As part of the multipass segmentation and classification system, described in the MaSS documentation and elsewhere in the AstroPathPipeline documentation, multiple segmentation algorithms can be used on a single panel for cells of different sizes. The cell tables are merged as part of the MaSS protocol here we use the information from the cleaned_phenotype_tables to select the resultant cell geometries from the binary_seg_maps. The geometeries are then saved in separate layers with the component_data into a component_data_w_seg file in the Component_Tiffs folder of the AstroPathPipeline directory structure. This module should be run after the classification and segmentation of slides have been verified complete according to the protocols laid out in the QA QC step of the inform_processing module. The module is set to run once across all cohorts after being launched, it checks the following conidtions before running on a set of slides:
- That the cleaned_phenotype_table files exist
- That the component_data_w_seg files exist
- If they do exist, that they were created after the last cleaned_phenotype_table file was created
The code should be launched through matlab. To start download the repository to a working location. Next, open a new session of matlab and add the AstroPathPipline to the matlab path. Then use the following to launch:
segmaps(<Mpath>)
<Mpath>[string]: the full path to the directory containing the AstropathCohortsProgress.csv file- description of this file can be found here
The segmentation maps, component_data_w_seg, start with the component tiff layers from the component_data files, the number of layers correspond to the layers added for unmixing in inForm. For a standard 7-color slide unmixed by inForm this would be:
- DAPI
- Opal 520
- Opal 540
- Opal 570
- Opal 620
- Opal 650
- Opal 690
- Autofluorescence
The next layers correspond to the segmentation label images from inForm. The layers start with the tissue segmentation, followed by the nuclear segmentation for each segmentation type and the membrane segmentation for each segmentation type. The segmentation types are defined in the <SegmentationStatus> column of the MergeConfig_NN.xlsx files. The layers in the segmentation maps are added in the same numeric order of the <SegmentationStatus> column, such that the segmentation map layers for a panel with two segmentation types and a 7-color panel would be as follows:
- 9: Tissue Segmentation
- 10: Nuclear:
<SegmentationStatus>1 - 11: Nuclear:
<SegmentationStatus>2 - 12: Membrane:
<SegmentationStatus>1 - 13: Membrane:
<SegmentationStatus>2
The values in the segmentation layers are set up as a label matrix with the values corresponding to the <CellID> column in the cleaned_phenotyped_tables.csv files. For example, in the nuclear layer all values with 1 correspond to <CellID>: 1; all values with 2 correspond to <CellID>: 2; and so on.