Toxicologic Pathology Forum*: Opinion on Considerations for the Use of Whole Slide Images in GLP Pathology Peer Review

Abstract

Whole slide imaging (WSI) technology has advanced to a point where it has replaced the glass slide as the primary means of pathology evaluation within many areas of medical pathology. The deployment of WSI in the field of toxicologic pathology has been delayed by a lack of clarity around the degree of validation required for its use on Good Laboratory Practice (GLP) studies. The current opinion piece attempts to provide a high-level overview of WSI technology to include basic methodology, advantages and disadvantages over a conventional microscope, validation status of WSI scanners, and perceived concerns over regulatory acceptance for the use of WSI for (GLP) peer review in the field of toxicologic pathology. Observations are based on the extensive use by AstraZeneca of WSI for the peer review of non-GLP studies conducted at Charles River facilities and represent the experiences of the authors.

Note: This is an opinion article submitted to the Toxicologic Pathology Forum. It represents the views of the author(s). It does not constitute an official position of the Society of Toxicologic Pathology, British Society of Toxicological Pathology, or European Society of Toxicologic Pathology, and the views expressed might not reflect the best practices recommended by these Societies. This article should not be construed to represent the policies, positions, or opinions of their respective organizations, employers, or regulatory agencies.

The pathology peer review should be conducted by a pathologist with appropriate training and experience. The peer review pathologist may work for a contract research organization, the sponsor, or a third party. The sponsor’s peer review pathologist may bring extensive knowledge to the peer review process that can improve the quality of the pathology data and interpretation. The study pathologist is ultimately responsible for the interpretation of pathology study data and for placing these data into context so that, in collaboration with the study director, the correct conclusions are drawn for the study. The Organization for Economic Co-operation and Development (OECD) is the only guidance that requires a formal peer review (carcinogenicity studies), but any peer review undertaken on a Good Laboratory Practice (GLP) study must be documented on the study protocol and should be GLP compliant or covered by an exception statement. Current guidance (OECD Guidance Document 16, 2014) does not explicitly dictate that peer review must be undertaken using glass slides.

Whole slide imaging (WSI) technology has evolved to a point where digital slide scanners are currently capable of producing high-resolution digital images with a relatively high throughput. WSI is used extensively in the medical diagnostic field worldwide, often for the primary diagnosis of surgical biopsy and postmortem specimens. While digital and telepathology guidelines have been drafted by several medical associations such as The Royal College of Pathologists (Cross et al. 2017; Lowe 2013) and The College of American Pathologists (CAP; Hipp et al. 2017), detailed guidelines for the use of WSI in the context of toxicological pathology are lacking. Long et al. (2013) provided an overview of the necessary steps for the validation of a digital pathology system including WSI, which was endorsed by the Scientific and Regulatory Policy Committee of the Society of Toxicologic Pathologists (STP). This article was focused on the technical aspects relating to the validation of a digital pathology system in the nonclinical environment rather than its context of use. While Malarkey et al. (2015) undertook a comparison of the use of WSI versus glass slides for reviewing histopathology findings in rodent studies run at the U.S. National Toxicology Program, their paper concentrated on the level of “diagnostic agreement” between the two approaches and did not provide details concerning any validation of the methodology employed (slide scanner, viewing software, etc.). The current article aims to provide an overview of the main considerations for the use of WSI for GLP pathology peer review based on AstraZeneca’s extensive use of WSI for pathology peer review of non-GLP studies run at Charles River Laboratories facilities and discussions with individuals experienced in quality assurance and GLP compliance.

Definitions

Digital pathology

The use of digital images for the diagnosis, review, and demonstration of pathology findings. Diagnostic digital pathology refers to the use of WSI for diagnosis, education, and research that will facilitate the development of pathology networks to assist pathologists in their work (Cross et al. 2017).

WSI

The process of obtaining a digital image from an optical image of a histopathology glass slide.

Slide scanner

It consists of a combination of hardware to move the slide and acquire the digital image and software to stitch and compress the image into a usable format for the viewing software.

Image chain

Image acquisition device (scanner), method of accessing the images, and workstation (viewing software, PC, monitor, tablet).

Telepathology

Electronic transmission of digital images from one location to another for the purpose of interpretation and diagnosis.

