NetNotes


Look at the histogram for each channel and use a variance filter.


If bits have been cut and pasted from different original images, the variance in the background may differ. A more sophisticated option is to look for the local Poisson noise. Tis should be the same in all areas of the image with a similar intensity, but will differ if cutting and pasting has been used. However, there is competition between those committing fraud and those chasing fraud, and both improve. Tere is also major interest in comparing papers to check if the same image has been re-used. Jeremy Adler jeremy.adler@igp.uu.se


I think that you might be referring to ORI Droplets that can be used


in Fiji. https://ori.hhs.gov/droplets Brian Armstrong barmstrong@coh.org For


a first quick check of images: https://29a.ch/photo-


forensics/#forensic-magnifier. For more detailed analysis: InspectJ in FIJI: https://github.com/ZMBH-Imaging-Facility/InspectJ and ORI Forensic tools: https://ori.hhs.gov/forensic-tools. However, the ORI tools are not recommended for teaching purposes, as this is based on Photoshop actions and you want to keep any next generation life scientist away from Photoshop as long as possible as this is clearly NOT what should be used on scientific images. Tere are also commercial solutions (Mike Rossner is now running a service). Oliver Biehlmaier oliver.biehlmaier@unibas.ch


Similar to the suggestion of searching for local changes in Poisson


noise, looking at the Fourier transform of an image can tell you a lot. A simple cut-and-paste or lossy compression can add high-frequency harmonics, and convolutions are easily discernible in Fourier space. Even convolutions that are hard to visually distinguish in image space, such as a Gaussian filter versus a mean filter, are easy to distinguish in Fourier space: http://bit.ly/2NBQQfO. Boundaries of dissimilar regions of an image can also be found easily with a simple high-pass filter. Ben Smith benjamin.smith@berkeley.edu


You don’t really need any of these tricks to discern fraudulent


images. As an extreme example, see Figure 5d here [https://www. sciencedirect.com/science/article/pii/S0920586118310848] or here if paywalled: [https://scihubtw.tw/10.1016/j.cattod.2019.01.024]. Te curves are just scaled copies of the same curve, as pointed out here [https://pubpeer.com/publications/71B5E2EF6A7716D7F7F3B27 3E86926]. Unbelievable! Worth noting, aſter a couple of retractions (for example, in Nature Communications, see [https://pubmed.ncbi.nlm. nih.gov/33239646]) and countless allegations of fraud, the group is still in business, publishing, and receiving grants. Don’t let anything like this happen (even unintentionally), as it could (and should) ruin your career! If anyone has a suspicion of undisclosed image manipulations in their group, just talk to your lab members. It’s OK to make figures nicer and easier to understand, just don’t hide it. Even Photoshopping is fine, as long as you disclose it (well, the reviewers might not be happy, but you can respond in the rebuttal that “we achieved the same result in ImageJ by doing this, this and this…”). For honest and open science. Zdenek Svindrych zdedenn@gmail.com


I completely agree with your concluding statements in your post.


Oſten, during image analysis, there are image processing steps that dramatically modify the image histogram and manipulate the images as part of the analysis workflow. However, I emphasis to the researchers to document every step of the workflow and justify each step. Tis way any image modification/manipulation is fully transparent and adequately justified for the reviewers and readers. In fact, I usually recommend to users to create an image stack that captures each


2021 May • www.microscopy-today.com


modification made as a slice in that stack. Te final stack can even be uploaded as supplementary images when possible. I believe that transparency is key. Praju Vikas Anekal p.anekal@auckland.ac.nz


I really like the supplemental movie stack idea. I’ve also used flow


charts to show the processing steps along with a link for downloading a macro that does these steps: http://bit.ly/3jYyC4e. Both of these ideas are a win-win, because not only does it clearly disclose the processing steps for people who may want to reproduce the analysis, but it also makes it much easier for the reviewers to understand how each step impacted the image. Ben Smith benjamin.smith@berkeley.edu


This “Z-stack of image manipulation” is indeed a great idea.


