Results 2010

Quality Scan
For quality assessment 10 slides of different origin (skin, breast, kidney, nerve,…) and different staining (H&E, PAS, toluidine blue, HER2,…) have been used. The slides have been provided by different institutions.

Every vendor had to scan these 10 identical slides with a time limit of 1 hour for two slides. The time limit was used to organize the circulation of 10 slides in packages of two slides.

The comparison of the WSI was established as a direct assessment and comparison of ROIs.  Therefore the providing pathologist marked 3 diagnostically relevant regions (DROI) in each slide, performed on a previously generated WSI. The regions had been marked as red rectangles on the WSI. Referees searched in all WSI for the corresponding DROIs using the original viewer of the corresponding vendor. The region was displayed 1:1 (1 camera pixel corresponds to 1 monitor pixel) and a 1000x1000-screenshot with identical center point position was taken.

So for each slide and each DROI 7 corresponding screenshots of the participating scanners were available. These screenshots had to be ranked by pathologists according to image quality. Due to 10 slides with 3 DROIs a set of 30 series of 7 screenshots per series had to be evaluated.

For the evaluation a special program was developed. The program was running on 4 identical PCs with two identical Barco Monitors each. The user had to order the randomized positioned screenshots with respect to image quality. The best image got rank 1 the weakest rank 7.

The program was used by 40 pathologists who performed the evaluation parallel to the 94th Annual Meeting of the German society for pathology. A time between 25 and 45 minutes was necessary to evaluate all series.

The mean rank (between 1 and 7) was estimated over all 30 image series was used as base for quality evaluation.

Mass Scan
For the mass scan 35 HE slides have been provided by more than 11 institutions. The series consists of biopsies and operational material from different origin (kidney, liver, breast, pancreas, skin …). From every slide we had minimal 9 serial sections which have been randomly distributed to 10 boxes with 35 slides. Every vendor got also the whole box randomly. So we had maximal equal conditions.

The scanning time was calculated by analysis of the log files (overnight scan) and parallel by stop watch (referees). Time measurements were performed always under supervision of a referee in cooperation with the responsible person of the vender.

The total scanning area (42,736 sq cm) was measured at 2.5x magnification with the image analysis function of the slide explorer.

To avoid fast scan without the generation of useful data we decided to measure the percentage of scanned area out of focus. Therefore we analyzed all WSI of all vendors by measuring the focus of all 256x256 pixel sized tiles. Tiles out of focus were recognized by a criterion and counted.

The only criterion which we used for the assessment of the mass scan was the focus corrected scanning time (FCS):

FCS =TAS x PAF/ TTS

with FCS - focus corrected speed, TAS - total area of series, TTS - total time for series (gross), PAF - percentage of area in focus (measured on 256 x 256 tiles).