A Call for Color Calibration
Why digital slide scanners must be correctly calibrated to yield trustworthy results
Richard Salmon | | Longer Read
At a Glance
- For histopathologists, it is vital to ensure that digital images of stained tissue accurately represent the original slides
- Color calibration for whole slide imaging can verify accurate color representation
- Calibration is a complex issue, especially given the variations between different scanners and imaging software
- A physical, slide-based device may offer a good approach to universal calibration
Many laboratories are moving into digital imaging – scanning slides, viewing them on computer monitors, and using software-based tools to assist with analysis. But there’s an understandable hesitation from some pathologists, who ask: “How can we be sure that the digital image we see is truly representative of the slide?” In particular, with histopathology’s need for accurate staining, how can practitioners be assured that nothing is being lost or inaccurately represented in the digitization process?
Guidance from the US Food and Drug Administration (FDA) has recommended color calibration for whole slide imaging (WSI), just as it does for many other digital systems. There is substantial variation in color between digital slide scanners, a fact the FDA recognizes, and standardization of color can increase the validity, reliability and quality of WSI.
Color calibration seeks to ensure color fidelity throughout the imaging process. The goal? To provide the pathologist – who may be looking at an image scanned in their own lab or one scanned a lab across the globe – with the confidence that the colors they see are representative of the true colors of the tissue sample and are not altered by the scanning process. This gives pathologists confidence that they are working from ground-truth data.
But what exactly is color calibration – and why should we trust it to provide a true representation of tissue slides?
Bringing color fidelity to biomedical imaging
Digital pathology scanners (often referred to as WSI scanners) are computer imaging devices for the in vitro examination of biopsy, cytology, and tissue specimens. They are capable of all sorts of tasks – scanning, digitizing, compressing, storing, retrieving, and even enabling the viewing of high-resolution, high-magnification digital medical images. Most pathologists are aware that digital pathology has only recently come into its own (go back 10 years, and it was far less common than it is today). Histology laboratories have shown reluctance to move away from the age-old microscope and into the digital space. Why? At least in part, their hesitance was due to the absence of color fidelity regulation and image quality assurance in WSI systems.
Global transition to WSI systems has been gradual, with key regulatory bodies, such as the FDA, requiring better color fidelity alongside other factors that determine diagnostic reliability. In addition, the FDA believes that software analysis of images from WSI systems will be very important in the future and has therefore indicated that the analysis software should be treated as a separate module in the medical approval process (1). And that is only achievable, if the colors in the image being analyzed are consistent and correct from one version of the same WSI system to another, and from one vendor’s WSI systems to those of another.
But color calibration and standardization is complex. A significant degree of color variation exists between different WSI systems, and the method by which WSI systems observe color is different from the way the human eye does. All of this means that images need to be processed to interpret these differences.
The path to standardization
How can we ensure that the colors of scanned specimens are the same when digitally presented to a pathologist as when viewed down a microscope? And how can we ensure that those colors are the same regardless of the systems on which they are scanned, stored, and viewed?
To make sure WSI systems produce images with an accurate replication of the stains used, one approach the FDA suggests is color management using a unique slide-based device. The FDA’s WSI guidelines (1) state, “The WSI system should be tested with a target slide. The target slide should contain a set of measurable and representative color patches. Ideally the color patches should have similar spectral characteristics to stained tissue.” The document also describes the need to create three datasets: truth, intended color, and output color – a responsibility that I propose should lie with both the vendor (to ensure that products are accurately calibrated prior to market) and the user (to ensure that systems remain calibrated and validated for ongoing diagnostic accuracy). Essentially, this means obtaining a measure of the “true” color of a stained sample and comparing it with the output of a WSI device to identify the device’s effect on color representation through the digitization process.
Creating a “truth” dataset is critical to the whole process and requires the use of specialist equipment calibrated to internationally accepted standards. Of equal importance is the precision of the color profile this calibrated equipment generates, which is why color calibration specialists work to International Color Consortium (ICC), Commission Internationale de l’Eclairage (CIE), and ISO standards.
What difference does it actually make?
Color calibration to device-specific ICC standards takes WSI scan fidelity to another level. Figure 1 shows an alcian blue-stained tissue section as viewed down the microscope alongside three different representations of the same scanned image on the same calibrated display monitor. The first corrected image is from a scanner with only standard white balancing applied – a color correction that is likely to vary between scanner models in relation to differences in illumination source. The next is shown with only RGB standard values applied. The final image displays the tissue with the scanner-specific ICC color profile applied. The color shows clear variation in intensity across all three images. Applying the RGB profile improves the color representation of the scan; however, it can’t compare with the improvement obtained by applying the ICC color profile, which most closely matches the original tissue as viewed through a microscope.
Looking to the future, applying ICC color profiles to WSI technology will also aid in the application and validation of artificial intelligence (AI) algorithms for automated diagnostics. The algorithms are then able to create, make decisions on, and work from data files to process digital pathology images across disparate locations and scanner types. For AI algorithms to be universally applicable and medically reliable, we will need rigorous validation and quality assurance of images fed into automated diagnostics – including stringent control of color management to ensure image fidelity. The application of color management using a slide-based device has proven its potential for ubiquity across WSI platforms and offers a subtle and integrable solution to this “pathology of digitization within digital pathology.”
- N Anderson, A Badano, “Technical Performance Assessment of Digital Pathology Whole Slide Imaging Devices” (2015). Available at: bit.ly/2PUmsgQ. Accessed November 26, 2018.