30 AM to 12:20 PM
Sponsor: Section on Statistics in Imaging
Abstract #306989 Presentation
Title: Reproducible Image Processing by Journaling
Author(s): Paul Thompson* and Norman Matloff
Companies: Thompson Biostatistical Solutions and University of California at Davis
Keywords: Reproducible research; Image fraud; Image manipulation; Journaling
Abstract:

In many areas of science, images are data. Information in images is used to show the existance of structures, to show the impact of treatments, and to show the effect of treatments. Images must be manipulated for scientific presentation. In image manipulation for scientific manuscripts, interactive programs (Photoshop, GIMP) are used to edit images. These manipulations are not reproducible. Changes in colors (which are legal), cropping, or other changes cannot be tracked, and are susceptible to fraudulent manipulations. Some studies show a high proportion of images in published documents to have evidence of fraudulent or inappropriate manipulations. A project, funded by Office of Research Integrity, is working to solve this problem by tracking and documenting all interactive events in an editing session. A "journal" or tracking log is created which indicates what editing operations were performed, and further can be executed to reproduce the edits. This introduces transparency into image editing, which will act to reduce fraud.


Authors who are presenting talks have a * after their name.

Back to the full JSM 2019 program