Microscopy is now entirely digital. This makes acquisition much easier, but it also make image manipulation easy for anyone to do. 1% of manuscripts that were accepted by the JCB had to be revoked because of fraudulent imaging practices (Rossner, 2006). These were instances of blatant fraud such as cut-and-paste or erasing image features. More troubling, 25% of accepted manuscripts had to remake at least one figure due to not following publication guidelines. Why is this important? Because at the core, images are numerical DATA created by scientists.
A very detailed paper on the dos and don’ts is Cromey, 2010. Here we will try to condense some of the guidelines from the paper.
Acquire images in a way that represent accurate numbers. Digital images are composed of individual pixels that each have a “value.” In an 8-bit image that number is between 0 (black) and 256 (white). If we artificially set the background to 0 or oversaturate bright features, we are no longer on a linear scale. If you won’t change an number you got in a different experiment, don’t do it to your image.
Always keep your raw data. When you are processing, copy the data and then begin your work. Not only is it good practice to always have two copies of your data, it could protect you from accusations of fraud.
Small changes that are applied to the entire image is generally acceptable. Changes in brightness & contrast are ok if they are not used to mislead the audience. Most importantly, keep a detailed record of what changes you made. Cropping out confounding data or manipulations to just specific areas are not ok.
Don’t treat your images like you would on Instagram. Filters can make the numbers no longer linear or obscure data. If you were only able to get the perfect image of the feature you were looking for once, does it represent reality? Remember: Treat your images like you would any other data, alteration to make it beautiful can remove important data.
More sources for image manipulation guidelines: