Digital Pathology Initiatives and Experience of a Large Academic Institution During the Coronavirus Disease 2019 (COVID-19) Pandemic

Since 2016, our pathology department, in conjunction with the Ohio State University Comprehensive Cancer Center (Columbus), has invested in a variety of DP tools and platforms, and we began scanning retrospective oncology cases, mostly for research and archival purposes.

In 2017 the US Food and Drug Administration approved the first pixel pipeline for primary diagnosis, the Philips DP system (Table 1), consisting of Philips IntelliSite Pathology Solution 3.2 systems (IMS software version 3.2.1, Ultra Fast Scanner [UFS] serial No. FMT0145 with software version 1.8, and Philips display PP27 QHD; Royal Philips, Amsterdam, Netherlands). Shortly after, we validated this system in house for clinical diagnostic purposes in accordance with the College of American Pathologists guideline.¹⁴  The validation process included a sample set of 60 surgical pathology cases with H&E-stained, formalin-fixed, paraffin-embedded tissue slides that reflected the spectrum and complexity of specimen types and diagnoses likely to be encountered during our routine practice; 20 cases of immunohistochemical stained slides; and 20 cases of special stained slides (Grocott-Gomori methenamine silver stain, acid-fast bacillus stain, etc). The validation study established diagnostic concordance between digital WSIs and glass slides for the same observer, which ranged from 91% to 100% (average, 95%).

In May 2017, we began retrospectively scanning oncology cases back to 2010 using validated Philips UFSs (Royal Philips). Shortly thereafter, a few pathologists chose to have their cases (oncology and nononcology) prospectively scanned prior to receiving the glass slides for review. All pathologists in the department were offered the opportunity, at their own pace, to develop a DP workflow.

In December 2019, we built our new histology laboratory with 6 Philips UFSs installed adjacent to the histology staining machines, which significantly improved our DP workflow (Figure 1). As of September 2020, we have scanned and archived approximately 1.7 million slides from more than 190 000 clinical cases. Currently, we prospectively scan on average 2138 slides each day in conjunction with daily quality assurance metrics: H&E and immunohistochemical staining intensity, scanning of entire tissue area, and slide labels. The average pass rate for our daily quality metrics is 97.43% (2083 of 2138). Slides that do not pass these metrics typically result from a failure to scan the entire tissue area on slide, weak H&E or immunohistochemical staining intensity, or incorrect slide labels. In cases of failure to scan the entire tissue area or weak staining intensity, an attempt to rescan the slide is made. If the slide continues to fail, a note accompanies the case in our LIS indicating that the glass slide needs to be viewed by the pathologist.

All WSIs are scanned at a magnification of ×40 equivalent resolution (0.25 μm/pixel). Scan failures by the scanners are uncommon during our scanning process, at a rate of approximately 1.22% (20 660 errors in 1 687 289 scanned slides from May 2017 to September 2020). These errors include failed tissue detection, slides skipping, slides dropping, and other errors. Scan failure is not frequent enough to impact turnaround time and need not be a concern for institutions transitioning to digital workflows.

In 2017, we had 2 petabytes of total storage available, and at the time of this write-up we have used 75% of that space. Over time, the storage structure has been modified to address clinical needs while being fiscally responsible. Currently our total storage capacity is divided into 3 tiers (Table 2). Tier 1, X Bulk (Dell EMC, Seattle, Washington), is where all metadata for digital images are stored. Tier 2, Isilon (Dell EMC), is our first-level image storage, with fast retrieval but also high cost. Most recent digital images are stored here for rapid access. After a certain period of time, digital images are moved to tier 3, Scality (Scality, San Francisco, California), for permanent storage, where the images take longer to retrieve but the cost is less.

