fidoptAllen hiPSC single-cell image dataset, bright field to DNA, 3D
One field of view of live gene-edited WTC-11 human induced pluripotent stem cells in 3D: the transmitted-light bright-field stack (source) and the Hoechst-stained DNA stack (target) of one acquisition (Allen Institute for Cell Science, hiPSC single-cell image dataset).
Read before comparing (9)
- Pool. Fields of view of live gene-edited WTC-11 hiPS cells of the Allen Cell Collection, 412 plates in metadata.csv imaged at 100x on Zeiss spinning-disk microscopes from 2017 to 2020, those of the pool from 2017-02 to 2018-07. One field a plate, its lowest FOVId, so no two cases share a culture: 328 plates once the 22 of AllenLabelFree_Syn and the 62 'alignV2' plates are left out, in 18 cell lines (each a tagged structure), the draw stratified by line.
- Left out: alignV2. From October 2018 the files are 'alignV2' TIFFs (600 x 900 pixels, 4 channels): there the bright field is taken by one camera and the Hoechst by another, the second camera's image brought onto the first by the provider's alignment, so the pair is registered rather than taken on one grid. The seven lines imaged only then (ATP2A2, HIST1H2BJ, NUP153, RAB5A, SLC25A17, SMC1A, SON) are therefore not in the pool.
- Images. Each field is 624 x 924 pixels of 0.108 µm, 43 to 82 planes 0.29 µm apart, written in micrometres as if millimetres; the CZI-converted OME-TIFF holds 7 or 4 channels, all taken by one camera, and prepare.py takes the bright field (Bright_100, Bright_100X or TL 100x) and H3342 by their OME names. Four of the 60 files are named '10X' (three test, one development), but their first image is the 100x acquisition, with the same voxel size.
- Values. Both stacks are uint16 camera counts as delivered. The DNA target sits on a camera offset near 410 (its median 401 to 423 on the test fields), the nuclei reaching 435 to 811 at the 99.9th percentile and up to 10,465; truth.dynamic_range, 573, is that percentile pooled over the 30 test fields. The bright field's level differs by plate, its median from about 17,000 to 43,000.
- Alignment. Both stacks come from one acquisition by one camera, nothing is registered.
- Mask. The whole field: a mask made from the target would give the methods the nuclei's places.
- Baseline. On the whole field of the 30 test cases, at a range of 573, a constant at the target's median scores an MAE of 10.7 (5.1 to 17.9), a PSNR of 30.9 dB (23.4 to 38.4) and an SSIM of 0.79 (0.67 to 0.88), means over the cases: the background between the nuclei is most of the field. The bright field as it is, on another scale, scores nothing.
- Outside numbers. The label-free models of the Allen Institute were trained on fields of this collection, possibly these; no published number is on this draw.
- Overlap. AllenLabelFree_Syn's 80 lamin B1 fields come from the same pipeline, 71 of them in this dataset under the same file names: its 22 plates are left out of the pool, so no field or culture is in both.
No method has a result on every case of this dataset yet.
Case by case
Click a score to look at that result: the target image, and over it the method's synthetic image. Underlined scores have their image; for the others you see the case without the result.
About the data
- Test cases
- 30 test fields drawn with seed 2026 by cell line among the 298 outside development, then 30 development fields, the last of the same seeded order, the lines in proportion. One field a plate, the plate's lowest FOVId; the 22 plates of AllenLabelFree_Syn's lamin B1 fields and the 62 'alignV2' plates are left out. The dataset gives no test set.
- Truth
- target image as released the field's Hoechst 33342 stack (channel H3342) as delivered, taken with the bright-field stack in one acquisition on one spinning-disk microscope, on the same grid, not registered
- Source
- open.quiltdata.com/b/allencell/packages/aics/hipsc_single_cell_image_dataset
allencell.s3.amazonaws.com/aics/hipsc_single_cell_image_dataset/README.md
www.allencell.org/terms-of-use.html
doi.org/10.1038/s41586-022-05563-7 - Licence
- Allen Institute for Cell Science terms of use: research and other non-commercial use, copies and derivative works allowed, with its citation policy and its 'freedom to innovate' clause
30 cases for tuning, never tested
3500000839_100X_20170417_F06_P35, 3500000680_100X_20170307_F04_P27, 3500001121_100X_20170726_2-Scene-03-P27-F04, 3500001988_100X_20180430_3-Scene-14-P55-F04, 3500001560_100X_20171120_1e-Scene-10-P28-F04, 3500001502_100X_20171106_2-Scene-20-P40-F06, 3500000961_100X_20170609_2-Scene-27-P48-F08, 3500001856_100X_20180326_4-Scene-1-P11-E07, 3500001597_100X_20171219_2-Scene-1-P3-F08, 3500002055_100X_20180522_3c-Scene-07-P21-E05, 3500001153_100X_20170804_2-Scene-10-P40-F07, 3500001345_100X_20170926_4-Scene-03-P27-E08, 3500001250_100X_20170829_3-Scene-07-P25-F05, 3500001859_100X_20180326_2e-Scene-11-P52-F04, 3500001827_100X_20180319_1-Scene-05-P12-E06, 3500001464_100X_20171024_5-Scene-7-P33-E07, 3500001917_100X_20180409_2-Scene-07-P17-E06, 3500001022_100X_20170630_1-Scene-5-P12-E05, 3500001309_100X_20170919_16_184_1-Scene-02-P6-E04, 3500000722_100X_20170317_E04_P08, 3500001093_100X_20170718_2-Scene-1-P20-E08, 3500002002_100X_20180504_2-Scene-01-P22-E07, 3500000989_100X_20170619_3-Scene-2-P5-E04, 3500001541_100X_20171114_1-Scene-07-P8-E05, 3500001519_100X_20171110_1-Scene-15-P45-F05, 3500001034_100X_20170705_3e-Scene-3-P17-E06, 3500001850_100X_20180323_2-Scene-04-P35-F04, 3500001610_100X_20180112_1-Scene-15-P45-F07, 3500000645_10X_20170301_E07_P27, 9999999314_100X_20180514_1-Scene-17-P32-F04