Image processing pipelines in Image Analyst MKII
Image Analyst MKIIPipelines Glossary
HomeWorkflowImage Processing BasicsFunctions GlossaryPipelines GlossaryProtocolsQuick How ToSearch

HE brightfield fat droplet analysis

Parameters:
Name # Type Description
Threshold level (pixel value) 1 real Pixel intensity value distinguishing fat droplets and cellular structures. 0-255 for 8-bitgrayscale or 24-bit RGB images. A lower value makes detection more sensitive.
Size cutoff between small and large droplets (pixels, approximate diameter) 2 integer Smaller objects than this value in pixels (diameter) will be detected as small droplets, larger object as large objects
Minimum roundness of droplets (0-1) 3 real 1 for disc, smaller for irregular shapes. 0 for not checking
Minimum size of droplets (pixels, diameter) 4 integer Diameter. The surface are calculated from this diameter is used to gate off droplets smaller than this size.
Maximum size of droplets (pixels, diameter) 5 integer Diameter. The surface are calculated from this diameter is used to gate off droplets or background areas larger than this size.
Histogram: number of bins 6 integer When the output is histogram, the histogram is calculated using this number of equal sized bins, where the lowest bin starts at the smallest data point and the highest bin ends at the largest data point. If explicit range definition is used below, the bins will bound the given range. Use zero bin size for automatic binning.
Histogram: range 7 string If left empty or set to "All", a histogram is generated where the lowest bin starts at the smallest data point and the highest bin ends at the largest data point. If a range is given here e.g. 100-500, then the lowest bin starts at the beginning of the range and the highest bin ends at the end of the range.
Description:
Analysis of lipid droplets in adipose or brown adipose tissue using brightfield RGB images of H&E staining. The algorithm recognizes round holes in the tissue as fat droplets. The dynamic range of detection is increased by separately detecting small and large objects defined by the “Size cutoff between small and large droplets” parameter. Non-parenchyma area is excluded by the “Maximum size of droplets” parameter.

Input: RGB color or grayscale brightfield images. The sum of all color channels is used for calculations. For TIFF files of images. Use the File/Open folder of TIF files menu point to open the folder where the files are. Then select Simple Rule and %position%.tif in the TIF folder of the Multi-Dimensional Open Dialog "TIF Folder" tab. If you are opening *.tiff files change the rule to %position%.tiff. Press Scan folder and go to the Open tab. For pyramidal image formats, such as *.ndpi, use Bio-formats for opening, select the highest magnification, shown as a position and an appropriately sized (up to 20,000 x 20,000) pixels crop area. The default pipeline configuration is for 20x-40x magnification (0.2um/pixel).

Output: Excel worksheet of the number and mean size of droplets in the image, and a histogram of size distribution. Use the "Run pipeline on all stage positions" option of the run pipeline button (double blue arrowhead) to evaluate all files in the folder.

Adjustments:
The threshold level is an absolute pixel value for the sum of three channels therefore is between 0 and 255 for 24 bit images or 0 and 65535 for 48-bit RGB images.
To tune the detection to less sensitive to small objects or irregularities of the cytoplasm, increase the minimum size, and size cutoff values.