When the pandemic hit during the first quarter of 2020, The Ohio State University (Columbus) undergraduate and graduate campuses, including the College of Medicine (Columbus, Ohio) and The University Hospital, Wexner Medical Center (Columbus, Ohio), experienced significant and ongoing disruption of campus activities and clinical workflows. Most academic activities transitioned from live to remote using digital platforms. In the hospital, most employees who could perform their duties from home were either required or encouraged to do so. This option was also available for a large number of medical specialties, leading to a massive transition of medical visits into the virtual realm. Radiology as a diagnostic specialty was also able to take advantage of its significant digital workflow to minimize staff time within the hospital. In the department of pathology, we were well positioned to continue, with minimal disruption, with our education and research activities, as well as to take advantage of the aforementioned CMS waiver allowing remote sign-out to perform clinical services while protecting our pathologists, their families, and the community from unnecessary exposure. All medical and nonmedical employees in the department who were able to complete their duties remotely were encouraged to work from home. For attending pathologists, all on-site, direct in-person interactions with trainees during sign-out (ie, double scoping) and colleagues for intradepartmental consultation were discontinued and all meetings and conferences were transferred to virtual platforms.

As previously described, at OSU we deployed a DP workflow for primary diagnosis 2 years before the pandemic. The Philips DP system was fully integrated into our pathology LIS, initially Sunquest CoPath (Roper Technologies, Tucson, Arizona) and more recently Epic Beaker (Epic Systems Corporation, Verona, Wisconsin), so that it is also fully integrated into the hospital's electronic medical record (Epic Systems Corporation). This integration allows WSIs to be opened directly through a link in the pathology case environment within our LIS.

After the CMS waiver allowing remote sign-out, we started an internal validation process before attending pathologists could issue primary diagnoses from home. The validation procedure required a test of their home equipment (Figure 2); an Internet connection with an optimal speed (100 Mbps or faster); establishment of a secure VPN connection to access the hospital EMR, pathology LIS, and IMS; and validation of the home monitor using the Point of Use Quality Assurance Remote Image Tool^15,16  v 2.0.20 (Figure 3). The Point of Use Quality Assurance tool was developed at the University of Leeds, Leeds, United Kingdom, initially to assess monitors for radiologists working remotely, and recently upgraded to test monitors for pathology remote use. An example home workstation of a pathologist is shown in Figure 4. This pathologist's setup included a Dell laptop computer with i9-9980HK CPU @ 2.4 GHz and 32 GB RAM (Dell Computer, Round Rock, Texas) connected to the secure institutional VPN network, an accessory 32-inch (81-cm), 4K UHD Philips monitor (Royal Philips) to display images, and a commercial Internet connection speed of 1 Gbps.

A laptop or personal computer comprises the required hardware, without a need to buy additional hardware unless the monitor fails the Point of Use Quality Assurance test. There are, however, optional additions, including larger monitors (as in the example above), ergonomic mice, track pads, etc, which may add some benefit in terms of ergonomics and comfort.

Early in the pandemic, some pathologists chose to review cases and order ancillary testing remotely, only coming to the hospital to verify completed cases or review slides/material that required further review of the glass slides. After the CMS waiver was issued, some pathologists chose to work almost exclusively from home.^1,2,17–19  These efforts allowed pathologists who were high-risk or in quarantine to work from home, facilitated a reduction in the number of staff required on-site, increased our ability to socially distance for staff members required to be on-site, and reduced the total number of staff touching and interacting with case materials (slides and paperwork).

After analyzing the steps involved in creating and delivering glass slides in our regular analog workflow, we found many potential risk points for cross-contamination considering the total number of touch points for glass slides, cardboard slide folders, and the paperwork generated to accompany the slides on each case. We assessed the number of touch points between initial slide staining/coverslipping and final filing in the slide archives and determined that the slide(s) and slide folder(s) for a case will have been handled on average by at least 2 histotechnologists, 1 courier, 2 people in slide distribution, a resident, a pathologist, 2 or 3 additional staff members in our central pathology delivery hub, 2 people in the scan center, and 2 or 3 staff members in the archive area. Total touch points in our routine digital workflow are 15 to 20, at best. In addition, those pathologists with a nondigital workflow receive paperwork that has been touched by the accessioner, grosser, transcriptionist who proofed the paperwork, courier who transported the paperwork to our delivery hub, and staff member who collated the paperwork and slides prior to delivery to the pathologist.

