Proofig AI has released a new system upgrade to help editors and institutions proactively check the integrity of scientific images before publication, founders of AI-enabled scientific image checking tool.
The new feature can identify potential instances of splicing in western band images and provide users with tools to investigate potential forms of image alteration.
As part of the new upgrade, Proofig AI software now scans a manuscript and its subimages, flagging potential instances of image duplications and of splicing in western blots, a prevalent form of image alteration.
After receiving the results, the reviewer can then investigate flagged images more thoroughly using the software’s enhanced visualisation tools, applying filters to emphasise details that may indicate alteration (Figure 1). The user can then determine how to proceed, seeking clarification from the researcher if necessary.
Dr. Dror Kolodkin-Gal, co-founder of Proofig AI, said: “AI software is not designed to be the judge of whether an image issue is deliberate or not. It’s there to highlight the suspected images that need attention, so a reviewer can focus their time on proactively maintaining image integrity.
“We’ve already seen how automating image integrity checks to find instances of duplication has helped editors and institutions save time, reduce the risk of human error and maintain the highest standards of integrity, and we expect this new upgrade will provide further support.
“Papers often contain hundreds of subimages, and many are difficult to differentiate — especially western blots — so thoroughly checking images manually can be time-consuming and inaccurate. Using AI to focus on potential alterations that are difficult to detect by eye therefore offers significant improvement both in time and accuracy.”
The main reason for creating Proofig AI was to provide the scientific research community with a tool to proactively check image integrity before publication.
According to research by image integrity analyst Jana Christopher: “Among the various journals where [she had] screened accepted manuscripts pre-publication, the percentage of manuscripts flagged up for image-related problems ranges from 20-35%.”
This post originally appeared on TechToday.