cellprofiler load time series multi page tiff

Cellprofiler load time series multi page tiff how to#
Cellprofiler load time series multi page tiff manual#

The variability was analyzed by calculating intraclass correlation coefficient (ICC) using a two-way mixed single measure’s model.

Variability between measurements was determined by the difference in mean nuclei density per digital slide. Nuclei per area were defined as the number of CD3 positive cells per tissue area. For each slide, measurements per tile were combined. Baseline characteristics of these measurements can be found in table 4. We used R (CRAN R version 3.0.2) to collect, concatenate and analyze the.csv files. With the exception of the CellProfiler pipeline, all steps were unique for each 606 slides. The number of CD3 positive cells per tissue area for each digital slide were compared between two runs. From these 606 digital slides, masks were created, reviewed and adjusted. The 303 slides were scanned in two runs at 40x, using the Roche iScanHT slide scanner. Tools, wall-clock times and instructions can be found in table 3. Other steps ran on a high-performance cluster (HPC) (8x Intel(R) Xeon(R) CPU E5-2630 0 2.30 GHz, 38x Intel(R) Xeon(R) CPU E5-2640 0 2.50 GHz, 11x Intel(R) Xeon(R) CPU E5-2630 v2 2.60 GHz, all with 12 cores and 128 GB RAM per node) was used. For some steps, we ran the slideToolkit on local hardware (MacMini with 2 GHz i7 and 16 GB RAM). Measurements can then be gathered and further analyzed using preferred statistical software, like R. We store CellProfiler measurements, like cell count, cell position, tissue surface area and other information in a database file (e.g. To verify identified objects, an image was saved with an overlay of each object using the ’OverlayOutlines’ and the ’SaveImages’ modules (blue outline for tissue, green outline for DAB positive nuclei, figure 3). Occupied area for the ’Tissue’ and the ’DAB’ objects were measured using the ’MeasureImageAreaOccupied’ module and exported to a.csv data file. Clumped objects were distinguished by intensity. The threshold correction factor for pixels was entered as ‘1.3’, the lower and upper bounds on the threshold were ‘0.5’ and ‘1.0’, respectively. An ‘Otsu Global’ with a ‘three classes’ threshold was used with pixels in the middle intensity class assigned to the foreground. Objects outside this range were discarded. Positive nuclei were defined as objects with diameters between 8 and 26 pixels. In this DAB channel, nuclei were identified using the ’IdentifyPrimaryObjects’ module. The ‘UnmixColors’ module extracted the brown DAB color as a separate grayscale channel. Positively stained nuclei were identified as follows.

Cellprofiler load time series multi page tiff manual#

We selected only objects larger than 40 pixels and with an absolute manual threshold of 0.03. Tissue was then identified as objects using the ’IdentifyPrimair- yObjects’. Tissue was defined using the gray-scale version of the image, applying a Gaussian blur with a 20 pixels diameter, and inverting the resulting image (using the ’ColorToGray’, ’Smooth’ and ’ImageMath’ modules respectively).

We created a pipeline to assess the tissue surface area and the number of positively stained nuclei.

Cellprofiler load time series multi page tiff how to#

An illustrated example on how to create pipelines in CellProfiler is described by Vokes and Carpenter.