Removing the resident(s) and attending(s) from this scenario, in a digital and paper-free workflow, decreased the total touch points to 10 to 14, as well as eliminating paperwork beyond the initial scanning of the requisition and demographics sheet. In a paperless, glassless, and folderless environment, the touch points for attendings and residents are virtually zero, and the number of touch points for technical staff is reduced to fewer than 5.

Before the COVID-19 pandemic, 23 (54.8%) of 42 subspecialized pathologists at our institution had completed the initial training and validation for on-site primary diagnosis using DP. Of these 23 pathologists, 18 were using a combination of glass slides and WSIs for primary diagnosis (Table 3); furthermore, 2 pathologists (4.8% of the total of 42) were using exclusively WSI for primary diagnosis without ever receiving glass slides.

Since the pandemic began, a total of 30 of 42 pathologists (71.4%) have been trained and approved for on-site primary diagnosis using WSI. Of those 30, 21 individuals were also approved for remote primary diagnosis, including 5 of 42 pathologists (11.9%) who reported an exclusive DP workflow (either on-site or remote without receiving/viewing glass slides). At present, 24 of 42 pathologists (57.1%) report using a combination of WSI and glass slides for primary diagnosis. Because of the limitations of WSI for some subspecialties such as hematopathology and cytopathology (smears not scanned) and renal pathology (immunofluorescence-stained slides not scanned), subspecialized pathologists in these areas have to continue to use glass slides exclusively or predominantly. If pathologists in those subspecialties are excluded, 3 (9.4%) of 32 pathologists are exclusively using glass slides for primary diagnosis during the pandemic, and the majority (29 of 32; 90.6%) of pathologists are using WSI for primary diagnosis either exclusively or in combination with glass slides. In summary, the number of pathologists using WSI in our department increased significantly during the pandemic (Table 3).

Based on survey results, data collected, and personal experience of the authors, the advantages and disadvantages of a DP workflow before and during the COVID-19 pandemic are summarized below and in Table 4.

The first significant advantage is the prevention of errors during slide scanning and case entry when reviewing slides and completing reports, as the system verifies the bar code label on glass slides digitally, prior to attaching the image(s) to the correct case.

Second, there are advantages inherent to using the viewing software, including WSI annotation and a measurement tool allowing precise microscopic measurements of tumor size, depth of invasion, and distance to surgical margins. Slides can also be viewed side by side, which allows for easier comparison of an H&E image and its corresponding immunohistochemical or special stained slide (Figure 5). The system also allows efficient sharing of images with colleagues, sending either the case link or using the built-in collaboration function of the IMS. In the collaboration function, 2 pathologists can simultaneously move the slide, change magnification, or annotate the slide remotely in real time. This is particularly useful for intradepartmental consultations as well as for case sign-out with trainees (Figure 6).

Third, a digital workflow allows for increased flexibility in staffing, as slides can be rerouted in real time to available pathologists at different locations. A related advantage is the ability to reduce or prevent virus transmission by decreasing the number of in-person/face-to-face interactions as well as the number of individuals handling case material. Improvements in efficiency and efficacy are reported previously by Retamero et al.²⁰ 

Fourth, a digital workflow can improve office ergonomics by allowing more flexibility in shifting arm, head, and body position. A digital workflow allows for a more natural position—simply looking forward at a screen rather than bending forward and downward to use a microscope.

Some challenges of a digital workflow cited by some colleagues are also summarized. First, there is the requirement for high-speed Internet connectivity if signing out from home, as well as potential need to purchase new hardware for home use. A second commonly cited challenge was difficulty identifying microorganisms (in particular Helicobacter pylori, some fungi, and acid-fast bacilli, where fine focusing is often required) and evaluating special stains such as Congo red stain, which requires polarization with a microscope at this time. Third, pathologists involved in head and neck diagnosis mentioned interpretation of nuclear features of papillary thyroid carcinoma as a challenge. Fourth, rare circumstances of suboptimal slide scanning due to scant tissue, weak or overly intense staining, thick sections, or tissue folds were mentioned.

Anatomic pathology services have unique workflows that bring additional considerations when using DP tools. The experience of these divisions is presented below.

