Histopathology, which involves the microscopic examination of patient tissues for the identification of tissue abnormalities, is a largely manual process. It requires slide preparation by fixing, followed by staining for specific cell and tissue markers, and finally, visual inspection by a pathologist.
As technologies continue to evolve, vendors are looking to develop solutions to automate this intensive and time-consuming process. Automation can help move the largely qualitative field of pathology to a quantitative assessment that avoids human bias and enables the precise and reproducible extraction of data from slides. The digitization of pathology slides through whole slide imaging (WSI) represents a major step toward this automation.
In WSI, slides are prepared and stained in the same way as in conventional microscopy, but instead of examining the slide with a microscope, the slide is scanned and visualized on a computer screen.1 The user can navigate the tissue and annotate any findings using software.
While this technology has been used for slide archiving, remote consultations, and education, among other applications, the use of WSI for diagnosis in the clinical lab is still in its infancy.
Uses of whole slide imaging
Environmental factors can degrade tissue mounted on slides over time. Slides are also prone to breakage, misplacement, or mislabeling, and they take up physical space. Digital slide archives maintain the quality of the slide image over time and provide long-term storage solutions so that only tissue blocks need to be physically stored.1
The digitization of histology slides allows them to be accessed anywhere by anyone.2 Specialists around the world can be sent digital slides in minutes and examine the entire slide instead of relying on the sender to choose a representative section.
Developing and maintaining tissue sets of histology slides is challenging. First, instructors need to find high-quality specimens that are free of artifacts and produce enough representative sections. When slides are digitized, an entire class can have access to the same tissue set anywhere.2 Only a single tissue section needs to be scanned, and slide-to-slide variation is eliminated. Digitization also promotes sharing and distribution of histopathologic specimens so that resources across training programs are broader and more homogenous.
FDA approval—a major step forward for the clinical lab
In 2017, the United States Food and Drug Administration (FDA) approved the first WSI scanner for primary diagnosis in surgical pathology.3 The scanners are defined as Class III medical devices, and the FDA regulates these instruments to help ensure that images being analyzed for clinical use are safe and effective for their defined purpose.4
Before approval was conferred, the whole slide imager was thoroughly validated to show that it produced results comparable to conventional microscopy. Many studies have investigated whether there is a difference in diagnosis when pathologists use conventional microscopy versus WSI.5 These studies have shown high concordance rates among these two imaging types; however, study participants found that WSI was too slow for routine use when examining slides and that digital images were more difficult to evaluate than were glass slides.6,7
It has been shown that WSI performs just as well as does conventional microscopy. However, there are several challenges that are discouraging clinical labs from adopting this technology.
Standardization and data management
Vendors of WSI platforms use proprietary formats to store image data and metadata, which makes it challenging to organize images acquired from a different scanner. As a result, labs are limited to using a single type of scanner when performing WSI, which can impact interoperability and scalability. One solution is Digital Imaging and Communications in Medicine, which is an international set of standard file formats and communication protocols that provide a vendor-neutral and universal language for medical imaging.1
Another major issue is the amount of data created when WSI is used. If the average image is between 200 MB and 1 GB, and the average number of slides per surgical case is 12.2, then anywhere from 2.4 to 19.5 GB of storage will be required for each case.8 The costs can add up when storing this much data, but discarding images can be almost as costly, especially if the glass slides aren’t being retained. One solution might be to have pathologists flag images that contain information important for diagnostics and discard the remaining slides from each case.8
There are many factors that can impact image quality when performing WSI. Pre-analytical variables during slide preparation, such as tissue procurement, fixation time, fixation type, and antigen retrieval protocol, all need to be standardized to ensure consistency.2 A single slide scanner can also produce different quality of images of a single slide, depending on external factors such as temperature and mechanical shifts.7 Finally, slide scanners aren’t standardized, so image quality from one scanner to the next could differ, making the images obtained incomparable.2
As technology advances, there is potential for WSI to be used for more detailed and complex analysis using fully automated processes.