Z-stacking

Images are gathered at several different focal planes at every focus point and the image focus can be shifted by switching between the layers.

Method

Pathologists should be aware that a digital image is a high-resolution copy of a tissue section on a glass slide, but it is not an exact replica in every way. The slide scanner is the product of an engineering process in which the design of the hardware and software will affect the nature of the information that is captured and presented to the pathologist. Pathologists should be aware that the same tissue sample may look different when imaged/scanned with different instruments, viewed on different displays, or even assessed with different software viewers. Digital pathology therefore consists of an end-to-end imaging chain. For GLP use, all hardware used in the imaging chain will need a fit for purpose validation, and its use should be standard operating procedure (SOP) driven to ensure consistency.

Resolution has a major influence on image quality. In terms of digital images, spatial resolution refers to the number of pixels utilized to construct the image. The spatial resolution of a digital image is related to the spatial density of the image, and optical resolution of the lens used to capture the image. The number of pixels contained in a digital image and the distance between each pixel is known as the sampling interval, which reflects the accuracy of the digitizing device. Many of the scanners currently used (Leica Aperio, Hamamatsu NanoZoomer, Philips IntelliSite) support capture resolutions of 0.5 μ/pixel (effective viewing magnification of 20×) or 0.275 μ/pixel (effective viewing magnification of 40×). The resolution of the acquired whole slide image is therefore affected by several stages in the optical imaging chain, including the magnification of the objective lens used, the camera chip specifications, and other electronic components inside the device. The scan quality also depends on the scanner’s optics. The numerical aperture (NA) of the lens relates to its property to resolve fine detail. Higher NA values have better resolution but provide less depth of field. Most scanners utilize a proprietary microscope lens designed to work best with the optical setup of the scanner. For some scanners, a higher magnification scan may only utilize an inferior quality lens that slides in front of the microscope objective. This can produce an inferior digital image when compared with utilizing a higher magnification microscope objective. Toxicologic pathologists will generally use a low-magnification objective to screen the tissues. At the lowest magnification, whole slide images contain sufficient detail to replicate this approach. Some software viewers allow a “digital zoom” which increases the apparent magnification on screen. While this feature can be useful to provide a small window of magnification within an image, no additional information is being added to the image during a digital zoom. When training with a WSI system, pathologists should compare the original glass slide to the WSI at varying magnifications to familiarize themselves with the similarities and differences between the two. Careful consideration also needs to be given to the number of focus points required to produce a fit for purpose scan. Heterogeneous specimens such as bone marrow smears and lung may require a larger number of focus points compared to more homogeneous tissues such as liver. Bone marrow smears and blood films may also require manual verification of the focus points to ensure a consistent focus across the slide. Standard scanner configurations generally only cater for 40× magnification, which may not be sufficient for critical evaluation of bone marrow smears/aspirates/cytospins, and blood films. The use of oil immersion objectives generally requires a dedicated lens and slide tray and scanning time can be prolonged due to the effort required to set up the parameters for each slide. The authors are not aware of WSI being routinely used for the evaluation of blood and bone marrow smears derived from routine toxicology studies, but their use in the medical field has been documented (Chen et al. 2014).

There is a delicate balance to be struck between the number of focus points and the speed of scanning, to maintain an adequate throughput. The more focus points generated, the slower the scanning process. At Charles River facilities, 20× scans are used for routine non-GLP peer review, while 40× scans are used when additional resolution is required, or when a higher magnification is needed to capture a field of view for illustrative purposes.

A scanner that enables Z-stacking will produce an image that allows the pathologist to track through the focal planes of the image akin to the fine focus on a microscope. This is especially useful for techniques such as fluorescence in situ hybridization, where the signal is not always located in the same focal plane. For routine peer review of hematoxylin and eosin (H&E) stained slides, Z-stacking is not essential on a well-scanned high-resolution image as it can also add considerably to the image file size and time to scan.