The practice of cytology uses a variety of slide preparations, including direct smears, cytospins, liquid-based cytology, and cell block sections. Cytology slides (direct smears, liquid-based cytology, cytospins) may contain single cells and 3-dimensional cell groups. Cytology slides can be very thick, with cells overlapping, and may contain obscuring materials (eg, mucus, blood, lubricant, and ultrasound gel) that make scanner focus on diagnostic material when scanning cytology slides challenging. In addition, when cytology material is scant the scanner may not detect and focus on rare cells. Therefore, routine scanning of cytology slides for clinical diagnosis is not practical by currently available DP technology; however, all slides prepared from cell blocks are routinely scanned for diagnostic use at our institution.

Hematopathology uses a variety of slide preparations, including aspirate smears, touch preparations, and paraffin-embedded tissues (eg, core biopsies, clot sections, and lymph node sections). Although slides prepared from paraffin blocks are easily digitized, many of the same issues encountered in cytopathology also limit use in the evaluation of other preparations. Aspirate smears, touch preparations, and peripheral blood smears usually require oil emersion and high magnification (×50 or ×100) to adequately evaluate the morphologic features. Currently, the use of digital technology is limited for the hematopathology service with respect to these latter considerations; however, all slides from formalin-fixed, paraffin-embedded tissue specimens including lymph node cases are scanned routinely.

Digital pathology tools are routinely used by pathologists and trainees for autopsy case slide review and teaching conferences. Prior to the pandemic, all autopsy slides were prospectively scanned to facilitate case review and intradepartmental consultations. Additionally, residents had prior access to narrated instructional videos for evisceration and dissection techniques as well as annotated WSIs for common cardiac and pulmonary autopsy findings.

Teaching conferences that require in-person collaboration (eg, brain cutting) were limited during the pandemic to the attending pathologist and service resident. As medical students begin to return to clinical rotations, the implementation of live audio/video streaming is underway to enable clinicians, medical students, and other trainees to interact with the autopsy service in conference settings. Conferences previously using digital slides/images have continued without significant changes.

Intraoperative consultations are a unique aspect of pathology practice. At our institution, digital telepathology was available and frequently used prior to the pandemic. Because of the time-sensitive nature of intraoperative consultations, the digital tools used for other aspects of pathology are not feasible. As such, digital intraoperative consultations are performed using a VisionTek digital microscope and conferencing software (Somagen Diagnostics Inc, Edmonton, Alberta, Canada). Prior to the pandemic, digital tools were principally used by the neuropathology faculty (day service, evening service, and after-hours call) or surgical pathology faculty called after hours (11 pm–7 am; weekends/holidays). During our normal day (7 am–5 pm) and evening services (5 pm–11 pm), intraoperative consultations for surgical pathology were almost exclusively performed on-site using glass slides. In contrast to the small samples submitted for neuropathology consultations, surgical pathology consultations range from small biopsies to large resections. As such, having faculty available on-site remains critical for proper tissue sampling and diagnosis. Overall, the intraoperative consultation service did not see significant changes in the use of DP tools as a result of the pandemic. The specimen volume remained fairly steady, with some changes in the handling and processing from high-risk tissues (head and neck specimens and thoracic specimens).

At our institution, a minimum of 1% of a total of any given pathologist's cases during a given period of time are rereviewed as part of our formal quality assurance procedure. These cases are randomly selected and reviewed monthly. The results of these reviews are reported on a quarterly and annual basis to the hospital quality assurance department and departmental leadership.

The percentage of major and minor diagnostic errors prior to and during the pandemic demonstrated little, if any, variation. Cases with no major or minor diagnostic errors in the last quarter of 2019 (prepandemic) numbered 696 of 703 (99%). During the first quarter of 2020, the total number of cases with no major or minor diagnostic errors was 749 of 759 (98.6%) and during the second quarter of 2020 where the data included remote sign-out, the total was 606 of 611 (99.2%). This suggests that accuracy was not significantly impacted by the increase in use of DP for primary diagnosis and by the deployment of remote sign-out. However, these data do not distinguish between cases signed out digitally on-site versus those signed out digitally at a remote location (home, etc). A separate internal study is currently underway to further evaluate any differences in the accuracy of on-site versus remote digital sign-out.

