Rapid Validation of Whole-Slide Imaging for Primary Histopathology Diagnosis

A Roadmap for the SARS-CoV-2 Pandemic Era

Megan I. Samuelson, MD; Stephanie J. Chen, MD; Sarag A. Boukhar, MBChB; Eric M. Schnieders; Mackenzie L. Walhof; Andrew M. Bellizzi, MD; Robert A. Robinson, MD, PhD; Anand Rajan KD, MBBS


Am J Clin Pathol. 2021;155(5):638-648. 

In This Article


The profiles of the cases scanned are shown in Table 1 and Figure 2. Of the 180 initial cases, 9 were excluded from digital conversion. The main reason for exclusion was a high rate of scanning failure at first attempt within the case. Three breast resection cases were removed before scanning because they had high numbers of slides per case, and scanning failures occurred most frequently with breast tissue sections. A total of 88 glass slides were rescanned, with the highest number of rescans occurring in breast cases (Table 2, Figure 1). The WSI file sizes ranged from approximately 500 MB to 2.5 GB. The average number of slides per case was roughly equal among pathologists (6–7.6), excepting one (pathologist 3, 11.1 slides per case) who had a greater proportion of breast resections. The number of cases reviewed per pathologist ranged from 32 to 37. Following case adjudication, cases were classified as being concordant or with minor or major disagreement with the original conventional light microscopic ("glass") diagnosis. Table 3 lists all disagreements adjudicated in the study. There were disagreements in grading of breast carcinoma (n = 5), prostate Gleason grade and score (n = 2), and oral dysplasia (n = 2). Other notable ones included misidentification of small diagnostic features in GI biopsies (n = 2) and dysplasia recognition in GI biopsies (n = 2). Counting all cases, the base concordance rate including major and minor disagreements for each pathologist ranged from 71.8% to 96.9% Table 4. The concordance including only major disagreements ranged from 93.7% to 96.9%. The mean pathologist concordance rate in sampled cases (n = 18) ranged from 72.2% to 94.18% Table 5. Mean concordance excluding minor disagreements ranged from 90.49% to 97% (Table 5). Assessment of the concordance rate for the group in sampled cases (n = 90) showed a mean of 83.62% counting all discrepancies and 94.72% counting only major disagreements Figure 3. No significant associations were identified between the occurrence of discrepancies (major or minor) and study pathologists (χ2 test, P = .19) or subspecialty (χ2 test, P = .14). No significant associations were identified between the occurrence of major discrepancies and pathologists (χ2 test, P = .91), case type (large vs small; χ2 test, P = .06), or subspecialty (χ2 test, P = .31). A small but significant increase in major disagreements was found in large cases compared with small cases (χ2 test, P = .04; see Figure 4, supplemental data).

Figure 2.

Distribution of digital cases (n = 171) by small (light gray) and large (dark gray) types (A) and numbers of slides per case (B) and by subspecialties (C) across study pathologists.

Figure 3.

Distribution of percentage of agreement in 90 cases (1,000 samples) drawn from total (n = 171). A, Concordance rate counting minor and major diagnostic discrepancies, with a mean of 83.66 (95% confidence interval [CI], 83.49–83.83; range, 75.56–92.22). B, Distribution of concordance counting only major discrepancies, with a mean of 94.58 (95% CI, 94.48–94.68; range, 90–100).