At2 digital slide scanner

Stroma-tumour fraction was determined on H&E stained histological sections. H&E stained sections were routinely estimated under 10 × 10 lens microscope to select the sections rich in tumour (tumour tissue > 50%, necrotic tissue < 10%). The stroma-tumour fraction was measured by Stroma Analyzer (Room 4, Kent, UK), as was described by Danielsen et al. ^{29 (link)} Briefly, the whole slides images of H&E stained sections were scanned with an Aperio AT2 digital slide scanner at ×40 (Leica, Germany), giving a resolution of 0.23 µm per pixel. Images with a resolution of 1.82 µm per pixel were used for image processing. A senior pathologist (Li Zhang) marked the tumour areas on the scanned images by using the software tool (Stroma analyzer, Room 4, Kent, UK). The stroma fraction in the selected tumour region was automatically calculated by the software (Stroma analyzer, Room 4, Kent, UK). Tumours with stroma fraction less than or equal to 0.50 were labelled low stroma, while those with stroma fraction greater than 0.50 were labelled high stroma.

Scanned digital image data files were created as whole slide images of 3 independent primary tumors of each genotype after 14 days using the Leica AT-2 digital slide scanner at a resolution of 0.50 μm/pixel (20× objective). Scanned slide images were imported into HALO imaging software. Slides were manually annotated using the HALO pen drawing tool to segment the tumor tissue area to be analyzed based upon visual assessment of the mass lesion. A unique machine learning pattern classifier algorithm was created using study specimens to designate areas within the tumor tissue segmented image area to classify tumor, tumor necrosis, loose areolar connective tissue (fat), or glass. Data output included area measurements for each tissue class for respective specimens. HALO Area Quantification algorithm was programed with appropriate intensity thresholding for the existing vimentin intermediate filament IHC chromogenic signal development, and subsequently used to quantify the vimentin expression in the tumor as seen by the brown DAB signal in each tissue section. Image analysis QA was performed by visual inspection on post processed image markup files.