The COVID-19 pandemic has had a significant impact on trainee education and workflow. Teaching institutions have implemented a wide variety of solutions, including the installation of plexiglass barriers between pathologists and trainees to facilitate double/multiple-headed scope instruction as well as the use of conferencing software to share images projected from a microscope camera or WSI. Because of the prior establishment and use of DP tools prior to the pandemic, our residents and attending pathologists were well equipped to transition quickly to DP modalities for trainee education.^12,21,22 

Before the COVID-19 pandemic, most pathologists engaged in one-on-one teaching at the microscope using glass slides (aka double scoping). Only a small number of attending pathologists used WSI to review cases with the resident in an in-person setting.

During the pandemic, however, in-person interactions were suspended and both attending pathologists and trainees had to quickly adapt to an entirely digital sign-out. The majority of these interactions used 1 of 2 methods: (1) videoconference sharing of the attending pathologist's monitor projecting the WSI and (2) use of the Philips IMS collaboration function, which allows the attending pathologist and trainee to simultaneously interact with the WSI with a concurrent phone conversation (Figure 7). Of note, both methods allow trainees to annotate slides during their case previews for later review and discussion. Other less commonly used modalities included videoconference sharing of live images captured through a microscope camera as well as reviewing slides separately with subsequent electronic or telephonic communication to address trainee questions and the final diagnostic interpretation (Figure 8). Our trainee survey results demonstrated that the collaboration function was preferred as a means to improve interaction with the cases and the attending pathologist (50%; 7 of 14) over videoconference sharing of the attending pathologist's monitor projecting the WSIs (21%; 3 of 14) and other methods, including double scope with glass slides with social distancing (7%; 1 of 14), review of glass slides separately (7%; 1 of 14), and review of WSIs separately (14%; 2 of 14).

For specialties that relied heavily on multiheaded scope conferences (in particular cytology and hematopathology), images were typically projected from a microscope camera onto a large television screen in a room that allowed for adequate social distancing. Alternatively, live images obtained through a microscope camera were shared via videoconferencing.

Similar to case review and sign-out with the attending pathologist, most unknown slide sessions were conducted using videoconferencing platforms to project WSIs or live images from a microscope camera. Overall, this method was similar to an in-person session and did not significantly impact this educational activity.

Since 2018, our residents have maintained a WSI teaching set in the Philips IMS. To date, the collection includes nearly a thousand surgical pathology cases organized by subspecialty. A PDF file is available with the final diagnosis for each case. During the pandemic, use of the WSI study set significantly increased as a way to supplement exposure to case volumes that were impacted by initial restrictions on elective procedures.

During the early period of the COVID-19 pandemic, all research activities (with the exception of COVID-19–related activities) were suspended at our institution. This resulted in the closing of core laboratory facilities and a transition to research support staff working from home. As a result, studies requiring new histologic or immunohistochemical techniques were put on hold. However, studies that could be conducted using digitally scanned material continued.

The majority of our scanned images are from oncologic cases and have been supported by a $10 million investment by the Ohio State University Comprehensive Cancer Center. This unique initiative has allowed for remote collaborations and substantial data elements (examples: area-based measurements, cell-based measurements, and noncellular measurements).¹  Although many of these data can be obtained visually by a trained pathologist, image analysis and artificial intelligence platforms can be introduced to improve precision and reproducibility.²³ 

For laboratories with graduate students, novel strategies were used to maintain their progression within the degree program. Methods included the use of WSI data for translational projects as well as using clinical trial data available from the National Institutes of Health and material from our institution's biospecimen repository.

To assess the DP experience at our institution during the pandemic, a voluntary survey was created and sent to all anatomic pathologists/hematopathologists and trainees. Of 42 active anatomic pathologists/hematopathologists, 35 responded to the survey. Of 16 residents, 13 responded to the survey.

Of the 35 respondents, the majority had been in practice for more than 20 years (13 of 35; 37%). Attending pathologists in practice for 11 to 20 years and 4 to 10 years were evenly distributed, each containing 8 of 35 (23%) respondents. Finally, a minority of attending pathologists responding had been in practice for less than 3 years (6 of 35; 17%).

