In just the last twelve months, there has been a wave of hype around artificial intelligence (AI) image generation tools. DALLE-E, Midjourney, and Stable Diffusion have brought joy to many users, who can create novel images from just a few words. In science, as in art, AI tools are also changing the way things are done. Here Dr. Dror Kolodkin-Gal, founder of Proofig, explores how AI is changing the image integrity landscape.
Research from Morten P. Oksvold found that 25 per cent of all life science articles published may contain duplicatedor manipulated images. Many of these duplications are not the result of fraudulent research, but honest mistakes.
It is important we reduce this, to protect the accuracy of the scientific literature. However, a single paper could contain hundreds of sub images, and checking them manually can be time consuming, as well as carry low probability that the researcher will identify all of the issues, which are hard to spot with the human eye and may require hundreds or thousands of comparisons per paper.
With journal editors under increasing pressure, there is similarly low probability the issues will be detected at peer review. In many cases, they are picked up post-publicationand result in investigation and potential retraction or erratum, putting reputations of researchers, journals, and even science itself, at risk.
Overcoming image integrity issues
To address this challenge, we launched Proofig, an automated image duplication and manipulation tool that can be used by researchers and journal editors alike. Thissoftware carries out a range of checks so they can be addressed before publication. It can detect full duplications (include scaling or rotating), partial duplication or overlapping, alongside flipped images, figure manipulations, copy and paste, cloning — and any of these issues combined. Each image is compared with itself (for internal duplications) and with all others (for image issues as described), as well as comparing each flipped image with the other images within the uploaded files.
Professor Ofer Mandelboim, head of the department of immunology and cancer research at Hebrew University Faculty of Medicine, Israel said: “I’m happy to provide my department with the peace of mind that comes with knowing our manuscripts’ images have been properly verified pre-publication using Proofig, and only manuscripts with no mistakes will follow for publication.“
Proofig is increasingly being adopted by publishers as part of the review process. For example, the American Association for Cancer Research (AACR) found that checking for duplication using Proofig took 4.4 minutes compared with eight minutes manually, and identifiedeleven duplications in eight manuscripts, compared with five duplications in three manuscripts manually.
As AI-based image software for art continues to gain traction, AI tools are also advancing the scientific industry. By adopting this approach, publishing houses, universities, and authors can help maintain the integrity of the scientific literature and prevent harm to the field of science itself through enhanced quality control.
Do you want to learn more about the benefits of AI tools for maintaining image integrity? Find out more about Proofig and try the software yourself by visiting us at https://www.proofig.com.
