INTRODUCTION
Research integrity is based on adherence to core values – objectivity, honesty, openness, fairness, accountability, and stewardship.1 Scientific papers can contain inaccuracies stemming from various causes, such as unintentional errors (incorrect calculations, using the wrong reagent or flawed methodology) or deliberate errors that constitute research misconduct (manipulated, omitted, manufactured or misrepresented data) in order to achieve a desired outcome.2
Duplicated and manipulated images in published articles are a major concern as it reduces the integrity and credibility of biomedical science.3 A study published in 2016 revealed that 3.8% of published articles have been identified to contain such figures, with at least half of those suggestive of deliberate manipulation.2 We have recently observed inappropriate image manipulations in immunofluorescent (IF) images during the review process of rhinology research articles.4 The objective of this study is to assess the prevalence of duplicated images in rhinology research articles using artificial intelligence (AI) powered software.
METHODS
The PubMed/Medline database was queried for English-language studies using the following search terms: “immunofluorescence” plus “sinus” or “nasal”, up to October 2022. Studies without IF images were excluded for analysis. Once portable document format (PDF) file formats were retrieved, identified articles were then uploaded to the “ImageTwin (imagetwin.ai)” AI-based software to detect integrity issues. Each report from “ImageTwin” was reviewed by the authors and divided into three groups: Clean (no detection of duplicated images), Possible (suspicious but unable to verify), and Definite (containing duplicated images). The duplicated images were classified into three major categories2: simple duplications (Category 1), duplications with repositioning (Category 2), and duplications with alteration (Category 3): 1) Simple duplication means figures containing two or more identical panels, either within the same figure or between different figures within the same paper; 2) Duplications with repositioning include images with a clear region of overlap, where one image had been shifted, rotated, or reversed with respect to the other; 3) Duplications with alteration consist of images that were altered with changing background or contrasts, thereby adding or removing data points or a part of images with repositioning (Figure 1). Images falling under category 1 (simple duplication) are most likely the outcome of either an oversight to include identical images twice, or a deficiency in explanation if the author(s) using a control panel from a different experiment due to the lack of an actual control. Categories 2 (repositioning) and 3 (alteration), on the other hand, are more likely to result from misconduct, as they require conscious effort to complete.
Figure 1.
Schematic examples of simple duplication (Category 1), duplication with repositioning (Category 2), and duplication with alteration (Category 3). In simple duplication, the third row of FIGURE B is identical to the top image in FIGURE A (red boxes). In duplication with repositioning, yellow boxes (the third row of FIGURE A and FIGURE C) appear to show a region of overlap, suggesting these images were obtained from the same sample. In duplication with alteration, the second row of FIGURE A and third row of FIGURE D appear to be different, but they are identical images with different contrasts (purple boxes). An extra-finding (person) was added in the top row of FIGURE D (green box).
RESULTS
Our initial search identified 69 articles through the PubMed/Medline database query, and 2 articles were excluded due to lack of IF images. Of 67 articles, ImageTwin AI based software flagged 18 articles (26.8%) with duplicated images. Of those 18 articles, nine articles (13.4%) were found to contain duplicated images and the other nine articles contained possible duplicated images, but without confirmation.
The details of nine articles with duplicated images are summarized in Table 1. Of the nine articles with duplicated images, a total of 19 incidents were identified and a maximum of five incidents were noticed in one article (all category 3). The majority of duplications were classified as category 3 (n=14) [category 1 (n=2), category 2 (n=3)]. One article with category 1 duplications utilized an identical representative figure in two separate images, which could be an oversight, or an absence of explanation provided within the articles. Eight articles containing category 2 or 3 duplications were from different journals. The mean impact factor of those eight journals was 5.9+/−1.9 [2.3–8.786]. Based on the location of corresponding author(s)’ institution, these articles are originated from Asia-Pacific countries [China (3), South Korea (3), Australia (2)]. No corrections or erratum were published in the journal at the time of analysis.
Table 1.