During the COVID-19 pandemic, the proportion of attending pathologists using DP tools in their daily workflow increased, including for primary diagnosis (on-site or off-site), intradepartmental consultations, tumor board, education, and research. The most significant changes occurred for primary diagnosis (off-site), intradepartmental consultations, tumor board and education (sign-out with trainees) (Figure 9).

Following the CMS waiver, 21 of 35 (60%) reported taking advantage of DP tools for remote primary diagnosis. If restrictions were permanently lifted, 30 of 35 respondents (85.7%) said they would continue using DP tools for remote primary diagnosis.

Trainees reported a smooth transition to DP tools for educational conferences and sign-out with attending pathologists. Of the methods used for digital sign-out, trainees preferred use of the Philips IMS collaboration tool over other digital methods. However, both residents and attending pathologists expressed a desire to resume in-person double scoping when it becomes safe to do so.

The early adaptation of DP at our institution allowed for the quick and seamless transition to a predominantly digital workflow during the COVID-19 pandemic. We understand that the transition period can be slow, and the learning curve can be steep (at least initially) for some pathologists before they reach a level of comfort with the technology. However, the pandemic served as a catalyst for more pathologists to realize the advantages of a digital workflow and integrate these tools into their daily practice.

Recently, Hanna et al¹¹  reported their validation of a DP system including remote review during the COVID-19 pandemic. They found good interobserver concordance between diagnosis from remote WSI review using the Leica Aperio GT450 System (Leica Biosystems, Wetzlar, Germany) and glass slide diagnosis, with 100% major diagnostic equivalency and an overall concordance of 98.8%.¹¹  Liscia et al²⁴  demonstrated similar findings of good concordance between remote sign-out using WSI captured by the Hamamatsu Nanozoomer system (Shizuoka, Japan) and sign-out using glass slides. Both studies validated remote diagnosis using WSI as equivalent to diagnosis using glass slides, but without real-time implementation of digital workflow in their institutions. Stathonikos et al²⁵  summarized their experience using WSI scanned by the Hamamatsu Nanozoomer and managed by the Sectra Picture Archiving and Communication System (Sectra Medical, Linkoping, Sweden) at Universitair Medisch Centrum Utrecht (Utrecht, The Netherlands) for clinical service, education, and multidisciplinary meetings. Whereas the Leica Aperio GT450 System is a Food and Drug Administration–approved system, the Hamamatsu Nanozoomer system is not. In our study, a different Food and Drug Administration–approved system, the Philips IMS together with UFSs), has been implemented in our clinical workflow for clinical diagnosis, both remotely using home monitors and at the workplace using Food and Drug Administration–approved monitors. Our quality assurance process revealed that the discordance rate has not changed during the pandemic, with an increased number of cases reviewed and/or signed out digitally compared with before the pandemic. Additionally, we have summarized our experience in using DP in education and research, from the perspective of both attending pathologists and trainees regarding their experience of using DP in our institution.

Our experience has demonstrated that implementing DP in our routine clinical workflow increases flexibility in both staffing and the sign-out process, while maintaining clinical productivity and patient care quality. Digital workflows reduced in-person interactions and improved efficiency of education and research during the pandemic. We also recognized some challenges, such as difficulty in identifying microorganisms on WSIs and interpreting some histologic findings such as nuclear features of papillary thyroid carcinoma. Minor challenges included the requirement of high-speed Internet connectivity and the potential need to purchase new hardware for remote digital sign-out.

Undoubtedly, the use of DP tools will remain increased, compared with the use before the pandemic, following resumption of normal activities. Within our own department, we observed an increase in enthusiasm and motivation among our faculty to use DP tools and a desire to continue or increase their use of DP in their future practice. New faculty hires and trainees will quickly adapt these DP tools and be able to take advantage of their growing capabilities. We hope that the CMS waiver will be considered for more permanent use, as it has allowed some faculty to continue to carry out high-quality patient care and academic work from home when they might otherwise have been unable to work on-site.