In the United States, review of digital whole slide images (WSIs) using specific systems is approved for primary diagnosis but has not been implemented for intraoperative consultation.
To evaluate the safety of review of WSIs and compare the efficiency of review of WSIs and glass slides (GSs) for intraoperative consultation.
Ninety-one cases previously submitted for frozen section evaluation were randomly selected from 8 different anatomic pathology subspecialties. GSs from these cases were scanned on a Leica Aperio AT2 scanner at ×20 magnification (0.25 μm/pixel). The slides were deidentified, and a short relevant clinical history was provided for each slide. Nine board-certified general pathologists who do not routinely establish primary diagnoses using WSIs reviewed the WSIs using Leica Aperio ImageScope viewing software. After a washout period of 2–3 weeks, the pathologists reviewed the corresponding GSs using a light microscope (Olympus BX43). The pathologists recorded the diagnosis and time to reach the diagnosis. Intraobserver concordance, time to diagnosis, and specificity and sensitivity compared to the original diagnosis were evaluated.
The rate of intraobserver concordance between GS results and WSI results was 93.7%. Mean time to diagnosis was 1.25 minutes for GSs and 1.76 minutes for WSIs (P < .001). Specificity was 91% for GSs and 90% for WSIs; sensitivity was 92% for GSs and 92% for WSIs.
Time to diagnosis was longer with WSIs than with GSs, and scanning GSs and uploading the data to whole slide imaging systems takes time. However, review of WSIs appears to be a safe alternative to review of GSs. Use of WSIs allows reporting from a remote site during a public health emergency such as the COVID-19 pandemic and facilitates subspecialty histopathology services.
Whole slide imaging involves high-resolution scanning of glass slides (GSs) prepared from formalin-fixed, paraffin-embedded tissue to create digital images that can be reviewed through digital platforms. The main components of whole slide imaging systems are whole slide scanners, software for image viewing, and digital monitors. Since its introduction in the late 1990s,1 whole slide imaging has gradually made its way into daily pathology practice. Initially, the main applications of whole slide imaging were nonclinical, including research, education, and conference presentations. In 2017, the US Food and Drug Administration approved the first commercial system for the use of whole slide imaging for review and interpretation of specimens and primary diagnosis in surgical pathology.2 This step opened the door for more widespread adoption and incorporation of digital pathology in the daily practice of pathology departments.
Whole slide imaging has a wide array of clinical applications. It allows provision of specialized histopathology services to less accessible locations, facilitates consultations and expert second opinions, provides easy access to previous pathology material for comparison with new material, facilitates maintenance of records of consultations and legal cases, facilitates multipurposing of limited tissue from biopsy specimens (when digital slides are obtained before the tissue is processed for ancillary studies), permits image analysis and computer-assisted quantification of immunohistochemical studies, permits sharing of cases with colleagues in tumor boards, and permits maintenance of pathology operations during emergencies, like the COVID-19 pandemic.3–6 Additionally, implementation of a digital pathology workflow with whole slide imaging can provide operational cost savings in the long run.7
Numerous studies have shown that diagnoses rendered on review of whole slide images (WSIs) and review of GSs using conventional light microscopy are highly concordant.8–16 In a study that compared review of WSIs and GSs for primary diagnosis in surgical pathology and included 2045 cases with 5849 slides, Borowsky et al8 found that the rate of major discordance between findings on review of WSIs and findings on review of GSs was 0.44%. A similar discordance rate was seen in another study of WSIs versus GSs for primary diagnosis in surgical pathology by Mukhopadhyay et al.9 The authors of both studies concluded that WSIs are noninferior to GSs for primary diagnosis in surgical pathology. However, in a systematic review that included 38 studies on the diagnostic concordance of findings on review of WSIs and GSs, the authors found a mean diagnostic concordance rate of 92.4%,15 lower than the concordance rate implied by the findings of the Borowsky et al8 and Mukhopadhyay et al9 studies.
