Information Integrity: Visual Images
- Page ID
- 130587
\( \newcommand{\vecs}[1]{\overset { \scriptstyle \rightharpoonup} {\mathbf{#1}} } \)
\( \newcommand{\vecd}[1]{\overset{-\!-\!\rightharpoonup}{\vphantom{a}\smash {#1}}} \)
\( \newcommand{\id}{\mathrm{id}}\) \( \newcommand{\Span}{\mathrm{span}}\)
( \newcommand{\kernel}{\mathrm{null}\,}\) \( \newcommand{\range}{\mathrm{range}\,}\)
\( \newcommand{\RealPart}{\mathrm{Re}}\) \( \newcommand{\ImaginaryPart}{\mathrm{Im}}\)
\( \newcommand{\Argument}{\mathrm{Arg}}\) \( \newcommand{\norm}[1]{\| #1 \|}\)
\( \newcommand{\inner}[2]{\langle #1, #2 \rangle}\)
\( \newcommand{\Span}{\mathrm{span}}\)
\( \newcommand{\id}{\mathrm{id}}\)
\( \newcommand{\Span}{\mathrm{span}}\)
\( \newcommand{\kernel}{\mathrm{null}\,}\)
\( \newcommand{\range}{\mathrm{range}\,}\)
\( \newcommand{\RealPart}{\mathrm{Re}}\)
\( \newcommand{\ImaginaryPart}{\mathrm{Im}}\)
\( \newcommand{\Argument}{\mathrm{Arg}}\)
\( \newcommand{\norm}[1]{\| #1 \|}\)
\( \newcommand{\inner}[2]{\langle #1, #2 \rangle}\)
\( \newcommand{\Span}{\mathrm{span}}\) \( \newcommand{\AA}{\unicode[.8,0]{x212B}}\)
\( \newcommand{\vectorA}[1]{\vec{#1}} % arrow\)
\( \newcommand{\vectorAt}[1]{\vec{\text{#1}}} % arrow\)
\( \newcommand{\vectorB}[1]{\overset { \scriptstyle \rightharpoonup} {\mathbf{#1}} } \)
\( \newcommand{\vectorC}[1]{\textbf{#1}} \)
\( \newcommand{\vectorD}[1]{\overrightarrow{#1}} \)
\( \newcommand{\vectorDt}[1]{\overrightarrow{\text{#1}}} \)
\( \newcommand{\vectE}[1]{\overset{-\!-\!\rightharpoonup}{\vphantom{a}\smash{\mathbf {#1}}}} \)
\( \newcommand{\vecs}[1]{\overset { \scriptstyle \rightharpoonup} {\mathbf{#1}} } \)
\( \newcommand{\vecd}[1]{\overset{-\!-\!\rightharpoonup}{\vphantom{a}\smash {#1}}} \)
\(\newcommand{\avec}{\mathbf a}\) \(\newcommand{\bvec}{\mathbf b}\) \(\newcommand{\cvec}{\mathbf c}\) \(\newcommand{\dvec}{\mathbf d}\) \(\newcommand{\dtil}{\widetilde{\mathbf d}}\) \(\newcommand{\evec}{\mathbf e}\) \(\newcommand{\fvec}{\mathbf f}\) \(\newcommand{\nvec}{\mathbf n}\) \(\newcommand{\pvec}{\mathbf p}\) \(\newcommand{\qvec}{\mathbf q}\) \(\newcommand{\svec}{\mathbf s}\) \(\newcommand{\tvec}{\mathbf t}\) \(\newcommand{\uvec}{\mathbf u}\) \(\newcommand{\vvec}{\mathbf v}\) \(\newcommand{\wvec}{\mathbf w}\) \(\newcommand{\xvec}{\mathbf x}\) \(\newcommand{\yvec}{\mathbf y}\) \(\newcommand{\zvec}{\mathbf z}\) \(\newcommand{\rvec}{\mathbf r}\) \(\newcommand{\mvec}{\mathbf m}\) \(\newcommand{\zerovec}{\mathbf 0}\) \(\newcommand{\onevec}{\mathbf 1}\) \(\newcommand{\real}{\mathbb R}\) \(\newcommand{\twovec}[2]{\left[\begin{array}{r}#1 \\ #2 \end{array}\right]}\) \(\newcommand{\ctwovec}[2]{\left[\begin{array}{c}#1 \\ #2 \end{array}\right]}\) \(\newcommand{\threevec}[3]{\left[\begin{array}{r}#1 \\ #2 \\ #3 \end{array}\right]}\) \(\newcommand{\cthreevec}[3]{\left[\begin{array}{c}#1 \\ #2 \\ #3 \end{array}\right]}\) \(\newcommand{\fourvec}[4]{\left[\begin{array}{r}#1 \\ #2 \\ #3 \\ #4 \end{array}\right]}\) \(\newcommand{\cfourvec}[4]{\left[\begin{array}{c}#1 \\ #2 \\ #3 \\ #4 \end{array}\right]}\) \(\newcommand{\fivevec}[5]{\left[\begin{array}{r}#1 \\ #2 \\ #3 \\ #4 \\ #5 \\ \end{array}\right]}\) \(\newcommand{\cfivevec}[5]{\left[\begin{array}{c}#1 \\ #2 \\ #3 \\ #4 \\ #5 \\ \end{array}\right]}\) \(\newcommand{\mattwo}[4]{\left[\begin{array}{rr}#1 \amp #2 \\ #3 \amp #4 \\ \end{array}\right]}\) \(\newcommand{\laspan}[1]{\text{Span}\{#1\}}\) \(\newcommand{\bcal}{\cal B}\) \(\newcommand{\ccal}{\cal C}\) \(\newcommand{\scal}{\cal S}\) \(\newcommand{\wcal}{\cal W}\) \(\newcommand{\ecal}{\cal E}\) \(\newcommand{\coords}[2]{\left\{#1\right\}_{#2}}\) \(\newcommand{\gray}[1]{\color{gray}{#1}}\) \(\newcommand{\lgray}[1]{\color{lightgray}{#1}}\) \(\newcommand{\rank}{\operatorname{rank}}\) \(\newcommand{\row}{\text{Row}}\) \(\newcommand{\col}{\text{Col}}\) \(\renewcommand{\row}{\text{Row}}\) \(\newcommand{\nul}{\text{Nul}}\) \(\newcommand{\var}{\text{Var}}\) \(\newcommand{\corr}{\text{corr}}\) \(\newcommand{\len}[1]{\left|#1\right|}\) \(\newcommand{\bbar}{\overline{\bvec}}\) \(\newcommand{\bhat}{\widehat{\bvec}}\) \(\newcommand{\bperp}{\bvec^\perp}\) \(\newcommand{\xhat}{\widehat{\xvec}}\) \(\newcommand{\vhat}{\widehat{\vvec}}\) \(\newcommand{\uhat}{\widehat{\uvec}}\) \(\newcommand{\what}{\widehat{\wvec}}\) \(\newcommand{\Sighat}{\widehat{\Sigma}}\) \(\newcommand{\lt}{<}\) \(\newcommand{\gt}{>}\) \(\newcommand{\amp}{&}\) \(\definecolor{fillinmathshade}{gray}{0.9}\)
Global Challenges |
Integrity of Information Visual Images |
Literature-Based Guided Assessment (LGA) |
Introduction
In the emerging world of generative AI, it is evermore imperative that the integrity of scientific research and publication be maintained. If we can't trust the veracity of images and data presented in publications, how can the overall trust in the scientific enterprise be maintained? Science skepticism has dramatically increased as more people question the utility of vaccines, the truth of human-caused climate change, and the causes and treatments of diseases. Researchers must live by personal and professional ethics as they use public dollars and communicate research findings that impact the lives of everyone. Students must learn about the process of scientific discovery and realize that scientists are people who have biases and pressures to publish that can lead to the dissemination of data and findings that contain errors, are misleading, and in some rare cases fraudulent. This Literature-Based Guided Assessment is based on publications that present "information" that is likely derived from simple unintentional errors to information that is likely based on fraudulent data. In the latter cases, the papers have been publicly retracted (or should be).