Most WSI devices capture contiguous images from the glass slide either as tiles or as stripes, by moving the objective lens/imaging system relative to the glass slide. These subimages are then stitched together to create the whole slide image. The tile technique assembles image tiles and stitching algorithms are used to put the images together. In linear scanning, no stitching is used, and there is continuous image acquisition in real time. This requires a separate focus run to find the right focal plane. Pathologists should be aware that where stitching is used, it can produce artifacts such as visible striping or misalignments across tiles. In the worst case, misalignment could result in small areas of tissue being hidden. During assessment and use of WSI devices, pathologists should ensure stitch artifacts are minimized.

The final image quality also depends on the resolution of the monitor and graphics card. The colors displayed on any viewing device should be vivid with good white balance capabilities and the display should be glare free. Pathologists should be aware that ambient lighting and reflections can also affect the performance of a display.

Pathologists must also recognize the potential for differences in color reproduction between scanning devices. In procuring devices, pathologists may wish to include an assessment of color accuracy in their testing. Experience suggests that many standard histochemical stains and immunohistochemical stains using 3,3′-diaminobenzidine (DAB) reagents are most often affected by color differences in WSI.

Most scanners cater for standard glass slides (75 × 25 mm), but separate scanning trays can be acquired for the larger slide sizes. Horizontal stages allow for scanning of frozen sections, while vertical stages can also be used as long as the mounting medium prevents coverslip movement. Vertical stages have a smaller footprint and tend to have more reliable loading mechanisms.

Benefits of WSI

What are the potential benefits of scans over a conventional microscope? Whole slide images are detailed and high contrast and retain this detail when resized to a low magnification. The use of scans also enables two or more images to be compared side by side, allowing for the comparison of test article and control groups, or between an HE and a special stain.

Whole slide images provide a greater field of view at the corresponding magnification, often allowing for the whole tissue section to be viewed at lower magnifications. The large field of view can facilitate the assessment of structural information on the scanned image, for example, nasal turbinate atrophy in an inhalation study. Most viewing software allows a considerable degree of image rotation that is difficult to replicate with a microscope stage. This allows for an easier comparison of areas of interest, for example, when assessing changes such as centrilobular hypertrophy in the liver.

Being digital, WSI images can be easily shared. Images can be encrypted onto portable media, with password access to the encrypted drive provided via secure portals. They can also be accessed remotely via secure servers, with the potential for encryption in transit and at rest, to provide added levels of security. With the proper security protocols in place, this makes the sharing and retrieving of images much easier. Digital images obtained from WSI can also be included in pathology reports and scientific publications without the need to obtain bespoke images prior to slide archiving. Illustrative images are generally not considered to be raw data because they are not used for data generation (Tuomari et al. 2007).

WSI also provides a much easier means for obtaining second opinions, as WSI images can be shared real time using many proprietary applications such as Webex and Skype. This removes the need to ship glass slides, where the movement of some histopathology material is subject to additional control such as Convention on International Trade in Endangered Species permits or farm animal movement permits (e.g., with pig tissue into the United States and rabbit tissue into Canada).

The portability of scans also provides sponsors with the potential to undertake peer reviews remotely, providing cost avoidance for travel, and any associated disruption to work routines. AstraZeneca currently utilizes WSI to undertake all non-GLP peer reviews for studies placed with Charles River Laboratories. The use of whole slide images also allows sponsors to build up an easily accessible database of project-related toxicology studies, allowing for easier project transfer between individuals, and a valuable resource for training and education.

Disadvantages of WSI

Viewing large numbers of scans will require the pathologist to move and scroll the images using some form of handheld device (mouse, trackpad, or possibly a touch screen), particularly when screening tissues at low magnification. A typical dose range finding study in a rodent may contain several hundred scanned images, requiring careful consideration for workstation ergonomics.

Slide scanners tend to create numerical file names that will generally not align with the information printed on the slide label. Database software for storing and cataloging WSI (e.g., Aperio E-slide manager) often allows for the image labels to be displayed as a thumbnail but is frequently restricted by the amount of time it can take for images to load and render, given that databases will generally be hosted on a networked server. When images are stored on portable media, some older operating systems will not display WSI files as thumbnails. This can make it more challenging for a pathologist to pick the desired animal or tissue from a list of numerical file names (although the slide label will be displayed once the image has been loaded by the viewer). Some WSI databases allow for the export of limited metadata relating to fields such as the animal number, tissue, and so on, in the form of Microsoft Excel or comma separated values files that can provide filterable spreadsheets as a means of aligning image file names with animal numbers or tissues. The adoption of bar coding for routine use in toxicology studies may provide a solution for the easy transfer of metadata in the future. AstraZeneca and CRL are currently collaborating with a software developer to produce a bespoke slide viewing interface that will facilitate ease of peer review by displaying image thumbnails and allowing multiple relevant fields (e.g., animal, dose group, sex) to be selected. The time taken for slide scans to open and render when using a remote server can significantly hamper ease of viewing, and for this reason, AstraZeneca currently utilizes portable media when undertaking digital peer reviews.

