fidoptBioSR wide-field to linear SIM, clathrin-coated pits, 2D
A 2D wide-field fluorescence image of clathrin-coated pits in one cell at BioSR's lowest signal level as source, and the linear SIM image of the same field as target (BioSR, Qiao et al., Nature Methods 2021).
15 test casescellsource wide-field, target SIM2D, distances in µm
Read before comparing (10)
- Structures. BioSR's four structures are four datasets (BioSR_CCP_Syn, BioSR_ER_Syn, BioSR_MT_Syn, BioSR_Actin_Syn), each cell in one zip of the deposit. BioSR's models (DFCAN and the paper's other networks) are trained one a structure, and a method gets only the image: on a mixed set no method could pick its model.
- Source. The mean of the cell's nine raw SIM images (three angles, three phases) at signal level 1, the noisiest, as the provider makes a wide-field image (DL-SR): a low-signal image to denoise and sharpen, as in BioTISR_Syn. It stays in camera counts, with the camera's offset of about 100.
- Target. The SIM image the provider reconstructed from the cell's high-signal raw images: 1004 x 1004 pixels of 31.3 nm, micrometres written in the headers as if millimetres. The provider scaled it to a peak of 65535; it is divided by its largest value, as nothing in the source predicts its brightness. It is 93 to 98 % zeros, the background set to 0.
- Alignment. The source is put on the target's grid by linear interpolation, its pixel k on SIM pixel 2k, as for BioTISR_Syn. On every case the shift left between the two is at most 0.31 SIM pixel; prepare.py refuses a case over one pixel.
- Mask. The cell's labelled structure: the SIM image smoothed by a Gaussian of 0.3 µm, above Otsu's threshold, grown by 1 µm, as for BioTISR_Syn. It is made from the target and given to every method. 86 to 96 % of it is zero in the target, so the target's median there is 0.
- Baseline. Inside the mask on the 15 test cases, the zero image, a constant at the target's median, scores an MAE of 0.012 (0.007 to 0.025), a PSNR of 24.8 dB (21.5 to 27.8) and an SSIM of 0.81 (0.70 to 0.88), so SSIM is shown, not ranked. Beating it takes the pits in the right place and nothing elsewhere.
- Least squares. The source mapped linearly onto the target by least squares inside the mask, which no method can do, scored on the first ten test cases an MAE of 0.022 (0.015 to 0.042) and 27.7 dB: worse than the zero image in MAE, better in PSNR.
- ccp_050. Its level-1 wide-field image is nearly all noise (band-passed correlation with the SIM image 0.28, against 0.86 for ccp_036): the pits are barely seen. It is kept, a provider test cell not chosen by eye with a sound target, but it weighs on the means: read it per case.
- Overlap. No other dataset of the corpus holds BioSR images. BioTISR_Syn comes from the same laboratory (Dong Li, Institute of Biophysics, Beijing) but is another deposit: 3D-SIM time-lapses of live COS-7 cells on Zenodo, other cells, no file shared.
- Public. The provider's code trains on cells 6 to 35 and validates on 1 to 5, but all of BioSR is public with its targets: a model trained on all of it has seen the test cases.
No method has a result on every case of this dataset yet.
| Accuracy | Cost | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| # | Method | MAE | Change | PSNR µm | SSIM * | Time s | RAM GB | GPU GB | Images |
| Source as given | 104.8 ±0.9 | -40.42 ±0.08 | 0.000 ±0.000 | ||||||
* shown, not ranked: on this dataset it is checked against the output of a program.
MAE: mean absolute error of a synthetic image against its target, inside the body, in the image's units. PSNR: the same by signal-to-noise ratio, higher is better. SSIM: the same by structural similarity, 1 is perfect.
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
The first test caseClick to look at every case.
- Test cases
- The provider's 15 test cells of CCPs.zip taken whole (Cell_036 to Cell_050, the default test_cell_list of the DL-SR repository's DataAugumentation_ForTest.m), and 10 development cells drawn among its training and validation cells (Cell_001 to Cell_035). One cell is a case, at signal level 1.
- Truth
- target image as released the provider's linear SIM image of the cell (SIM_gt.mrc), reconstructed from the cell's high-signal raw SIM images, divided by its largest value
- Source
- figshare.com/articles/dataset/BioSR/13264793
doi.org/10.6084/m9.figshare.13264793.v9
doi.org/10.1038/s41592-020-01048-5
github.com/qc17-THU/DL-SR - Licence
- CC BY 4.0 (the figshare record's licence field)
10 cases for tuning, never tested
ccp_004, ccp_026, ccp_006, ccp_035, ccp_017, ccp_023, ccp_022, ccp_016, ccp_003, ccp_032