However, for this one, would also need the appropriate tools. It is clear that you can’t present such an “image processing stack” for all your data. A 3D multicolor time-lapse movie, such as an experimental file, can be quite large and with a “processing stack” the size of this file would be multiplied. The solution would be to select one representative (single plane, single color, single time point) image to demonstrate the process. But, selecting a representative image is not easy, not only conceptually but also technically, and the concept fails if you want to do a manipulation along the time axis. Gabor Csucs gabor.csucs@scopem.ethz.ch


I think the best approach is to keep primary data together with


the program script that produced the final image, in the same folder. We previously used IDL but have switched to MATLAB for all image processing and analysis, so our code is available and code parts can be re-used (such as complicated segmentation routines). Of course, there is a steep learning curve to using/developing such scripts, but at least we can be sure of the reproducibility of the results, and no intermediate images need to be stored, so it is space efficient. The downsides might be: 1) Steep learning curves (but the increased depth of understanding offsets this). Most undergrads I’ve met are able to get to grips with, and can do, simple image processing in these environments; 2) The time to write a program to open a data set, run a gaussian filter, and store the results takes a bit longer than clicking on buttons in ImageJ. This difference disappears if many images need to be processed in the same way; 3) Cost can be prohibitive. Some universities have site licenses, but if you must pay for the license it is a problem if it has not been budgeted in grants. I know that Python/SciPy is a free tool that is powerful, but the learning curve is (I think) steeper because it is somewhat lower level than IDL/MATLAB. In addition, documentation is generally weak, and the user interface is poor. There may be fewer user-submitted and -tested library routines, but this may improve. I am not sure how easy it is to develop complicated image processing programs in this environment (you get what you pay for), and since I’ve never encountered anything that can’t be done with MATLAB plus extensions, I’ve never felt the need to use Python/SciPy/NumPy; 4) Reluctance to come to grips with programming, but computers are a slide rule to today’s scientist so why not learn to unleash its full power if you want to be a professional scientist?; 5) There is often a lack of local support in use of the tool, but help groups exist. Mark Cannell mark.cannell@bristolac.uk


You can do the same as Mark suggests using ImageJ/Fiji,


no costs involved. ImageJ/Fiji includes a relatively easy macro language and there are many online resources out there for help and advice. ImageJ/Fiji also includes a recorder that allows you to record analysis steps in ImageJ macro language, JavaScript,


83


Page 1  |  Page 2  |  Page 3  |  Page 4  |  Page 5  |  Page 6  |  Page 7  |  Page 8  |  Page 9  |  Page 10  |  Page 11  |  Page 12  |  Page 13  |  Page 14  |  Page 15  |  Page 16  |  Page 17  |  Page 18  |  Page 19  |  Page 20  |  Page 21  |  Page 22  |  Page 23  |  Page 24  |  Page 25  |  Page 26  |  Page 27  |  Page 28  |  Page 29  |  Page 30  |  Page 31  |  Page 32  |  Page 33  |  Page 34  |  Page 35  |  Page 36  |  Page 37  |  Page 38  |  Page 39  |  Page 40  |  Page 41  |  Page 42  |  Page 43  |  Page 44  |  Page 45  |  Page 46  |  Page 47  |  Page 48  |  Page 49  |  Page 50  |  Page 51  |  Page 52  |  Page 53  |  Page 54  |  Page 55  |  Page 56  |  Page 57  |  Page 58  |  Page 59  |  Page 60  |  Page 61  |  Page 62  |  Page 63  |  Page 64  |  Page 65  |  Page 66  |  Page 67  |  Page 68  |  Page 69  |  Page 70  |  Page 71  |  Page 72  |  Page 73  |  Page 74  |  Page 75  |  Page 76  |  Page 77  |  Page 78  |  Page 79  |  Page 80  |  Page 81  |  Page 82  |  Page 83  |  Page 84  |  Page 85  |  Page 86  |  Page 87  |  Page 88  |  Page 89  |  Page 90  |  Page 91  |  Page 92