# The hiPSC single-cell image dataset
​
**Maintainer**: Jianxu Chen (jianxuc@alleninstitute.org) and Kim Metzler (kimberly.metzler@alleninstitute.org)


**Usage**

If you would like a deeper understanding of what all you can do with Quilt, please refer to this [documentation](https://docs.quiltdata.com/). Code examples and demos of how to use this package in different scenarios can be found at https://github.com/AllenCell/quilt-data-access-tutorials


The package has the following hierarchy:

```
(remote Package)
 └─README.md
 └─crop_raw/
   └─447101df_raw.ome.tif
   └─c217c356_raw.ome.tif
   ...
 └─crop_seg/
   └─27a34691_segmentation.ome.tif
   └─b25ec91e_segmentation.ome.tif
   '''
 └─fov_path/
   └─022e902b_3500000635_100X_20170227_F07_P23.ome.tiff
   └─10fbc8ee_3500000635_100X_20170227_E07_P07.ome.tiff
   '''
 └─fov_seg_path/
   └─0fa7a17f_3500000635_100X_20170227_F07_P23_CellNucSegCombined.ome.tiff
   └─15d8db4e_3500000635_100X_20170227_E07_P07_CellNucSegCombined.ome.tiff
 └─struct_seg_path/
   └─0ea40dcb_3500000635_100X_20170227_F07_P23_struct_segmentation.tiff
   └─1caaa3ae_3500000635_100X_20170227_E07_P07_struct_segmentation.tiff
 └─metadata.csv
```

The dataset is organized by `metadata.csv`, which has the following columns:
* CellId: a unique identifier of each cell (match with the CellId in cfe.allencell.org)
* FOVId: a unique identifier of the FOV image this cell is segmented from
* structure_name: the cell line this image was acquired from
* crop_raw: the file path to access cropped single cell image (raw image), which is an unsigned 16-bit multi-channel 3D image. The name of each channel can be found in the column `name_dict`.
* crop_seg: the file path to access cropped single cell image (segmentation), which is an unsigned 8-bit multi-channel 3D image (pixel value = 0 or 255, 255 indicates the segmentation mask). The name of each channel can be found in column `name_dict`.
* name_dict: a dictionary of channel names in `crop_raw` and `crop_seg`. The python dictionary can be programmatically extract by the `eval()` function. 
* roi: the ROI position of this cell in the full field-of-view (FOV)
* scale_micron: the size (in micrometer) of a single voxel along each dimension
* fov_path: the path to the FOV image this cell is segmented from.
* fov_seg_path: the path to the cell and nuclear segmentation of the FOV image this cell is segmented from. The segmentation is an unsigned 8-bit multi-channel 3D image. Pixels values are sequetial integers from 0 to the total number of cells in this FOV. Each integer represents one unique cell. The image contains 4 channels: nuclear segmentation, cell segmentation, contour of nuclei segmentation, and contour of cell segmentation.
* struct_seg_path: the path to the structure segmentation of the FOV image this cell is segmented from, which is an unsigned 8-bit single-channel 3D image (pixel value = 0 or 255, 255 indicates the segmentation mask).
* this_cell_nbr_complete: a boolean value indicating if all the direct neighboring cells in the colony are successfully segmented and included in the dataset
* this_cell_nbr_dist_2d: a dictionary of the distance (2D euclidean distance in XY dimension) from this cell to its neightbors in the dataset.
* edge_flag: a boolean value indicating if this cell is on the edge of a colony
* this_cell_index: the index of this cell (i.e., the integer value in the segmentation image) in this FOV.
* cell_stage: a label indicating the cell cycle stage of this cell, which could be "M0", "M1M2", "M3", "M4M5", "M6M7_single" or "M6M7_complete". See more details about M0, M1, ..., M7 at https://www.allencell.org/hips-cells-during-mitosis.html. Here, M6M7_single vs M6M7_complete indicate whether the two daugher cells are both included in this cell. These labels were generated by the combination of rule-based criteria and a deep learning based classifier. See more details and the training data used to build the deep learning based classifier here: https://open.quiltdata.com/b/allencell/packages/aics/mitotic_annotation.


* There are several columns about meta data: "PlateId", "WellId", "meta_fov_position", "meta_imaging_mode", "meta_fov_outside_overview", "meta_fov_xcoord", "meta_fov_ucoord", "meta_fov_edgedist", "meta_colony_label", "meta_colony_centroid", "meta_colony_area","meta_plate_bad_segmentation", "meta_plate_confluency", "meta_well_passage_at_imaging", "meta_well_passage_at_thaw", "InstrumentId", "WorkflowId", "ProtocolId", "PiezoId". Please find the details of all these meta data in paper supplementary information.
* There are a few columns of single cell features: 'dna_position_depth_lcc', 'dna_roundness_surface_area_lcc', 'dna_shape_volume_lcc', 'mem_position_depth_lcc', 'mem_roundness_surface_area_lcc', 'mem_shape_volume_lcc', 'str_shape_volume', 'NUC_shcoeffs_L0M0C_lcc', ..., 'MEM_shcoeffs_L16M16S_lcc', 'NUC_MEM_PC1' ..., "outliers". These values can be used to reproduce all the figures in the paper. See the paper for more details.                                                                                                                                                                                                                                                                                                                                                                                                                                                          
### Usage
For documenation on how to use and interact with this dataset please refer to [https://docs.quiltdata.com/walkthrough/reading-from-a-package](https://docs.quiltdata.com/walkthrough/reading-from-a-package).
### License
For questions on licensing please refer to [https://www.allencell.org/terms-of-use.html](https://www.allencell.org/terms-of-use.html).