getmor_segmentation

GeTMoR (GEnomic, Transcriptomic, and MOrphological profiling of Rare cells) is a protocol for or detecting rare cancer related cells to simultaneously image and profile the genome and transcriptome from single rare cells. This repo contains the code that we used for image analysis of scanned Immunofluorescence slides.

Dependencies and compilation

1. We make extensive use of the ITK toolkit for image analysis. Download and install the ITK toolkit:

2. Set the ITK_DIR variable in src/CMakeLists.txt

3. Create a build director:

mkdir build; cd build

4. Create makefiles using:

cmake <path_to_rerp>/getmor_segmentation/src/

5. Compile the code with:

make

Procedure for GeTMoR rare cell detection

The parameters for the programs below are based on the assumption that the cells in the cytokeratin (CK) channel are rare and so at most 3 cells are present in a frame. If this assumption is not met, the parameters need to be changed accordingly to achieve high sensitivity of detection.

It is assumed that the scanned image file names are of the format Tile\%06d.tif, that 2304 images are generated per channel, and that the different channel images are named sequentially after each other. If differing number of frames are imaged, adjust the frame offsets in the commands accordingly.

• Run the command below in a cmd prompt window after 20% of the frames are scanned on the CK channel to segment the cell:

 ck_segment <scan_output_dir> <segmentation_dir> <sample_name>
 1 2304 2305 5 0.995 0.3 3 1

where scan_output_dir is the path to the directory containing scanned images, segmentation_dir is the output directory to store the segmented images, and sample_name is an identifier for the sample.

• Run the command to extract cell features:

$ feature_extraction <scan_output_dir> <segmentation_dir>
<sample_name> 1 2304 <channel_start>

where channel_start is a comma separated list of channel start offsets to user for feature extraction. This should be set to "1, 2305" when scanned on DAPI and CK channel and to "2305" when scanned only on the CK channel.

• Run the command to filter cells based on feature values:

$ filter_features <sample_name>_feature_vec.txt
<sample_name>_feature_vec_filt.txt <col(1-based),min,max>*

where <col(1-based),min,max>* specifies the column in the feature vector file and the minimum and maximum values in that column to retain. To filter events corresponding to typical cell sizes use "5,100,10000". In addition, filters can also be used for DAPI and CK mean intensities.

• Run the command to visualize and confirm all the segmented events:

$ cell_image <scan_output_dir> <segmentation_dir>
<sample_name>_feature_vec_filt.txt "1,1,0,0" "2305,1,0,0" 150

Citation

Contact information

Rishvanth K. Prabakar kaliapp@cshl.edu