There is also a potential uncertainty over the longevity of archived WSI images in terms of whether future image viewing software platforms will remain compatible with legacy scan file formats. However, in the context of peer review, the archived glass slides and or wax blocks will remain the raw data, unless regulatory authorities will not accept this approach.

Validation Status of WSI Scanners

The U.S. Food and Drug Administration (FDA) considers WSI scanners to be medical devices. With respect to this, the U.S. FDA (Center for Devices and Radiological Health) recently permitted marketing of the Philips IntelliSite Pathology Solution (PIPS) WSI for review and interpretation of digital surgical pathology slides prepared from formalin-fixed paraffin-embedded biopsies (FDA News Release 2017). This is the first time the FDA has permitted the marketing of a WSI system for this purpose. The FDA reviewed the data for the PIPS through the de novo premarket review pathway, a regulatory pathway for devices of a new type with low to moderate risk that are not substantially equivalent to an already legally marketed device. The FDA evaluated data from a clinical study of approximately 2,000 human surgical pathology cases using tissue from multiple parts of the body (anatomic sites). Results of the study found that clinical interpretations (diagnoses) made based on the PIPS images were comparable to those made using glass slides. In this authorization, the FDA outlined “special controls” that must be met to assure the digital imaging system’s precision, reliability, and clinical relevance. These included a detailed description of the device, and bench testing results at the component level, to include the following: slide feeder, light source, imaging optics, mechanical scanner movement, digital imaging sensor, image processing software, image composition techniques, image file formats, image review manipulation software, computer environment, and display system. They also required bench testing and results at the system level for color reproducibility, spatial resolution, focusing test, whole slide tissue coverage, stitching error, and turnaround time (FDA letter dated April 13, 2017). These requirements are in line with the original guidance FDA issued on the technical performance and assessment of WSI devices (FDA Guidance for Industry 2016). This document comprehensively describes test methods for each component of the imaging chain and is aimed at the scanner manufacturer, not the end user. Sections IV(A)(1) to (11) describe tests that should be carried out by the manufacturer of the device. Sections IV(B)(1) to (6), IV(C)(1) to (2) should be repeated at the test facility as part of the end user acceptance test. While the FDA regulates the device manufacturer, they do not regulate the medical professional’s use of the device. They may however require line of sight to the results of the bench and system testing outlined above.

Regulatory Acceptance for the Use of WSI for GLP Peer Review

There are no regulatory authority guidelines for the technical specifications of scanners for nonclinical use, but it is anticipated that they would be similar to the bench and system testing expected for clinical use. Given the complexity of these tests, it would seem logical that the results of any such validations will be provided by the device manufacturer and would be used alongside any “imaging chain” or system validation conducted by the test facility and or end user. For a GLP facility, the scanner will need to be maintained and operated under the GLP footprint and its use driven by documented SOPs. However, some debate remains as to the nature and level of any additional validation required to incorporate WSI for use in GLP peer review (i.e., validation beyond that supplied by the manufacturer). There is also debate about what constitutes pathology raw data, both in the context of primary reads and peer review, particularly with reference to potential differences in guideline requirements (FDA, U.S. Environmental Protection Agency, and OECD, Bennet et al. 2018; Settiagounder 2017a, b). In the context of peer review, the authors would consider WSI as certified copies of the glass slides and, as such, do not constitute raw data. The original glass slides would be available for examination, should the peer review process need to be reconstructed in the future. However, notes or conversations that resulted from the review of the WSI could be archived with other study material. On the basis of this, it is anticipated that any validation for using WSI for peer review would need to be less stringent than for primary pathology reads. Whether this position will be accepted by regulators is currently uncertain.

