Multi-centred validation of digital whole slide imaging for routine diagnosis

Digital pathology (DP) refers to the use of high throughput slide scanners to digitise diagnostic histopathology slides that are then reported on computer workstations as opposed to a microscope. The development allows electronic distribution of work to pathologists helping to address the pathologist capacity problem in many NHS hospitals. Sharing difficult cases more easily may help reduce error, and improve high complexity specialties such as renal pathology, that also require out of hours reporting. Prior to adoption, it is important to demonstrate that DP is fit for purpose of providing tissue imaging of sufficient quality to ensure diagnostic accuracy equivalent to light microscopy (LM), the current standard of care. Few comparison studies published have been adequately powered and cancer screening samples have not been specifically studied. Since the image produced in DP systems is inferior to resolution provided by LM it is questionable if DP is suitable for specialties requiring fine resolution such as renal biopsy interpretation. In addition, none of the published trials has examined the use of DP in immunofluorescence, which is essential if the DP is going to be used for renal biopsies. This study uses teams of 4 pathologists all examining the same series of samples, on both LM and DP. Samples will be randomised (excepting renal samples) to which modality is used first, and there will be a 6 week washout period between viewings. Reports will be scrutinised for differences, which will then be classified by an independent pathologist into major (would alter clinical management) and minor (would not). The original report gives the reference diagnosis. The ground truth (GT) for each case is decided on conclusion of the readings, by consensus of the study pathologists, using multihead LM viewing and taking into account the reference diagnosis if needed. This allows comparison of each pathologist’s performance on LM and DP against the GT. The study will examine 2000 complete histopathology samples including 600 samples each of breast, gastrointestinal (both including 200 cancer screening samples) and skin, and 200 renal samples performed for native or transplant related renal disease. The renal biopsies will include immunofluorescence for detection of immunoglobulin deposits, necessitating the first read to be performed on LM. The population of cases recruited will be selected sequentially from study centres’ archives and (except for renal) be enriched with 10% moderately difficult and 10% difficult samples. A qualitative study will examine, by means of structured interview, perceptions and experiences of pathologists prior to and during the progress of the study; assessing acceptability, timing and attitudes to DP. Eye tracking software will be used to examine pathologists’ examination techniques and this will be analysed alongside reporting discrepancies to identify if poor technique contributes to error in reporting. A Health Economic study will use lean working based value mapping to study the costs associated with review of cases for multi-disciplinary meetings and measure the potential efficiencies achieved by digitisation. Time for reporting cases with both modalities will be recorded to provide measurements of pathologist time saved. Clinically significant discrepancies will be further analysed to assess the likely variance in treatment cost compared to best standard of care, and the estimated impact on the patient.