Our goal is NOT to cause students to lose faith in the integrity of the scientific enterprise (which includes the funding, performance, publication, review, and replication of research). Rather it is to humanize the process of science and make students better consumers, performers, and evaluators of scientific research.
This assessment will focus on papers containing suspect images, mostly pictures of Western Blots and cell culture plates. The person who has issued the loudest clarion call to hold researchers and scientific journals accountable for image integrity is Elisabeth Bik. She trained her eye to see problems in published images and has added to her investigatory repertoire AI software that can find many more across multiple publications. Since it is difficult to spot alternations made to images in programs such as Photoshop@, she focused her attention on three main types of image alterations, simple duplication (category I), duplication with repositioning (category II), and duplication with alteration (Category 3). Examples of each are shown below.
All images shown below, unless otherwise stated, are used as examples in a paper by Bik, Casadevall, and Fang (Creative Commons Attribution 4.0 International license) and her repository of images. The references to the papers with the images below can be found in the Bik et al. paper. Most of the images have been corrected in the original papers or the entire paper was retracted. We have attempted to just show the images without referencing the actual paper and its authors, which for this assessment, is not necessary. The presence of various duplicated images in some cases might be explained by a simple error in assembling the document for publication.
Category I: Simple duplications
In this type, a single figure containing multiple images in different panels shows evidence of duplication. Alternatively, the same image is used in another figure in the paper or even in a completely different paper. A common duplication in Western Blot images is the reuse of the same control showing a protein from a housekeeping gene like β-actin that should not change in different experiments. The same control image should not be used in pictures of Western Blots from different experimental conditions and cells.
Try to find duplications in the exercises below which show Western Blots and cell culture plates.
The image below shows a series of Western Blots. Find evidence for duplication.
- Answer
-
The blots in the green rectangles look identical.
Now try another example.
These images, found in two separate figures in a paper, show cells visualized with fluorescence microscopy. Find the image duplication issues.
- Answer
-
The panels outlined in red and blue in the top row below are duplicated and transposed in the bottom row.
Category II: Duplication with Repositioning
In this category, one image or part of it has been moved, rotated, and/or rotated from one image to create a new image.
Try to find duplications with repositioning in the images below.
The images below show the result of cell migration assays. Gy is a unit for the amount of radiation absorbed.
Look for duplications between panels within the photo.
- Answer
-
The colored rectangles indicate the areas of duplication
The duplications are clearly evident.
Here is another duplication with repositioning
The images below show Western Blots. Find the duplication with repositioning. In the experiments, the genes for IPPK, PPIP5K1, and PPIP5K2 were silenced using appropriate si-RNAs, and the cellular location of IRF3 was determined, + polyinosinic:polycytidylic acid (poly (I:C). Look for duplication with repositioning.
- Answer
-
The red rectangles show the duplicated and repositioned images.
The blots in the red boxes seem identical with a 2-lane shift.
Category III: Duplication With Alteration
This perhaps is the most egregious type of image manipulation as it would be hard to attribute it to simple unintentional error. These images contain duplications (parts or whole) that can be rotated/reversed not only within a single panel but between panels in one figure or between figures. A specific area might also be duplicated within the same panel (stamping). In addition, different backgrounds from sections of an image can be overlayed on another image (patching). Points (from cell plates or FACS analyses) can also be erased or added.
Try to find duplications with alterations in the images below.
Finding the duplication with alteration
- Answer
-
Lanes 9 and 10 are suspect.
Examples of photos from a single paper that documents the effects of a regulatory micro-RNA (miR-23a) in human adenocarcinoma cells are shown below. Find the likely problem in the photos below. The paper was published in a journal that has Creative Commons BY reuse permissions. Our goal with these images is not to understand the experimental design or the result, but simply to see if the images show evidence of duplication. Hence we won't give the reference unless a request is made by email. It can likely be found by a web search if users would like the original reference.