The CAP guidelines were an early attempt to provide information for professionals about validation and implementation of digital pathology (Pantanowitz et al. 2013). The CAP validation procedure mixes aspects of laboratory quality assurance procedures (e.g., comparing standard light microscopy with digital pathology for a validation set of cases). An important point they make is the need to consider the whole system (from scanner to pathologist workstation = “imaging chain”) in the validation/verification process. The CAP guidance focused on using WSI for primary diagnosis, as opposed to peer review, but supports the notion of validating the “image chain.”

Initial consultation with nonpathologists experienced in GLP quality assurance (AstraZeneca and CRL quality assurance personnel) suggests that a fit for purpose validation of the “imaging chain” might suffice for the utilization of WSI for GLP peer review. This opinion has yet to be scrutinized or agreed by any external GLP inspectorate. There are several key themes (areas of concern) that emerge: The requirement for some type of formal assessment of image quality; whether WSI represents raw data or a certified copy of the raw data; the ability to tag the image to the original glass slide; ensuring all tissue in the block has been scanned; the potential to alter or corrupt the image files during acquisition, transfer, or viewing; and under who’s GLP umbrella would pathologists work, if they view slides at a test site that has left a GLP compliance program.

The acceptance by FDA of the Philips IntelliSite Pathology Solution signals that U.S. regulators acknowledge that WSI can be used to both review and interpret digital images of surgical pathology slides. The main challenge for toxicologic pathologists is persuading those tasked with assuring GLP/regulatory compliance (both internal and external stakeholders), that WSI can replace glass slides for the purposes of pathology peer review. One way to approach this is to divide the components of the imaging chain into a sequence of steps and decide what level of validation, if any, is required for each. Some areas of concern, such as image quality, can be affected by multiple steps in the imaging chain and could be addressed by utilizing an “end to end” and/or “study by study” approach.

Within the themes highlighted above, it is image quality that appears a significant concern for using WSI for peer review. This partly stems from nonpathologists unfamiliarity with WSI and perceived quality issues of the images that are produced. There is also a potential lack of appreciation that viewing of WSI can be approached in a similar manner to reviewing a glass slide.

One key aspect of image quality relates to color fidelity (color accuracy and consistency). Since differences in color can play a key role in the diagnosis of histopathology changes in H&E slides, demonstrating accurate color fidelity will be an essential element in the acceptance of WSI in toxicologic pathology. Color fidelity will be governed both by the scanner and display device the image is viewed on. It is assumed that any color variation inherent in the histological processing and staining of tissues to produce the glass slide will be covered by the existing GLP framework of the histology laboratory. Studies have confirmed the potential for differences in color fidelity either produced over time with the same scanner or variation of color fidelity when comparing the same slide scanned using different scanners (Gray et al. 2015). As a potential solution to this, the use of color calibration slides has been advocated (Bautista, Hashimoto, and Yagi 2014; Badano et al. 2014) with a range of colors tailored to those dominant in H&E-stained sections. While these allow for accurate color calibration of the scanner, many are bespoke and lack commercial availability. They have also not proved particularly durable. Scanner manufacturers generally supply simple color calibration slides for the purpose of routine scanner calibration, and these can be incorporated into any routine calibration activities. A simple solution to ensure a fit for purpose color reproduction could be the use of a tissue microarray containing several tissue cores affected by different underlying pathologies. Two slides from the microarray would be cut with the tissues from the study being peer reviewed. The original glass slides from the microarray and equivalent scans are then supplied to the peer review pathologist. Following calibration of the viewing device, the pathologist could compare the colors present in the microarray slide with those of its equivalent digital image. This would provide a simple method to achieve an “end to end” color calibration and on a “study by study” basis. It would, however, require the presence of a light microscope at the peer-review site.

Like color fidelity, the resolution of WSI also appears to be a key concern—that is, would reviewing a WSI allow the peer review pathologist to observe subtle changes in the tissues. The tissue microarray approach outlined earlier might also provide a means to assess resolution if the microarray contained a tissue core affected by a subtle finding such as single cell necrosis or pigment accumulation. This approach would allow a real-time comparison of the original glass slide and equivalent WSI to ensure the resolution of the scan allows for an accurate diagnosis. From a practical standpoint, the generic issue of image resolution should be addressed by the validation documentation provided by the scanner manufacturer, as this should have been assessed as part of the scanner validation process.

