The capabilities of generative AI in digital imaging is astonishing. This technology doesn't just create simple objects—it fills in background information and integrates this material seamlessly into the initial picture, making it nearly impossible to distinguish between original data and synthetically generated content.
In the realm of using images within scientific research papers, we need to question the general acceptance of a cropped Western blot image as proof for peer review. How can we be sure that those blurred lines truly represent the supporting data in an article? A minor, sometimes even accidental, edit of an image can have significant implications for the authenticity of your work.
A metadata study reviewing 1,300 papers found that 10% had at least one instance of image manipulation (1). Of these images, 80% contained gel electrophoresis images. This means that over 20% of papers featuring gel electrophoresis images involved some form of image manipulation.
Scientific publishing houses are heavily invested in monitoring peer-reviewed papers submitted with manipulated images to prevent this issue. However, the AI tools they currently use are time-consuming. Each image analysis takes considerable time and can never 100% ensure that an image hasn’t been manipulated. Consequently, researchers face delays in manuscript reviews and experience publication rejections.
It is all too easy for scientists to unknowingly manipulate the images they submit. Powerful image software not dedicated to scientific imaging can easily and subtly alter image properties in the raw data. Scientists filter images and modify the intensity of certain parts, leading AI software from publishing houses to flag these changes as image tampering.
There have been tremendous advancements in information security across various sectors, and scientific research is no exception. To establish image standards and preserve data integrity across project stages, the industry must enforce securing digital data. Standards like CFR 21 Part 11 and EU GMP Annex 11 are well-established operating procedures that customers must comply with when developing and marketing products.
To ensure data traceability, Cytiva has implemented the Image integrity checker and associated protocol to generate images with a hash-tagged fingerprint based on the SHA-256 algorithm. This fingerprint remains unchanged only if the original image is unmodified. Any modification to the raw data alters the associated tag, preventing it from reverting to its original value.
The Image integrity checker verifies the hash-tagged fingerprint to inform users if an image has been altered. This free tool is available to publishers and researchers, allowing them to check up to a thousand images simultaneously in seconds. Research labs with limited budgets or lacking standard operating procedures can now access a software tool to ensure their images are handled safely and provide robust backup data for their publications.
Safe image handling and easy verification of image authenticity are essential for modern researchers. Image capture, analysis, and data submission must follow a process that ensures images are safely acquired:
Image acquisition
Algorithms enhance dynamic range while preserving real relative intensities in digital imaging. SNOW (signal to noise optimization watch) maintains relative intensities by capturing images at subsequent time points and averaging the slightly different values of the same pixel location across these images. This operation is critical for every pixel to avoid enhancing or filtering specific parts of the image. Learn more about SNOW capabilities.
Image analysis
The analysis software must preserve image data, regardless of the operations performed. In any ImageQuant™ TL software image, data cannot be modified, ensuring secure handling.
Image submission
Some image modifications, like cropping for PDF manuscripts, are standard. However, publishers and peer reviewers should access the original image file to verify its authenticity. With the free Image integrity checker software, researchers can confirm their image data is raw and submit a certificate of authenticity with their manuscript and raw image. Publishers can also use the software to quickly verify image sources.
References
Bucci, E.M. Automatic detection of image manipulations in the biomedical literature. Cell Death Dis 9, 400 (2018). https://doi.org/10.1038/s41419-018-0430-3