Summary of publications with duplicated images
| No | PMID (PMCID) |
Journal Title | Impact Factor | Number of duplications (category) | Figures flagged by ImageTwin™ |
|---|---|---|---|---|---|
| 1 | 34975898 (PMC8716742) |
Frontiers in Immunology | 8.786 | 2(3,3*) | 1F, 2D |
| 2 | 34977034 (PMC8718617) |
Frontiers in Cell and Development Biology | 6.081 | 1(3) | 1B, 4A |
| 3 | 34734505 (PMC8569026) |
Allergy, Asthma, & Immunology Research | 5.096 | 2(3,3) | 3 |
| 4 | 28492611 | Rhinology | 6.634 | 5 (3,3,3,3,3) | 4A, 4B |
| 5 | 25835956 (PMC4507357) |
Stem Cells and Development | 4.390 | 3 (2*,2*,2*) | 4C, 7A |
| 6 | 36187946 (PMC9515799) |
Frontiers in Microbiology | 6.064 | 1 (3) | 4 |
| 7 | 33736454 | American Journal of Rhinology and Allergy | 2.3 | 2 (1*,3) | 4C, 5A, 5D |
| 8 | 32623837 | International Forum of Allergy and Rhinology | 5.426 | 2 (3*,3) | 2B, 3A, 2C |
| 9 | 29935218 | J Allergy Clin Immunol | 14.29 | 1 (1*) | 1A, 2B |
PMID: PubMed Identification Number
PMCID: PubMed Central Identification Number
non-immunofluorescence image
DISCUSSION
Doubts regarding the caliber and veracity of the scientific literature have been on the rise.2,5–7 In the present study, we investigated the prevalence of duplicated images in rhinology research articles containing IF images using AI powered software. When analyzing nearly 70 articles from 10 primary research journals, the following are the major findings from this study: 1) Figures containing inappropriately duplicated images were not easily identified through visual inspection and AI based forensic software was required to identify category 2 or 3 duplications; 2) Approximately 1 of every 8 published articles with IF (12%) may contain inappropriately duplicated images (mostly category 3); 3) Category 2 or 3 duplicated images were published in higher impact journals with the average impact factor approximately 6; and 4) A group of two authors repeatedly duplicated images another paper. The results of these findings carry significant implications for the field of biomedical research, indicating a pressing need to enhance quality of our literature through heightened vigilance and education.
Publishing high-quality research in a journal involves more rigors and transparent reporting guidelines about experimental design and statistical descriptions.4 At present, a majority of journals only inspect certain images, with only a limited number utilizing automated procedures.8 Editors of the biomedical journals should have a responsibility to the readers and authors to ensure that what they publish is scientifically sound and meets the highest ethical standards.9 Prager et al. argued that rigorous reporting guidelines and publication standards cannot guarantee compliance.10 In order to achieve transparent and reproducible research, training and education on key aspects of research design and analysis are vital. This pilot analysis demonstrated a nearly 12% prevalence of inappropriate image duplication, compromising research integrity.
KEY POINTS.
Duplicated images in research articles erode integrity and credibility of biomedical science.
Forensic software was necessary to detect figures with inappropriately duplicated images.
This analysis reveals a significant issue of inappropriate image duplication in our field.
Funding Sources:
This work was supported by National Institutes of Health (NIH)//National Heart, Lung, and Blood Institute (1 R01 HL133006-05)/National Center for Complementary and Integrative Health (R21AT01223-01) and the Cystic Fibrosis Foundation Research Grant (002481G221) to BAW; and NIH/National Institutes of Allergy and Infectious disease (K08AI146220, 1R21AI168894-01), Triological Society Career Development Award, and Cystic Fibrosis Foundation K08 Boost Award (CHO20A0-KB) to DYC.
Footnotes
Bradford A. Woodworth, M.D. is a consultant for Cook Medical, Smith and Nephew, and Medtronic. The authors whose names are listed other than Bradford A. Woodworth, M.D. certify that they have no affiliations with or involvement in any organization or entity with any financial interest in the subject matter or materials discussed in this manuscript.
REFERENCES
- 1.National Academies of Sciences E, Medicine. Fostering Integrity in Research Washington, DC: The National Academies Press, 2017. [PubMed] [Google Scholar]
- 2.Bik EM, Casadevall A, Fang FC. The Prevalence of Inappropriate Image Duplication in Biomedical Research Publications. mBio 2016; 7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Bik EM, Fang FC, Kullas AL, Davis RJ, Casadevall A. Analysis and Correction of Inappropriate Image Duplication: the Molecular and Cellular Biology Experience. Mol Cell Biol 2018; 38. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Cho D Otolaryngologist Shares Experience with Image Manipulation in Research and How to Prevent It ENTToday. www.enttoday.org: The Triological Society, 2022.
- 5.Collins FS, Tabak LA. Policy: NIH plans to enhance reproducibility. Nature 2014; 505:612–613. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Fang FC, Steen RG, Casadevall A. Misconduct accounts for the majority of retracted scientific publications. Proc Natl Acad Sci U S A 2012; 109:17028–17033. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Gopalakrishna G, Ter Riet G, Vink G, Stoop I, Wicherts JM, Bouter LM. Prevalence of questionable research practices, research misconduct and their potential explanatory factors: A survey among academic researchers in The Netherlands. PLoS One 2022; 17:e0263023. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Butler D Researchers have finally created a tool to spot duplicated images across thousands of papers. Nature 2018; 555:18. [DOI] [PubMed] [Google Scholar]
- 9.Martin C, Blatt M. Manipulation and misconduct in the handling of image data. Plant Cell 2013; 25:3147–3148. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Prager EM, Chambers KE, Plotkin JLet al. Improving transparency and scientific rigor in academic publishing. Brain Behav 2019; 9:e01141. [DOI] [PMC free article] [PubMed] [Google Scholar]