Context-based analysis of kidney biopsies
The image is showing a kidney biopsy stained for various markers including DAPI (blue), immune cell marker (white), glomeruli detection marker (red) and a marker to detect the tubules (green). The aim of this StrataQuest-based analysis was to detect morphological entities using the machine learning-based classifier, a nuclear segmentation using the DNN, and furthermore a proximity map was used to detect immune cells within a certain distance to the glomeruli. All the features needed for these tasks are provided in StrataQuest 7.
The DNN is a powerful tool for nuclear segmentation in tissues with extremely high cellular density, weak signal intensity, or high variation in nuclei size and texture. Based on the detected nuclei a cellular mask was generated to assess the CY5-positive immune cells.
The classifier works by marking just a few areas representative for the specific morphological entities of interest (i.e. glomeruli, tubule, interstitium) and the background. Based on these defined areas the classifier is able to separate the tissue into specific tissue classes, including the background, and will automatically generate binary masks for the detected areas. These obtained classifier masks can be further examined by analysing tissue class-specific parameters including area (mm2), cell count, percentage of marker-positive cells within the tissue class, etc. The final stage of this analysis included a proximity measurement in which only Cy5-positive cells within specific distance ranges from the outside edges of detected glomeruli were counted.