WSI remains a 2D imaging technique, and because of this, it leaves a gap between recorded slide observations and the original state of the tissue.1 Stereology is the study of 3D representations generated from a random sampling of 2D images.9
3D imaging would allow for the evaluation of entire tissues rather than single representative samples.2 These images could also be compared with other diagnostic images such as those from magnetic resonance imaging and conventional computer tomography to identify diagnostic patterns.1 However, stereology requires a significant amount of time and tissue as well as a skilled stereologist and a specially trained histologist to correctly prepare the samples. For this reason, pathologists continue to argue about whether the benefits are worth the effort.2
Tools are being developed to extract information from tissue slides to avoid the error-prone and repetitive nature of manual assessment. The basic principle involves a mathematical algorithm that can process images and segment picture elements into regions of interest based on color, texture, and context.2
Commercial software for image analysis differs in the amount of supervision required or allowed.2 Unsupervised software is pre-built and easy to use right out of the box, whereas supervised software allows the user to program specific algorithms to develop unique analyses and requires extensive training. Either way, a pathologist needs to be involved to design the study, determine what the biological end point will be, define the regions of interest, and evaluate whether the software is correctly measuring that end point.
As pathologists know, abnormalities exist on a spectrum rather than in discrete groups. The advantage of image analysis compared with manual analysis is that it can measure variables as continuous rather than ordinal, and these values can then be analyzed using statistics.2 This evaluation can help decrease observer variability and increase reproducibility.
Artificial intelligence and machine learning
The use of artificial intelligence (AI) and machine learning in the evaluation of digitized slide images is still far from reality. In theory, artificial intelligence will allow for the discovery of patterns in tissue images that can be used to derive insights and make predictions. One day, this technology might be used for computer-aided diagnosis.1
On a more basic level, AI could relieve pathologists of mundane tasks and simplify more complex tasks. It could also analyze individual pixels of images more deeply, unlocking diagnostic information that might not be available when slides are visually inspected with a microscope.10
While there are clear opportunities for AI in pathology, there are many challenges that must be overcome before this technology can be implemented.10 For example, pathologists still need to be heavily involved in manually delineating the region of interest in images before automated analysis is conducted. The variability found in tissue can also challenge AI because the number of patterns that the software would need to identify in tissue could be nearly infinite.10
WSI has the potential to revolutionize the field of pathology, but it will in no way eliminate pathologists. These experts will continue to play an important role in slide and image quality assurance, labeling slides and selecting regions of interest, and evaluating algorithm performance. Nonetheless, use of this technology could transform pathology from a largely qualitative field to one that is data-driven and relies on quantitative rather than qualitative analysis.
1. Zarella, Mark D., et al. "A practical guide to whole slide imaging: a white paper from the digital pathology association." Archives of Pathology & Laboratory Medicine 143.2 (2018): 222-234.
2. Webster, J. D., and R. W. Dunstan. "Whole-slide imaging and automated image analysis: considerations and opportunities in the practice of pathology." Veterinary Pathology 51.1 (2014): 211-223.
3. Evans, Andrew J., et al. "US Food and Drug Administration approval of whole slide imaging for primary diagnosis: A key milestone is reached and new questions are raised." Archives of Pathology & Laboratory Medicine 142.11 (2018): 1383-1387.
4. FDA. "Technical Performance Assessment of Digital Pathology Whole Slide Imaging Devices." (2016) https://www.fda.gov/regulatory-information/search-fda-guidance-documents/technical-performance-assessment-digital-pathology-whole-slide-imaging-devices.
5. Goacher, Edward, et al. "The diagnostic concordance of whole slide imaging and light microscopy: a systematic review." Archives of Pathology & Laboratory Medicine 141.1 (2016): 151-161.
6. Onega, Tracy, et al. "Use of digital whole slide imaging in dermatopathology." Journal of Digital Imaging 29.2 (2016): 243-253.
7. Jayakumar, Rajeswari, et al. "Can whole slide imaging replace conventional microscopic evaluation? A comparative study over a spectrum of cases." Journal of Applied Clinical Pathology (2018): 4.
8. Balis, Ulysses G. J., et al. "Whole-slide imaging: thinking twice before hitting the delete key." AJSP: Reviews & Reports 23.6 (2018): 249-250.
9. Aeffner, Famke, et al. "Introduction to digital image analysis in whole-slide imaging: A white paper from the Digital Pathology Association." Journal of Pathology Informatics 10 (2019).
10. Tizhoosh, Hamid Reza, and Liron Pantanowitz. "Artificial intelligence and digital pathology: Challenges and opportunities." Journal of Pathology Informatics 9 (2018):38.