Can You Trust What You Read?
Part 2: What Is an Image Worth?

Audra E. Cox, PhD, ELS (Managing Editor, AJP and JMD)

Guts & Glory, by Carole Nicholas (quilt, based on cross-section of a single villus; part of Art Under the Microscope: Bioartography Quilts project in conjunction with University of Michigan Center for Organogenesis) http://www.fiberartists-looseends.com/Bioartography/GutsGlory.html

 

Images are the currency of scientific discourse. Whether appearing in oral presentations, journal articles, or lab reports, images convey the results of your experiments. The validity of those images and the data they report are of great importance. However, when the integrity of published images falls into question, the results and conclusions aren’t worth much more than the paper on which they are printed.  So if a picture is worth a thousand words, how much is a scientific image worth?  

The Problem with Images
Recent years have witnessed an explosion of retractions and adulterated data finding their way into the literature. Research on recent publishing trends suggests that the increase in retractions is more frequently due to overt misconduct than to author error.1 Mike Rossner (former Managing Editor of The Journal of Cell Biology, JCB) has written several editorials on the subject of figure manipulation. His first was penned shortly after JCB implemented electronic manuscript submission, when he realized not that manipulation was happening, but that “there is the potential for it to be detected less frequently by our reviewers.”2 And there lies the rub—image processing is no longer in the hands of Medical Arts departments but in the hands of researchers, postdocs, and students, many of whom have not had appropriate training in using Adobe Photoshop as a tool or in the ethics of figure preparation. 

Since then, Rossner and his colleagues at The Rockefeller University Press have been on the leading edge of image preparation guidelines for authors.  In 2004 Rossner and Yamada3 wrote an editorial that outlined many of the sins of image manipulators, and they provide examples of figures in both original and manipulated form (I encourage its reading for a more detailed exposition of the trappings of image misdeeds). Around the same time, The American Journal of Pathology and The Journal of Molecular Diagnostics adopted Scientific Integrity Policies to lay out more clearly what conduct we expected of our authors, and we drew upon Rossner’s experience as well as the recommendations of other organizations.

Image Manipulation Defined
The US Office of Research Integrity (http://ori.dhhs.gov/) defines falsification as “manipulating research materials, equipment, or processes, or changing or omitting data or results such that the research is not accurately represented in the research record.” That description seems rather broad, so you may be wondering what manipulations or corrections to images are allowed and what are not. In general, it is considered acceptable to make adjustments of brightness, contrast, or color balance to an image if the alterations are applied to the entire image, whether experimental or control, and as long as they do not obscure or eliminate any information present in the original.

Many adjustments seem benign on the surface. The judgment lies not only in the final product but also in the motivation—any changes that enhance, obscure, move, remove, or introduce data are never appropriate. It would not be acceptable to increase the brightness or contrast of an image with the goal of making a band on a gel or blot fade into the background. Manipulating individual color filters on a fluorescence image to “improve” staining for a specific molecule is inappropriate. Further, the grouping of images from different parts of the same microscope field equals data fabrication. Manipulation can even be as simple as using a single control experiment and duplicating it in multiple figures.

 

A frequently overlooked manipulation is the “splicing” of different gels and blots or of lanes within a single gel or blot. At times an author may crop the blot images and paste them back together—splicing—to obtain the desired presentation, whether to present all data points in a single blot or to reorder them for presentation. Although seemingly innocent, there can be deceitful reasons for doing so, such as pasting together unrelated results to obtain the desired answer or to remove unwanted negative or contradictory findings. Splicing will surely illicit the concern of editors and reviewers unless the manipulation is explicitly disclosed by the arrangement of the figure (eg, with dividing lines) and in the figure legend. If a gel is not large enough to run all samples or the samples were not run in the “correct” order, either repeat the experiment or clearly label your final composition.