A few studies have been published on the utility of WSIs for intraoperative consultation.17–20 In a validation study of WSIs for intraoperative assessment of oncology and transplant cases, the authors reported rates of concordance between findings on review of GSs and WSIs of 97% and 86%, respectively.19 Another study reported a rate of concordance between findings on review of GSs and WSIs of 92.6% for intraoperative assessment of thoracic pathology specimens.20
A major barrier to wide adoption of WSIs for primary diagnosis is efficiency. Longer times to reach diagnosis with WSIs than with GSs have been reported in some studies.21,22 In a large study by Hanna et al,12 a 19% median decrease in efficiency was reported for use of WSIs compared with use of GSs. In contrast, in a study in which synchronized parallel immunohistochemistry slides were viewed on a digital pathology workstation, the time needed for diagnosis was reduced by 81 seconds per case with WSIs compared to GSs.23 In yet another study, a workflow using WSIs was more efficient than a workflow using GSs when second opinions and expert consultation were required.24 The aim of the current study was to evaluate the safety of whole slide imaging and compare the efficiency of WSIs and GSs for intraoperative consultation.
MATERIALS AND METHODS
After obtaining the approval of the institutional review board at our institution, we retrospectively selected 91 cases previously submitted for frozen section evaluation in March, 2021 from 8 different anatomic pathology subspecialties: bone and soft tissue, cardiovascular and thoracic, dermatopathology, gastrointestinal and hepatobiliary, genitourinary, gynecologic, head and neck, and neuropathology. GSs from these cases were scanned using a Leica Aperio AT2 scanner (Leica Biosystems, Buffalo Grove, Illinois) at ×20 magnification (0.25 μm/pixel). Nine board-certified anatomic pathologists, referred to henceforth as reviewers, participated in this study. The reviewers were not part of the original frozen section diagnoses and did not establish primary diagnoses using WSIs in their routine practice.
The reviewers were given access to the deidentified WSIs and relevant clinical information, including the reason for intraoperative consultation for the selected cases. The reviewers were not given the diagnosis made during the original intraoperative consultation. The reviewers read the WSIs with Leica Aperio ImageScope viewing software (Figure 1) on consumer-grade 24-inch (61-cm) dual monitors with 1920 × 1080 native resolution (1080 pixels) connected to institutional Dell computer workstations. The reviewers recorded the diagnosis and the time required to reach the diagnosis for each case. After a washout period of 2–3 weeks, the reviewers were provided with the corresponding deidentified GSs and the same clinical information that was previously provided. The reviewers reviewed the GSs using light microscopes (Olympus BX43) and recorded the diagnosis and the time required to reach the diagnosis for each case. Time to diagnosis was measured with the starting time of the histologic examination of WSIs or GSs, including reading of the provided short relevant clinical information to the time when a diagnosis was entered into the data collection spreadsheet. The reviewers timed themselves using a stopwatch. All reviewers read the same WSI and GS cases.
For WSIs and GSs, we evaluated (1) the time needed to reach a diagnosis, (2) diagnostic accuracy compared to the original diagnosis rendered at the time of intraoperative consultation, and (3) intraobserver concordance. Diagnostic discrepancies (reviewers' diagnoses compared to original diagnosis) and discordances (diagnoses made on WSIs compared to those made on GSs) were identified by a board-certified general pathologist who did not participate as a reviewer in this study. The original frozen section diagnoses of all cases were confirmed by the specialists who reviewed the permanent sections of the entire case at the time of rendering a final diagnosis of the patient. The discrepancies and discordances were classified as minor if they were deemed to have no impact on patient care and major if they were deemed to have such an impact. Efficiency was based on the time required to reach the diagnosis. The difference in time needed to reach a diagnosis for WSIs and GSs was summarized and assessed using the Wilcoxon signed-rank test. Diagnostic accuracy of WSIs and GSs compared to the original diagnosis was determined by sensitivity and specificity.