Another area of concern is the ability to tag the WSI to the slide label on the glass slide it originated from. In addition, being able to confirm that all the tissue contained within the block has been included in the scan is also a prerequisite. Most scanners produce metadata that sits alongside the digital image. This allows the viewing software to display thumbnails of the slide label, a macro image of the whole slide to assess the extent of the tissue present, and a variety of other data relating to the image capture. There will probably need to be some assessment of the robustness of this metadata, in terms of the potential to corrupt, or alter it during image capture, “transport” and viewing. The provision of such metadata alongside the digital image may allow it to be used as part of an audit trail, or chain of custody, particularly if the metadata includes some form of time stamping relating to the date of image acquisition.

Given the complexity of the technology behind WSI acquisition and viewing, what might a fit for purpose validation of the “imaging chain” entail? This is probably easiest to conceptualize when broken down into the main components of the process, beginning with the scanner. As outlined earlier, the technical validation of the slide scanner should be covered by the documentation provided by the manufacturer. The scanner should not be modified from production specification, maintained under the GLP footprint at the test facility, and its use controlled by an appropriate SOP. Such an SOP should cover areas such as user access, details of the method and timings of routine calibration and servicing, a standard set of parameters relating to the scanner acquisition settings (e.g., focus points), and details of the image file type, file location, and convention for naming files. In addition to the latter, an extra layer of robustness could be provided by a simple user acceptance test that covered the use of the scanner within the context of the test facility. This would document that the scanner performed as expected when used in accordance with the SOP referencing its use.

The next step in the imaging chain would cover how the scans were made available to the peer review pathologist. This would be either via remote access using a secure server or via the transfer of images on portable media. For portable media, files can easily be encrypted and access to the files password protected to provide an adequate level of security. For the purposes of peer review, it is not anticipated that assessment of file integrity with approaches such as checksum would be required.

The final step in the imaging chain involves the viewing of the images using a proprietary viewer and display device. Guidance regarding the specification of display devices for viewing WSI is lacking and results of trials using different monitor sizes and resolution conflicting. Cucoranu et al. (2014) showed that a smaller, lower resolution monitor was associated with a greater confidence in identifying morphological features. Other studies found no difference (Randell et al. 2014). Until consensus is reached among the medical community, monitor size and screen resolution should probably be left up to the discretion of the pathologist. Monitor refresh rates also have a significant impact on the viewing experience of WSI. Slow refresh rates can produce a perceivable blurring as the images are scrolled, particularly from side to side. High refresh rate and “digital sync” technology can help to alleviate this problem.

There is also some debate over the need for monitor calibration. It has been suggested that display calibration is sensible only when the entire imaging chain is calibrated. Portable devices for calibrating displays are readily available and could be incorporated into the SOP for WSI viewing. In a study where medical pathologists reviewed WSI of 250 breast biopsies on uncalibrated versus calibrated LCD displays, no difference in diagnostic accuracy was observed, although time to diagnosis was slightly shorter when using calibrated monitors (Kuprinski et al. 2012). Ambient light in the room where the WSI is viewed can also have an effect and should be adjusted accordingly. The tissue micro array approach outlined earlier could also provide a simple “end to end” means of calibrating color. Some scanners also provide International Color Consortium ICC color profiles as part of the image metadata that allow the end user to calibrate their monitor using these. In the approximately three years that AstraZeneca has been using WSI for non-GLP peer review, we have not undertaken routine calibration of our display devices (monitors) and have not encountered any issues as a result. High-resolution display devices that are now capable of 4K often require graphics processing units of sufficient power to ensure optimal image rendering.

Many aspects of the imaging chain overlap and in practical terms it may be easier to divide image acquisition (scanner) and file transfer and viewing into separate SOPs. The current lack of guidance around using WSI for GLP peer review will probably result in a variety of bespoke approaches being adopted by each test facility that may involve some degree of user acceptance testing for the scanner and viewing software. In addition, a small validation study comparing diagnoses using glass slides versus WSI might be considered, to cover the use of WSI within the setting of peer review.

