FranceSoir – midjourney Sprinkler watered before the backlash: Elizabeth Bik, "cherry picking" studies!
Following Elizabeth Bik's allegations of fraud, many of which are against the IHU-Mediterranean, we became interested in her work. Indeed, many gray areas were noted in one of her 2022 articles and she is far from being above suspicion.
For example, regarding her past, she declares having left uBiome, the American startup which in 2019 will be the subject of a resounding bankruptcy and an FBI investigation causing the two founders to hastily leave the United States to seek refuge. in Germany! Elizabeth Bik was the company's scientific director and wrote numerous articles, published with the founders of uBiome without ever requesting retraction even though she knew that the microbiota tests sold by the company did not work. The investigation carried out on this subject will show substantial insurance fraud since old samples were re-analyzed by being re-invoiced to the insurance a second time! Elisabeth Bik, as a conscientious scientist could not not know this. The darkness surrounding this situation has been confirmed by microbiome specialists. For them, “Bik made a deal with the founders or federal investigators”.
Another gray area concerns the effective date of its departure from uBiome. Bik said on Questioned by France-Soir on two occasions, she did not find the time to answer our questions…
However, by looking more closely at Elisabeth Bik's declarations of fraud, the analysis images of Charles V. attract attention. Elisabeth Bik reportedly specializes in identifying photo fraud and photo duplicates. Charles V., using his expertise in computer graphics, noted numerous inconsistencies in his analyses. Is this one of the reasons why she did not respond to our requests?
We present the analyzes below as well as the reasons why < strong>Elisabeth Bik is greatly mistaken in her conclusionswhich could only be linked to problems of definition of the published images. The consequences of these errors, because they contradict the accusations of fraud of many scientists, thus contributing to sully their reputation by casting doubt on their work.
Introduction to the analysis of verifying BIK claims
Elisabeth Bik specializes in “detecting studies which, according to her, would be tainted by fraud”. To do this, it uses the Pubpeer platform which presents a series of tools intended to strengthen the traditional peer review process, including allowing comments from scientists and members of the public. A form of social network for exchange on science.
Bik claims on its site an activity of “scientific integrity consultant” and “1069 retractions, 149 expressions of reservations, and 1008 Corrections (in November 2023 )”. Bik specifically targets the publications of the IHU-Mediterranean.
Charles V. was therefore interested in Bik's allegations on several articles from the IHU-Marseille.
- A first article reanalyzed leads to erroneous conclusions
Let's take a look at the article “Use of rpoB genetic analysis for the detection and identification of Bartonella species”published on February 1, 2001. Bik comments on Pubpeer ;the similarity of areas in a photo, thereby insinuating that this could constitute fraud.
Madame BIK's comments are as follows: “On images 3 and 4:
- Boxes of the same color show areas (some including bands) that look more similar than expected.
- Green boxes indicate sharp background transitions.
- The area above the bands in lanes 1 to 5 in Figure 4 also appears to show repetitiveness.”
She adds thatthe “image has been made darker to bring out more features. This is probably far from the case, but would the authors still have the original photos on gel?”
Checking the first article challenged by Bik
A verification of Bik's assertions is necessary. After all, in science, the important thing is replicability, but it is also a matter of tools and methods. The verification process is explained below.
In order to verify a photo, we extracted the image questioned by Bik from the original file in PDF format to guarantee its authenticity. Its properties and characteristics are as follows: img60.png PNG 1960×1198 1960×1198+0+0 8-bit Gray 256c 471046B 0.000u 0:00.000
Bik claims, without real evidence,that two areas of the photo would be duplicated.
Using simple photo analysis tools, and after detailed examination, it appears that the Bik's argument is flawed. It cannot be a duplication, but two areas which, although they have visual similarities, cannot result from a duplication.
As is taken who thought he was taken.
p>
Instead, let's see, based on two methods, how Bik concludes in a completely erroneous way.
First Method: check if “areas identified as duplicated” really are /strong>
The photo in which there would be duplicated areas is as follows.
Using image editing software (GIMP):
- The Bg zone is extracted. This area is 84 pixels wide by 51 pixels high;
- A copy and paste of this area is duplicated on the photo in order to have a test or control area. This therefore makes it possible to duplicate with certainty the area which is supposed to have been duplicated, the famous hypothesis put forward by Bik.
In addition, the colometry in Pubpeer is different from the original image in the PDF and Bik file writes in his article that “the images were further examined for evidence of image duplication or manipulation using the color adjustment tool in Preview software on an Apple iMac computer.”
Ironically, if Bik turns off her screen, or sets the brightness to zero, she will find a big copy and paste of black. Aside from this ironic trait, a screen can be calibrated, there are devices that allow you to do this.
In conclusion, these non-automated and non-reproducible image analysis techniques have no scientific character and are purely subjective.
Methodology 2: identification of points remarkable
Bik may not use an automated and reproducible scientific approach to identify its frauds, however it seems important to us to test another method.
The second method used is that of comparing remarkable pointsof the image from a very powerful detection tool, FAST Feature Detector.
This tool is very often used to do point tracking for image stabilization. The hypothesis is: if two areas are similar, they have the same remarkable points. (https://docs.opencv.org/4.x/df/d0c/tutorial_py_fast.html< u>).
Here is the result obtained on the same image. Clear result confirming the hypothesis that there is no copy-paste as suggested by the fraud hunter: between the original zone, the copied zone, the remarkable points are similar, on the other hand for the zone “ imagined” by Bik, they are different by their cardinalities, but also by their positions. Which is compatible with the first methodology
- Second article checked : Bik conclusions erroneous
As part of the second article identified as potentially having duplicate areas in an image (Survival of Environmental Mycobacteria in Acanthamoeba polyphaga) a simple analysis was carried out on the image which may seem ridiculous both on a methodological and scientific level. The image was saved and opened in GIMP image editing software.
We will therefore verify Bik's analysis by demonstrating that the online service used is not 100% reliable.
We recreate a duplicate of the area that Bik considers identical by checking with the matching application used in the very first check: at the threshold of 0.999, there is no duplicate area identified , but there is one at the threshold of 0.97. Déjà vu: same result and same conclusion.
Unscientific analysis, because Bik does not indicate how she managed to obtain this result, which is therefore not replicable.
However, we went further by using Sherloq, a software that Bik recommends on his blog. This shows that Bik knows how to search on GitHub. This software is recognized by the community as a forensic analysis tool.
Sherlog allows you to display significant points: although geographically “fairly” close, they are different, as are their cardinalities.
Sherlog offers several methods to detect copy-pasted
With the BRISK (Binary Robust Invariant Scale Keypoint) method: no detection. With the ORB Oriented FAST and Rotated Brief method, with a sensitivity threshold of 5% matching, no copy.
With a sensitivity threshold of 10%, we find results that could be similar to those disputed by BIK.
Let's take the analysis further and focus on the areas of different colors in image 4 which we reproduce below:
Simply by putting the areas side by side, the difference is visible to the naked eye.< /p>
Which are not confirmed with Sherloq! >