RESULTS
The review results of all 91 cases were recorded by each reviewer. Intraoperative consultation was requested for diagnosis in 56 cases, assessment of margin status in 26 cases, and diagnosis and assessment of margin status in 9 cases. Table 1 shows the case distribution by subspecialty and the reason for intraoperative consultation. The file size (range, 8 MB to 2 GB) depended entirely on the size of the tissue being scanned and the resolution, while median scan time per slide depended on the amount of tissue being scanned as well as network connection speed (ranging from 90 seconds to 10 minutes). For this study, using the Leica Aperio AT2 scanner at ×20 magnification with our institutional network connection, the average WSI file size was 333 MB, and the scan time per slide ranged from 1.5 to 5 minutes with an average of 2 minutes 45 seconds, depending on the amount of tissue on the slide.
All reviewers required more time on average to reach a diagnosis using WSIs than to reach a diagnosis using GSs. For WSIs, the individual reviewers' mean times required to reach a diagnosis ranged from 1.63 to 2.02 minutes. For GSs, the individual reviewers' mean times required to reach a diagnosis ranged from 0.43 to 2.07 minutes. Overall, the mean time required to reach a diagnosis was significantly shorter for GSs (1.25 minutes) than for WSIs (1.76 minutes), with a mean difference of 0.51 minutes (P < .001) (Table 2).
Time to Diagnosis With Whole Slide Images (WSIs) and Glass Slides (GSs) by Reviewer and Overalla
Intraobserver concordance between the results on review of WSIs and GSs was seen in 767 of 819 reads (93.7%), with 29 major discordances (3.5%) and 23 minor discordances (2.8%) (Figure 2). The subspecialties with highest rates of discordance were genitourinary pathology and neuropathology, while full concordance was seen in cardiovascular and thoracic pathology and dermatopathology (Figure 3). Table 3 summarizes the cases with major intraobserver discordance between the diagnoses made with GSs and WSIs. The diagnostic specificities for GSs and WSIs were 91% and 90%, respectively. The diagnostic sensitivities for GSs and WSIs were 92% and 92%, respectively (Table 4).
Rates of major and minor discordance overall and by subspecialty. Abbreviations: BST, bone and soft tissue; CV&T, cardiovascular and thoracic; DP, dermatopathology; GI, gastrointestinal and hepatobiliary; GU, genitourinary; GYN, gynecologic; HN, head and neck; NP, neuropathology.
Rates of major and minor discordance overall and by subspecialty. Abbreviations: BST, bone and soft tissue; CV&T, cardiovascular and thoracic; DP, dermatopathology; GI, gastrointestinal and hepatobiliary; GU, genitourinary; GYN, gynecologic; HN, head and neck; NP, neuropathology.
DISCUSSION
In this study, we found that almost all diagnoses based on WSIs were concordant with diagnoses based on the corresponding GSs from surgical pathology specimens previously submitted for intraoperative frozen section evaluation. We also found that review of WSIs and review of GSs had similar diagnostic specificity and sensitivity. The time required to establish a diagnosis was slightly longer on average with WSIs than with GSs. Taken together, these findings support the conclusion that intraoperative review of WSIs is a safe and relatively efficient alternative to intraoperative review of GSs.
