RRC-RHTIC: Previous Results
Aperio - Tissue Classification
Measuring Fibrosis using Masson Trichrome in Human Liver Biopsies - Scott Cotler, MD, Hepatology UIC
- Clinical Masson Trichrome biopsies were scanned on the Aperio ScanScope CS at 20x magnification at standard resolution. Genie classes were created for blank slide, hepatocellular parenchyma and portal tract/fibrosis. Example regions were drawn on several slides. The Genie classifier was trained and applied to all slides in the study. The percent of tissue area that was classified as portal tract/fibrosis was the outcome measure along with a manual counting of the number of portal tracts.
Masson Trichrome in Human Liver Biopsy Raw Image Genie
Measuring Fibrosis using Sirius Red and Fast Green in Mouse Heart - Ross Solaro, PhD, Yunbo Ke, PhD Physiology and Biophysics UIC
- Sections of mouse heart were stained using both Sirius Red and Fast Green. The slides were then scanned on the ScanScope CS at 20x magnification at the standard resolution. Genie classes were created for blank slide, non-fibrosis (green) and high-fibrosis (red). Example regions were drawn on several slides to capture the variability seen across all slides in the study. Two of these example regions are shown in the images below, color coded for the class. The Genie classifier was then trained and applied to all the images. The percent of tissue area that was classified as 'high-fibrosis' was the final outcome measure.
Mouse Heart Sirius Red / Fast Green Fibrosis Raw Image Processed Markup
Aperio - Brightfield Nuclear Marker
Cellularity of Bone Marrow - Amelia Bartholomew, MD, Department of Surgery, UIC
- This study focused on measuring the cellularity of bone marrow in mouse femurs. Regions of interest were manually draw on the study samples using Imagescope and Spectrum. The Aperio Nuclear algorithm was optimized to search for hematoxylin stained nuclei while ignoring RBCs. The measure is focusing on estimating the number of nuclei and not so much on getting their size and shape accurately measured. The final outcome measure is the cellular density or number of hematoxylin stained nuclei per square millimeter.
Normal Mouse Femur Raw Image Processed Markup
Injured Mouse Femur Raw Image Processed Markup
MCM2 in Human Prostate Biopsies - Peter Gann, MD, ScD, Department of Pathology, UIC
- In this study prostate biopsy slides were stained for the proliferation marker MCM2 and scanned at 20x standard resolution. Biopsy cores from the right side of the gland were inked prior to fixation and this black ink is present in most images. Regions of interest were drawn in Spectrum around areas of folded tissue and non-specific staining. In order to prevent inaccurate quantification of the ink, four Genie classes were created: blank slide(red), epithelium(green), stroma(yellow) and ink(blue). While the presence of the ink class had the desired effect of excluding the inked regions, it also excluded some valid regions of epithelium. The Aperio Nuclear algorithm was run within the epithelium Genie class to obtain a percent MCM2 positive measure. A weighted H-score was also calculated.
|Raw Image||Genie Tissue Classification||Nuclear Markup|
8OHdG in Rat Prostate Sections - Nur Ozten, PhD, Department of Pathology, UIC
- Rat Prostate Sections were stained for the oxidative stress marker 8-OHdG. A spatial sampling grid was created for each image using a custom built Visual C++ program that created a grid layer of user defined size and sampling density. Where the "+" structure of the grid fell on a glandular structure, that structure was drawn around for scoring using the Aperio Nuclear Algorithm. Intensity bins were adjusted to provide optimal contrast data. H-score and percent positive data were the final measures used.
|Raw Image||Nuclear Markup|
Aperio - Brightfield Microvessel Density
CD31 in Human Prostate Sections - Maarten Bosland, PhD, Department of Pathology, UIC
- Prostate sections were extensively mapped for cancer, HGPIN, normal near cancer and normal far from cancer compartments. Sections were then stained for the blood vessel marker CD31. The Aperio Microvessel algorithm was optimized to calculate the number of blood vessels and distribution of size of blood vessels. Parameters were optimized to the extent possible to combine nearby regions of staining under the assumption that they were continuous vessels. Very large vessels were ignored in the calculations.
|Raw Image||MVD Markup|
|Raw Image||MVD Markup|
Aperio Brightfield Tissue Microarray Quantification
ER, PR, Her2 in Breast Cancer TMA - Debra Tonetti, PhD, Szilard Asztalos, PhD - Biopharmaceutical Sciences, UIC
- A large core breast cancer TMA was constructed and stained for ER, PR and Her2 using the Ventana Autostainer at UICH. Slides were scanned at 20x magnification on the Aperio ScanScope CS and uploaded to the Aperio Spectrum Server. Images were segmented into cores by using the Aperio TMA Lab module of Spectrum. Either the Nuclear Algorithm or Membrane Algorithm were optimized depending on the marker. Each core was scored separately using the TMA Lab module and data for each core was exported.
|Marker||Raw Image||Markup Image||Zoomed Markup Image|
TCTP in Prostate Cancer Recurrence TMA - Andre Kajdacsy-Balla, MD, Department of Pathology, UIC
- The CPCTR recurrence tissue microarray was stained for Translationally Controlled Tumor Protein (TCTP). This marker is thought to be important in apoptosis, cellular differentiation and control of sperm functions in the prostate (Arcuri et al, Prostate 60:130-140, 2004). The TCTP stained slide was scanned at 20x standard resolution on the ScanScope CS. Genie classes were initially created for blank slide, epithelium, stroma and folded tissue. After many iterations of Genie training, the folded tissue class was removed as there was a high rate of misclassification of intensely staining epithelium as tissue folds. Drawing of further examples of folds and epithelium and further training had no positive effect on the classification. Folded tissues were then removed via manual drawing of regions of interest and the slide was then scored using the Aperio Positive Pixel Count algorithm. Per core TMA data was exported from Spectrum and further analyses were completed in SAS and SPSS.
|Raw Image||Genie Tissue Classification||Positive Pixel Markup|
pATM and H2AX in LNCap Cells - Alan Diamond, PhD, Anita Jerome, PhD - Pathology, UIC
- LNCap cells were irradiated using different doses and stained for H2AX using Alexa 488, pATM using Alexa 594 and Dapi. Four low powered fields with adequate cells were selected for analysis. Within each low powered field, 3 high powered fields with adequate cells were chosen. The example below is of highly irradiated cells. The outcome measure is integrated optical density units in the nucleus of each cell of H2AX and pATM.
|Raw Composite||Processed Composite Image|
|Dapi Channel||H2AX Channel||pATM Channel|
|Background Channel||Cell Segmentation|
MCM2 Expression in Luminal and Basal Cells in Human Prostate Sections - Peter Gann, MD, ScD - Department of Pathology, UIC
- Prostate RP samples were stain with Dapi, p63 (Alexa 488) and MCM2 (Alexa 594) and were analyzed on the Vectra. An inForm classifier was created for high powered field selected based on the density of epithelium found on the 5x images. The 10 fields with the most epithelium were selected for analysis. A second inForm classifier was created for identifying epithelium on the 20x images. Only cells that were classified as epithelium on the 20x images were counted in the analysis. A thresholded level of P27 expression was used to classify each cell as either luminal or basal. MCM2 integrated optical density was used as the outcome measure for both luminal and basal cells.
|Raw Composite||Processed Composite Image|
|Dapi Channel (All Nuclei)||P63 Channel (Basal Nuclei)||MCM2 Channel|
|Background Channel||Tissue Compartments Epi, Stroma, Blank||Epithelial Cells|