This is the first in a series of articles discussing issues with color imaging.

Researchers are facing increasing demands from colleagues, peers and publishers for process documentation including adequate controls, and for extensive documentation of experimental parameters. Without such consideration, there would be little chance to repeat, or even validate, findings.

Consider microscope images.  Images are visual representations of experimental outcomes and, therefore, should be considered and treated as data. In today’s research climate, scientists should anticipate and be prepared to accommodate heightened demands for scientific integrity by providing enhanced rigor in the capture, processing and communication of images.

Image adjustment

Researchers commonly adjust microscope images by adding contrast, fine-tuning gamma to help visualize some detail, or adjusting color and background to facilitate image comparisons, such as in figure plates. To be consistent with the scientific method, any changes to the original image should be documented and communicated with the image – journal editors often request the original image to ensure integrity with the submitted image.

Color is a particularly complex aspect of microscope imaging and is among its least regulated and understood.  Fidelity and consistency in color are vitally important for the accurate evaluation of results. The repeatability of images is further affected by the specimen, illumination, instrument, software, process, camera, viewing technology and technician. Ideally, color variation in stained specimens should accurately reflect the consequence of the experiment (i.e. be a response to an experimental variable). Too often, however, the variability is due to inconsistent imaging conditions. 

Color shift can result from halogen light source voltage differences. Scientific CCD camera was white balanced at the middle frame (8 volts), and auto-exposure was enabled to ensure the same overall intensity in the images. (Source: Datacolor)Variables in imaging

Today’s microscope systems, cameras and software are more sophisticated than ever, but there are a plethora of variables to account for— from light sources and intensities to optics and filters— and no two camera manufacturers use the same “recipe” for color adjustment.  Further, there is no single industry standard in place for the implementation of white balance, contrast, exposure control and color correction. The result is often-inconsistent color when comparing images from different cameras or microscope systems. Variances are also seen from imaging session to imaging session. In many cases, the color in a specimen image changes just by clicking the white balance button a second or third time. Further, a general lack of standardization across cameras, software and the language they use (i.e. exposure vs. shutter), leads to variation and confusion regarding imaging. It’s no wonder many researchers lack a comprehensive understanding of exactly what the software and camera were doing during the acquisition process, which makes it more challenging to compare images or replicate the experimental results.

Microscope light sources impart color in the final image, which can vary by technology. Tungsten (halogen) lamps are known to fluctuate in color temperature with changes in intensity (voltage), ranging from very yellow to almost blue (Figure 1). Balancing filters, such as Kodak 80A, or daylight filters help maintain color temperature; however, these are not universally used. Even when they are employed, color can still shift with changes in light-source voltage. Light-emitting diode (LED) light sources are becoming more common but also vary in color generation from manufacturer to manufacturer.

Despite all of the advances in imaging technology, we are still faced with the conundrum that our images are not readily comparable between microscopes and labs, or between imaging sessions on the same microscope.

Color correction via software

Color correction has become a necessity when trying to compare or analyze images, and best practices dictate that images be captured without any processing. For example, white balancing is acceptable, but color and contrast enhancements are not.  Most acquisition software does not support post-acquisition image correction.  As a result, scientists have turned to image-editing software to address the need for making adjustments to previously captured images, and the use of such tools has become an accepted practice.

However, there are few controls or limits in place when using image editing software for scientific images. Photoshop® is a popular choice for adjusting hue and saturation, white balancing and even bringing multiple images into uniformity in color and brightness for comparative purposes (i.e. preparing figures for reports). Unfortunately, not only is it possible to alter the original images without recording what was done in the process, but the process is also subjective (“by eye” adjustments), time consuming and lacking scientific rigor. Increasingly, publishers, editors and reviewers request original image files to compare against submitted images for evidence of alterations. In extreme cases, image forensics may even be applied to scrutinize images.

Color correction can help ensure consistency and accuracy of color data. Column A images are originals; Column B contains images after color-calibration using Datacolor ChromaCal. (Source: Datacolor)Datacolor recently introduced ChromaCal, a color calibration solution for brightfield microscope images that ensures comparability of color and brightness among images without the subjective adjustments that are common when using image-editing software (Figure 2). The concept is very straightforward.  An image of the calibration slide represents the “color fingerprint” of the imaging system.  The software generates a corrected color profile based on a comparison of the fingerprint and the known measurements of the calibration slide.  The profile is then applied to a copy of the original TIFF or JPEG specimen images, in single or batch mode. Details of the calibration are stored in the metadata of the calibrated images. The system also automatically white balances and adjusts brightness levels of the specimen images before calibration, resulting in more comparable color and brightness in calibrated images – exactly what is needed for figure preparation in reports and publications.

One current ChromaCal user is a toxicologic pathologist who relies on image color for assessing a drug’s toxicologic profile. It is crucial that both color balance and overall image brightness be comparable from image to image in order to distinguish the subtle differences among specimens and experimental conditions. Inclusion of the calibration profile in the image metadata adds confidence that the images have not been subjectively altered, rather objectively calibrated.

Color imaging technology has made tremendous advancements, yet since color reproduction falls short of expectations, after-market imaging editing tools will continue to be utilized. The adoption of standards in imaging instrumentation and software would allow researchers to better understand what happens during the imaging process, account for those settings and apply them across platforms. Even now, as publishers, editors, reviewers and scientists consider the implications of recent high-profile article retractions, researchers must be prepared for the enhanced scrutiny of scientific images and findings. Objective calibration of the color in images may be a good place to start.

Next month we’ll discuss the implication of color management and the interpretation of experimental results.

Photoshop is a trademark or registered trademark of Adobe Systems Inc. in the United States and/or other countries.

ChromaCal is a trademark of Datacolor Inc. in the United States and/or other countries.