Color in Focus
Color calibration paves the way to large-scale digital pathology rollout
Digital pathology or whole slide imaging (WSI) is a valid and reliable method for interpreting images at a distance, enabling expert interpretation of pathology cases in remote and/or medically underserved areas. It is also an efficient and effective means of obtaining second opinions, and it’s useful for teaching too. Although there is significant progress in the technologies used to acquire the images and the displays for viewing them, there has been less progress in developing methods for standardizing the presentation of WSI.
Radiology faced this problem when radiographic images first went digital. In a bid to solve the problem, the DICOM GSDF (Digital Imaging and Communications in Medicine Gray Scale Display Function) was developed to standardize the presentation of images so that, no matter what display was used, a given image would look the same. This has been quite successful for gray scale radiographic images, but pathology WSI (as well as other types of medical images, such as those from ophthalmology and dermatology) are inherently colored in nature and so require color not monochrome displays! Although there are methods available to calibrate color displays, there is no widely accepted or validated color calibration method for use in medical imaging.
Very few studies have assessed the impact of color display calibration on WSI interpretation, which caused us to ask the question: does color calibration actually matter? We conducted a study that developed a calibration protocol for color medical imaging applications using WSI as the application. We assessed the impact of this protocol on observer performance – pathologists viewing and providing diagnoses on a set of cases. We used 250 breast biopsy WSI regions of interest (half malignant, half benign) and showed them to six pathologists, once using the calibration protocol and once on the same display in its “native” un-calibrated state. We measured diagnostic accuracy and time to render a decision. With the calibrated versus un-calibrated display, we found that there was no significant difference in interpretation accuracy. However, when assessing speed of case interpretation, we found that this was significantly faster for the calibrated display compared with the un-calibrated display; cases were interpreted almost two seconds faster with the calibrated display.
Admittedly, there are several other factors that contribute to viewing and interpretation time of cases in clinical pathology, but if by simply using the proper calibration method you could reduce even one contributing factor, it should make a significant difference in overall efficiency and acceptance of WSI. There is also growing evidence in radiology that spending long workdays interpreting digital images increases fatigue and visual strain, which leads to increased errors. If WSI becomes more widely integrated into routine clinical practice, why wouldn’t fatigue become a problem here as well? I think it will, and so we need to start thinking about ways to avoid reader fatigue in WSI applications.
We were rather surprised that the calibration method did not have that much of an impact on diagnostic accuracy but there are potential reasons for this. We only used breast biopsy specimens with hematoxylin and eosin staining. There are numerous other stains of various colors that are used on a wide variety of tissues. We may not have selected a tissue and stain where color contributes significantly to diagnoses – other image features may be more dominant. Other tasks do require careful color display calibration so we need to conduct more studies.
WSI is clearly going to be more widely adopted, especially if the FDA clears a path for approval of novel acquisition and display devices. Some of the concerns the FDA has are related to color reproducibility and standardization. In my view, we need more studies to assess the impact of display standardization on diagnostic performance to speed this process up and get WSI into everyday routine pathology as quickly as possible!
Elizabeth A. Krupinski is professor of medical imaging at the University of Arizona, Department of Medical Imaging, Tucson, Arizona, USA.