Comment on “Quality evaluation of microscopy and scanned histological images for diagnostic purposes”: Are scanners better than microscopes?

Abstract

There have been animated discussions in the USA, at least in the popular press[1] and online,[2] on whole-slide imaging (WSI) and the Food and Drug Administration's (FDA) view that WSI systems are to be considered class III (highest risk) medical devices with plans to regulate them as such. In order for class III devices (e.g. implantable pacemaker, gynecologic cytology imaging systems) to be approved they require general controls (such as quality system regulation and good manufacturing procedures) and premarket approval.[3] By comparison, microscopes are class I devices. Concerns have been raised that this stance may hamper current efforts to employ WSI for primary diagnosis in USA laboratories in lieu of the optical (conventional) microscope. Several publications have reported that some attributes of digital images may perhaps be better than the “gold standard” tool for microscopic diagnosis in pathology - a microscope – such as portability, archiving, sharing, and performing image analysis.[4] However, what if studies show that digital images acquired from scanners are in fact of better “quality” than those obtained from a microscope? Will this provide reassurance that WSI is suitable for rendering primary diagnoses? We share the thought-provoking findings and our commentary on the recent article published by Redondo and colleagues from Spain who compared the quality of images captured with optical microscopes and scanners.[5] In their article, Redondo et al. compared whether images captured with optical microscopes and scanners provided better image quality in terms of contrast, color, and stain.[5] They digitized biopsies (breast, prostate) and cytology (effusion) slides prepared with different stains (HandE, immunostains) using a robotic microscope (ALIAS II) and whole slide scanner (Aperio Scanscope XT). All images were acquired at 40× magnification in bright field without compression. The authors evaluated images using quantitative, objective, and subjective means. Quantitative evaluations employed several perceptual features such as color decomposition, intensity adaptation (dependent on luminance), and contrast, among others. Subjective (perception) testing involved six expert observers who scored microscopy and scanned images. These investigators only evaluated a limited number of images (image fields) for their study when one considers that a digitized slide is made up of more than 1,000 images. Nevertheless, the results from this study indicate a slight preference for the WSI scanner over the microscope in terms of better image quality. The main cause for subjective discrepancies between the microscope and scanner was related to focusing, with color distortions indicated to be a secondary problem. However, even though images acquired with scanners triumphed, these authors do stress that the image quality of both devices is suitable for clinical, educational, and research purposes. Image quality in WSI is an important aspect of digital pathology, but one that has been remarkably difficult to define.[6–8] As WSI becomes increasingly popular and used in clinical practice and/or automated image analysis, such issues are becoming more serious. As the authors state “a good image should preserve those features which facilitate the pathological assessment of the tissues, i.e. whether the tissue is diseased or not.”[5] While image quality primarily involves image capture, it can also be modified during postcapture processing as well as by the display. Moreover, ambient light in the pathologist's office too can affect the perception of an image. In this study, glass slides were digitized at 40× magnification without compression. However, many WSI systems in use today may scan slides at various magnifications not just 40× (0.23-0.25 μm/pixel) and often employ compression. Future studies will need to explore the impact of image metrics such as compression on WSI in greater detail. Some investigators to date have shown that image compression of static images does not appear to negatively impact image interpretation[9] nor image analysis.[10,11] The finding reported by Redondo et al. showing a preference for the WSI scanner is interesting, given that some people believe that the microscope still represents the “gold standard” tool for microscopic diagnosis in pathology. The original concept for WSI scanners came from a fully automated robotic microscope. Therefore, Redondo's conclusion that the image quality of scanners is better than motorized microscopes is encouraging. Theoretically, a well-maintained automated robotic microscope imaging system could produce an optimized image by permitting the user to control all components such as the optics, focusing, and camera in real time. On the other hand, while a WSI scanner's adapted line sensor is capable of producing a digital slide very fast, it may not necessarily be as good at offering precise control of each component compared to the robotic microscope. However, WSI scanners are relatively new and continue to evolve, such as added functionality related to remote telepathology. When we scan one slide multiple times with one scanner, we do not yet always obtain exactly the same image. If we want to use WSI for primary diagnosis and/or decision support systems, we may need to incorporate some kind of mechanism to quickly and easily evaluate the entire WSI of each slide. Moreover, it may not be enough to just evaluate the image itself, but also the entire imaging process. Nevertheless, while further work in this area is needed, the article by Redondo et al provides hope in supporting the notion that a digital virtual slide created by a WSI scanner can adequately simulate a glass slide.