Whole-Slide Imaging Validation: Scanner-Specific Failure and a Major Discordance
An academic pathology department plans primary digital diagnosis for all surgical H&E slides. Validation included 180 cases, three scanners, three pathologists, and a two-week washout. Glass slides were read first, then whole-slide images; discordances were adjudicated by a fourth pathologist. Major discordance was 2.3% and minor discordance 9%. Scanner B had 7.8% focus failures at 40× oil, but 20× images were acceptable. The vendor recommends go-live at 20× on all scanners to reduce scan time. One breast core biopsy scanned on Scanner B was signed out as ductal carcinoma in situ at 20×; the original glass diagnosis was invasive carcinoma. The department wants a decision on go-live and a plan for the discrepant case. The department has no prior digital validation.
1. Does the validation support go-live for all three scanners at 20×? Justify your decision.
Read the worked answer
No. The validation does not support a blanket go-live at 20×. Validation must be scanner-specific and intended-use-specific; a 7.8% focus failure rate at 40× on Scanner B is a scanner performance problem that should be fixed or the scanner should be excluded. The proposed 20× workflow is a different intended use from the validated 40× diagnosis, and 20× may be inadequate for mitotic counts, nuclear grading, lymphovascular invasion, and subtle invasion. The 2.3% major discordance rate must be interpreted against a prespecified acceptance threshold, not accepted because it sounds low.
I would approve go-live only for scanners and magnifications that met validated acceptance criteria, with a requirement that equivocal or high-risk cases be reviewed at 40× or on glass. Scanner B should undergo service and revalidation. If 20× is desired for screening, it should be validated separately with cases enriched for the diagnoses it might miss.
2. The breast core biopsy scanned on Scanner B was called DCIS at 20× but was invasive carcinoma on glass. What is your immediate action and what does this case indicate about the validation?
Read the worked answer
This is a major, clinically significant discordance because DCIS versus invasive carcinoma changes sentinel lymph node management and possibly re-excision. I would immediately rereview the glass slide, confirm the invasive component, and issue a corrected report after review. The patient's care team must be notified, and the case should go through the department's discrepancy/adjudication process. I would also review all Scanner B breast cases scanned at 20× for similar misses.
The case indicates that the 20× intended use is not validated for breast core biopsy diagnosis, at least on Scanner B. It is not simply a pathologist error; the system permitted a magnification that did not show the invasive focus. The validation should include subspecialty-specific cases and a predefined major-discordance threshold. If the validation protocol allowed this case to pass without triggering a failure, the protocol itself needs revision.
3. What ongoing quality metrics and governance should be required after go-live?
Read the worked answer
Ongoing monitoring should include scan failure rate by scanner, deferral-to-glass rate, major and minor discordance on audit, turnaround time, and pathologist-specific feedback. I would require a random audit of a defined sample of digital cases each month, with targeted review of high-risk areas such as breast, prostate, and margin assessment. Any scanner software or magnification change should trigger change control and revalidation.
Governance should include a digital pathology committee with pathology, IT, and laboratory leadership; documented approval of each scanner and intended use; and a process for reporting discrepancies without blame. If an AI tool is added later, it needs separate validation and drift monitoring. The department should not treat the initial validation as a one-time event; it is the baseline for a quality system.