WSI examination is rapidly becoming an important tool that can be used by pathologists to serve a wide range of purposes. In some institutions, WSI examination has become the main tool for rendering primary diagnoses in surgical pathology.25 Previous studies have shown WSI examination to be noninferior to conventional GS examination in general surgical pathology10 and, more specifically, in evaluation of specimens from breast,26 skin,27 ovary,28 prostate,29 kidney (renal transplant setting),30 thorax,20 and the gastrointestinal system.31 Moreover, it has been shown that WSI examination is equivalent to GS examination for interpreting special stains and immunohistochemical studies.9
Our study adds to the growing body of evidence18–20 about the accuracy and efficiency of WSI compared to GS examination for intraoperative consultation. Our results on intraobserver concordance between GS and WSI examination for intraoperative consultation correlate with the findings of other studies.19,20 Cima et al19 reported a concordance rate of 86% in transplant cases and 97% in oncology cases. Additionally, Griffin et al20 reported a 92% concordance rate in thoracic specimens. Our findings regarding sensitivity and specificity of WSI examination are in line with those of a study on intraoperative consultation on oncology and transplant specimens, in which Cima et al19 reported specificity and sensitivity of diagnoses made by WSI examination of 96% and 100%, respectively, for oncology cases and 75% and 96%, respectively, for transplant cases.
Our findings show that the mean time to diagnosis was longer with WSIs than with GSs. To our knowledge, no previous studies on the efficiency of WSIs in the intraoperative consultation setting have been published. However, findings from earlier studies on the efficiency of WSI examination in non-intraoperative settings show some variability. In a large study performed at an academic cancer center, the authors found a 26% increase in the median time required to make the diagnosis using WSIs compared to GSs.12 Another study showed a 30% increase in the overall time required to sign out a case when WSIs were used.32 On the other hand, Randell et al22 reported that there was no statistically significant difference in the efficiency of WSI and GS examination for primary diagnosis in breast, gastrointestinal, and gynecologic pathology specimens.
Our finding that WSI examination was more time-consuming than conventional GS examination, with an average time difference of 0.51 minutes, may not appear significant on its face. However, the longer time required for WSI examination, compounded with the time required to scan the GSs and upload the digital files to the server, may result in a significant delay in patient care. Such delays may be particularly significant in the intraoperative consultation setting. Other factors that may contribute to patient care delay in this setting are errors in scanning or inadequate scanning. Additionally, parts of the tissue on GSs may be missed during scanning. Atallah et al33 reported a frequency of missing tissue in the WSI of 2% to 19%. Moreover, the authors found a negative linear correlation between rate of missing tissue and scanning time. However, the missing tissue did not result in significant changes in the final diagnoses in their study cohort.33
Limitations
In our study, we used the Leica AT2 to scan the GSs, which does not have a true optical ×40 magnification and uses a magnification doubler to achieve ×40 magnification, which may not give the same clarity as reviewing GSs at ×40. This may be an uncontrolled variable in this study. We included the time allocated to read the provided short relevant clinical information in the measurement of “time to diagnosis” in addition to reading GSs or WSIs. However, we did not account for various reading times, since only short relevant clinical information was provided for each case. In addition, it is important to emphasize that at our institution, WSIs are not routinely used for primary diagnosis, and thus the reviewers had limited experience with WSI examination for this purpose. Other studies have also reported that rendering diagnoses using WSIs was more time-consuming than rendering diagnoses using GSs; however, the time required to render a diagnosis on WSIs was reduced when pathologists were given a training set of cases.34 Additionally, several promising WSI viewing and control solutions have been proposed that have the potential to facilitate the technical component of WSI examination, including using tablet computers35 and video-gaming console controllers.36
CONCLUSIONS
Our study demonstrated high intraobserver concordance in the diagnoses made with WSIs and corresponding GSs from specimens previously subjected to intraoperative frozen section examination. Additionally, we show that WSI and GS examination have similar sensitivity and specificity, which is similar to the positive correlation between WSIs and GSs highlighted by a systematic review and meta-analysis by Azam et al37 in 2021. On the other hand, our results show that WSI examination requires more time to reach a diagnosis than does conventional GS examination.
The authors thank Stephanie Deming, ELS , senior scientific editor, Research Medical Library, The University of Texas MD Anderson Cancer Center, for assistance with editing of this manuscript and Michael Suchko from Pathology Imaging Lab, Department of Pathology, The University of Texas MD Anderson Cancer Center, for providing the information on WSI scanners.