The last examples below come from a database of image duplication compiled by Elisabeth Bik and do not come from the Bik, Casadevall, and Fang (Creative Commons Attribution 4.0 International license) paper. The images all have Creative Commons BY permission for reuse.
Compare these plates. Is there evidence of duplication?
- Answer
-
The plate labeled pcDNA3 (the control plasmid, far left) appears to be identical to the plate labeled ASO-23a (the knockdown, far right). The plate labeled pri-miR-23a appears similar to plates ASO-NC.
Here is another photo.
Image A below shows the same experiment as above but with a different type of cells.
Image A
Image B below represents results from a separate image in the paper showing results when those cells were transfected with pSilencer/sh-IRF1, pcDNA3/IRF1, or the appropriate control vectors
Image B
Compare just the pcDNA3 in Image A and the IRF1 plate in Image B. Are there image issues?
- Answer
-
Yes. It appears that the IRF1 plate is the same as the pcDNA3 plate but rotated 90 degrees clockwise as shown below. (The red stripe in the figure below is part of the word RETRACTED placed over the image in the journal.
Here is one last set of images from the paper to compare.
Here are photos of cell plates from 2 separate experiments using the same cell type, but the images are taken from 2 separate figures.
In Image C immediately below, the cells were transfected with pcDNA3, pri-miR-23a,
Image C
In Image D below, the cells were transfected co-transfected with pri-23a and either pcDNA3-IRF1 or pcDNA3; the pcDNA3 group was used as the control.
Image D
Any image issues here?
- Answer
-
Yes! The pcDNA3 and pri-miR-23a plates in the top figure above appear similar to the left two plates in the figure below for the pcDNA3 and pri-miR-23a/pcDNA3 plates.
Now test your eyes with this more complicated example with lots of duplications within and between the panels. A reference to the paper will be supplied on request.
Try to find these duplications in the images shown.
Click Hint for some help if you are struggling to find duplicated "motifs" within and between panels!
- Answer
-
The duplications are circled in various colored circles and ovals.
Using AI to find image manipulation
It should be clear from the above examples that we need computer analyses and AI to detect present and more sophisticated image manipulations made possible with generative AI. Several companies have formed to analyze image integrity. In the next exercise, you will analyze images in a selected PDF (with author and title information redacted) and compare our results to the analysis of the same PDF file by ImageTwin, a company that uses AI-based software for detecting integrity issues in figures of scientific articles.
The analyses have two components:
- plagiarism detection to see if a figure or portion of it has been used in other published articles. ImageTwin has over 21 million images in its database;
- manipulation detection to see if duplicates and fabricated data are found in articles.
The questions below were derived from one PDF file which was analyzed by the ImageTwin for this problem set. The anonymized images from the article are used in the questions below.
This is a transmission microscopy (TEM) image of nanoparticles containing a biologically active molecule in a biocompatible and biodegradable polymer. The size distribution of the particles in shown in the bottom graph.
Look for signs of image duplication. (Hint: focus in the area around the nanoparticles)
- Answer
-
Areas of clear duplication within the greed boxes are clearly evident.
sdsd
dfd
Here is another example that would be very difficult to discern without computational analysis.
Look for duplication of parts of Figure C in G.
- Answer
-
The analysis is shown below. The left-hand panel shows the area around the red arrow in C above, while the right-hand panel shows the area around the black area in G but it is rotated 1800.
Here is another set.
Look for duplications in D and E
- Answer
-
Look at the area in the Green boxes in E and D
-
They are blown up below.
Here is one more.
Look for the duplication within one image below.
- Answer
-
The duplications are shown in the green boxes and also blown up.
Here is the last example from the paper.
Check for duplications within the image.
- Answer
-
Here they are (green boxes and zooms)
Recommendations
It is going to be more difficult to find images with integrity issues as AI becomes more powerful and more widely used. Nevertheless, steps can be taken by journals to standardize figure submission. A recent paper offers a checkbox list of requirements that authors and publishers should follow in taking, analyzing, and publishing images of cells to maximize image integrity.