The importance of quality control in scientific images

Tuesday, 15 August, 2023

The importance of quality control in scientific images

Maintaining integrity is critical in upholding public trust in science and credibility in the scientific community. Principal investigators (PIs) work diligently to ensure all the information published by their team is accurate, but reviewing every image produced by the lab is a definite challenge. Dr Dror Kolodkin-Gal, virology and cancer researcher and founder of image integrity software company Proofig, explores how image issues occur and how PIs can more easily manage image integrity.

In a laboratory setting, the PI is typically responsible for ensuring that the research they and their team share is credible, repeatable and accurate, across both written content and images. But although PIs recognise the importance of maintaining image integrity, they usually find it difficult to effectively perform the image analysis needed before publication. Instead, PIs may only have time to carry out inspections by eye.

Manually checking for image integrity issues is a long, daunting and arduous process that often results in failure. Before submitting the research to a potential publisher the PI must not only review each image to look for minute duplications between one or two images, but check every image against itself and every other image included in the research. So if the paper contains 100 images, that could equate to about 10,000 comparisons. This is not a task for a person without the aid of computational capabilities, even if they have plenty of spare time.

As instances of retractions caused by image duplications and manipulations become more prevalent, PIs must consider how to ensure the entire laboratory prioritises image integrity.

The image integrity problem

Failing to identify image integrity issues prior to submission, be it for grant applications or publication, can lead to rejection. Alternatively, if an integrity issue remains undetected during review but is later reported either to the journal or online, the publisher is usually obligated to conduct an investigation.

Many in the scientific community believe that there is no malicious intent within their lab, and therefore a low risk of retraction due to image issues. Indeed, image integrity issues are often due to honest mistakes. But the investigation process can take years, as forensic investigators look at the allegation, the research origins and suspected outcomes — during which time researchers may encounter challenges in securing further funding, carrying out research or finding alternative publishing avenues. Consequently, regardless of the investigation’s outcome, researchers must work hard to rebuild their reputation.

According to leading image data integrity analyst Jana Christopher MA, between 20 and 35% of manuscripts are flagged for image-related problems1. While some of these issues are due to deception or misconduct, research suggests that fabricated content only accounts for a small portion of image issues reported. This was evidenced during a trial that ran from January 2021 to May 2022, where the American Association of Cancer Research (AACR) used Proofig’s automated image integrity software to screen 1367 papers accepted for publication2. Of those, 208 papers required author contact to clear up issues such as mistaken duplications, and only four papers were withdrawn. In almost all cases (204 cases), there was no evidence of intentional image manipulation.

How image issues occur

PIs and their teams must increase awareness throughout the lab of any potential image issues that can occur, particularly unintentionally, to reduce the risk of paper retractions and corrections post publication.

Research teams can take various measures to mitigate duplication, but completely avoiding this is not always possible, especially because minute issues are difficult to detect by eye. When collaborating on research, it can be particularly difficult to ensure good image management during the experimental phase across a wider team.

Consider this example. A cancer researcher is looking to study the efficacy of a treatment for pancreatic cancer. This involves looking at cross-sections of pancreas in a slide, reviewing a control sample alongside samples with different concentrations of the treatment. Over the entire experiment, the researcher may be required to collect hundreds to thousands of images of specimens. If imagery of the entire slide is unclear when examining specimens under the microscope, the researcher may move the lens and change magnification to capture different areas of the pancreas in detail. Depending on the magnification, the researcher may need to move the microscope from left to right and up and down to document every section of the slide.

If a paper is published with two overlapping parts of an image used, with scaled versions (taken by different lens magnifications) or with an image that was mistakenly saved twice with different names, the researchers and the PI could not notice these errors and their paper may be flagged for image duplication. Unfortunately, the microscope itself does not alert researchers to potential overlaps when capturing images, so researchers may unintentionally duplicate some images.

The future of image integrity detection

PIs can no longer rely purely on traditional, manual checks. Many are therefore considering how to streamline the review process by introducing proactive quality control measures that reduce the risk of image issues.

The good news is, advancements in AI and computer vision have led to the development of valuable tools for scientists. Researchers and publishers can now use online tools to check their content for grammar, readability and plagiarism. Similarly, publishers and researchers can now use software to automate the image checking process.

AI image proofing software can be trained to automatically scan images in a paper, checking each against itself and others in the paper to flag any anomalies. PIs can use the software to review all the research produced by the team, rapidly generating reports that outline any potential issues. These issues can then be investigated further by the PIs, who can pinpoint their origins.

If the detected error needs attention, the PI and their team can hold off until the paper is as accurate as possible. This ensures a higher standard for the paper and the conclusions of the experiments. Therefore, the AI-driven approach not only optimises the verification process, it also contributes significantly to enhancing the overall quality and reliability of scientific research publications.

Improving awareness and providing support about image integrity can help a PI minimise the risk of their research team introducing image issues to their manuscripts from the outset. Implementing quality control procedures and using the right quality control software for both text and images prior to paper submission is currently fundamental to safeguarding the reputation of the team and their manuscripts.

By understanding how image duplications occur, and establishing processes to more easily detect issues, PIs and their teams can ensure image integrity throughout the research and publication process.


Image credit:

Related Articles

Deep learning decodes brain signals to identify ADHD

Neuroscientists, neurologists, psychiatrists and psychologists alike struggle to connect...

Machine learning aids neurosurgeon during brain surgery

Deep brain stimulation implants a stimulus electrode in the brain that then delivers electrical...

Transforming life sciences and health care with graph technology

Graph technology can help life sciences users integrate and analyse large, complex datasets,...

  • All content Copyright © 2023 Westwick-Farrow Pty Ltd