MM3 can use either deep learning or a traditional machine vision approach (Otsu thresholding, morphological operations and watershedding) to locate cells from the subtracted images. For info on the deep learning-based segmentation widget, see SegmentUNet.
The following four parameters are important for finding markers in order to do watershedding/diffusion for segmentation. They should be changed depending on cell size and magnification/imaging conditions.
OTSU parameters
first_opening_size: Size in pixels of first morphological opening during segmentation.distance_threshold: Distance in pixels which thresholds distance transform of binary cell image.second_opening_size: Size in pixels of second morphological opening.min_object_size: Objects smaller than this area in pixels will be removed before labeling.
The working directory is now:
.
├── 20170720_SJ388_mopsgluc12aa.nd2
├── TIFF
├── analysis
│ ├── time_table.pkl
│ ├── time_table.txt
│ ├── TIFF_metadata.pkl
│ ├── TIFF_metadata.txt
│ ├── channel_masks.pkl
│ ├── channel_masks.txt
│ ├── channels
│ ├── crosscorrs.pkl
│ ├── crosscorrs.txt
│ ├── empties
│ ├── segmented
│ ├── specs.yaml
│ └── subtracted