Guidelines for Image Acquisition
At the ASIP 2014 Annual Meeting, Dr. Eduardo Rosa-Molinar offered exceptional guidance of proper image acquisition techniques. In his presentation, “Image Acquisition and Manipulation” (part of the XIVth Annual ASIP/AAA Career Development & Mentoring Program and Lunch), he gave examples of improper fluorescence image processing. However, proper acquisition of data may be more important than how the images are processed. The most basic advice involves understanding the microscope and its settings. Proper training will inform your decision making and may save you from having to repeat experiments that were acquired using the wrong settings. Then, always save original source files and make copies before making any adjustments to brightness, contrast, color, etc. If there are problems with the data presentation, you may be required to provide original copies. Finally, always add scale bars before changing the size of the image, and your scale bar will then be accurate regardless of final image size.
Taking photographs of the same source under varied fields of view, light intensity, magnifications, or contrast conditions without disclosing that the data are not unique to the present study constitutes suspect scientific conduct. If your staining results do not look as good as the data obtained for your last paper, you cannot go back to the computer and use an old saved image. Further, unless serial sections are used, the publication of identical-appearing images labeled with different staining techniques in different figures or papers raises legitimate questions, so keep your files clearly labeled to avoid mix ups. If you are interested in learning more about proper immunohistochemistry techniques and analysis, including image acquisition, The Marine Biological Laboratory (MBL) offers an intensive four-day course each Spring (see box for details).

Immunohistochemistry and Microscopy (IHCM) Course

Offered by: The Marine Biological Laboratory (MBL)
Course Date: March 14 –19, 2015
Deadline: January 16, 2015 | Online Information at http://www.mbl.edu/education/special-topics-courses/immunohistochemistry-and-microscopy-ihcm/
Directors: Eduardo Rosa-Molinar, University of Puerto Rico-Rio Piedras and Charles W. Frevert, University of Washington School of Medicine

Ensuring Image Accuracy
At AJP and JMD, we explicitly state what is allowed and what is not in our Instructions for Authors: “No specific feature within an image may be enhanced, obscured, moved, removed, or introduced. The grouping of images from different parts of the same gel or blot, or from different gels or blots, fields, or exposures must be made explicit by the arrangement of the figure (eg, using dividing lines) and in the figure legend” (http://ajp.amjpathol.org/content/integrity). We and other journals update our policies regularly, so it is wise to review all relevant instructions and imaging policies when preparing a manuscript for submission.

As with most areas of misconduct, tools have been created to identify manipulated images. ORI has developed a series of add-on applications, or Forensics Actions, that can be used in conjunction with Adobe Photoshop to automatically process images and detect image irregularities. Their series of tools (freely available at http://ori.hhs.gov/actions) covers fraud such as erasure or whitening, splicing/cloning, and duplication. The Actions are easy to use and can make assessment of suspect images more standardized rather than simply subjective. Some journals utilize automated software to scrutinize accepted figures for any traces of manipulation, proper or improper. Many editors and reviewers rely on the naked eye to find possible errors in data presentation, but once suspected, ORI’s tools do make the initial assessment quick and easy.

Closing Remarks
As noted in my previous article (Part 1: Plagiarism in Our Midst, see http://www.asip.org/Trainee_Newsletter/Jun2014/trust.html), the consequences of misconduct are serious, resulting in potential embarrassment, retraction, and/or funding agency sanctions for the authors involved. But beyond that, the loss of faith in the scientific process and the costs of misdirected research funds for falsified data are difficult to quantitate. Most science is conducted in small labs, behind closed doors, so the integrity of the images that come out of those labs is all we have to ensure the validity of the results reported.  The value of your images is higher than you think, so steer clear of counterfeit data. Rossner and Yamada3 sum up the advice best: “If you would have redone an experiment to generate a presentation-quality image in the days before the digital age, you should probably redo it now.”

References

  1. Fang FC, Steen RG, Casadevall A: Misconduct accounts for the majority of retracted scientific publications. Proc. Natl. Acad. Sci. U.S.A. 109: 17028–17033
  2. Rossner M: Figure manipulation: assessing what is acceptable. J Cell Biol, 2002 158:1151 (Editorial)
  3. Rossner M and Yamada KM: What's in a picture? The temptation of image manipulation. J Cell Biol 2004, 166: 11-15