Software associated with the acquisition and viewing of scans could be validated using simple user acceptance tests that incorporated the key areas of concern for each process. Any approach for utilizing WSI for GLP peer review will have to be acceptable to those tasked with enforcing GLP compliance, as well as those agencies that review nonclinical data submissions.

For the purposes of peer review, the WSI is not considered raw data. If the scanned image is used for image analysis, then the program/application performing that task should be validated as per Section 3.2.5 of the OECD Guidance Document 16 (ENV/JM/MONO 2014) 30, as the process would be generating raw data.

Peer-Review Workflow

Any peer review utilizing WSI on a GLP study will need to be documented on the study protocol. The details will need to comply with the recent OECD guidance on pathology peer review (OECD Guidance Document 16 [ENV/JM/MONO 2014] 30).

Slide scanning can take place as soon as slides are released from the histology laboratory following standard QC checks and before they are handed over to the study pathologist. Suitable chain of custody documentation will need to be supplied to the peer review pathologist if images are provided on portable media. For additional security, the files on portable media can be encrypted with password protection. The decryption key for the hard drive is sent electronically to the peer review pathologist along with the draft pathology report, individual animal data, and summary necropsy and histopathology tables. The STP position paper of 2007 describes the use of images for formal peer reviews. If the study pathologist evaluated glass slides, but the peer review pathologist reviewed virtual slides, then any discrepancies should be resolved using the glass slides. The virtual slides are considered to be the pathologist’s interim notes that can be discarded (Tuomari et al. 2007). As the original glass slides are archived, there is no requirement to archive the scans. These recommendations were endorsed by the American College of Veterinary Pathologists and the British Society of Toxicological Pathologists.

Use of Digital Slides for Peer Review at Charles River Laboratories

At the Charles River Edinburgh facility, there are dedicated staff (Research Associates) to scan the slides and dedicated IT staff for server administration and system maintenance. The use of WSI for non-GLP peer review has now become the default for some sponsors (AstraZeneca). This approach is documented on the study protocol. For AstraZeneca, the slides are scanned prior to the primary read and shipped to the sponsor on encrypted portable media. The password to access the files is provided to the peer review pathologist once the study pathologist has reached draft report status. This approach allows AstraZeneca pathologists based in the United States, Sweden, and UK to peer review studies run or read at any of the other CRL laboratory test facilities worldwide. Any scans accessed by server are routinely deleted from the Charles River servers after 90 days but may be archived with other study materials prior to finalization at the request of the sponsor.

Certification of the Peer Review

Any GLP peer review undertaken utilizing WSI would be subject to the same approach as used for glass slides. This would need to be SOP driven with a mechanism to provide resolution in cases of nonconsensus. It is recommended that any significant disagreements between the peer review and study pathologist should be resolved by reviewing the original glass slides.

Conclusions

The use of WSI has significantly impacted the field of medical pathology, with WSI replacing glass slides for primary diagnosis in many instances. Within the context of GLP toxicology studies, the glass slide has remained the de facto medium, and this may in part relate to the large numbers of slides that require evaluation. Many sponsors and test site facilities are unwilling to deploy WSI within the GLP space, mainly due to uncertainties over the nature and extent of validation required and likelihood of regulatory challenge. Due to the large file sizes of WSI, there are also logistical implications for securely handling the large volume of data that would result. There is also a level of concern among some pathologists that WSI may be less optimal than glass slides for the evaluation of subtle lesions as outlined by Malarkey et al. (2015). The acceptance of WSI for use in GLP peer review should be the logical first step toward the potential adoption of WSI in the routine evaluation of pathology from regulatory toxicology studies. For WSI to be more widely accepted in the regulatory space, it may require a reevaluation of WSI versus glass slides using optimal image viewing hardware and software. Many of the issues encountered by Malarkey et al. (2015) related to the time taken for images to render (static and when scrolled). With the use of high refresh rate monitors, adequate graphics cards, and images stored on local caches, seamless rendering of images is now possible (personal observation). If toxicologic pathologists can be persuaded that WSI can really replace glass slides for peer review and primary reads, it should allow for an industry wide debate on how the technology should be deployed and regulated. This in turn may herald the opportunity to deploy WSI from regulatory toxicology studies in the emerging field of artificial intelligence and machine learning.