- 2024-04-15: version 1.0.0, first release.
- 2024-04-16: version 1.1.0, add demos.
The following videos are the results from a single experiment. (P.S.: w3_ND16_2024-04-10_23-28-30
in our server).
Check full videos of this experiment on YouTube.
up-stream.m
.tif
->is_outlier.mat
,.mp4
,intensity.mat
down-stream.m
-
2
is_outlier.mat
->is_outlier_union.mat
-
intensity.mat
,is_outlier_union.mat
->intensity.mat
,intensity_volume.mat
-
2
intensity_volume.mat
-> figures of red and green channel
Data
│
├── w1
│ ├── 0_Camera-Red_VSC-10629
│ │ └── *.tif
│ └── 1_Camera-Green_VSC-09321
│ └── *.tif
│
├── w2
│ ├── 0_Camera-Red_VSC-10629
│ │ └── *.tif
│ └── 1_Camera-Green_VSC-09321
│ └── *.tif
│
├── w3
│ ├── 0_Camera-Red_VSC-10629
│ │ └── *.tif
│ └── 1_Camera-Green_VSC-09321
│ └── *.tif
...
When using up-stream.m
, select a folder of the w1
level.
When using down-stream.m
, select a folder of the Data
level.
Answer: Use binarization.
note:
-
How to detect outliers? (Don't need fine-tune)
- Tukey test of the number of bright pixels of certain binary frame. IQR_index = 1.5
- Tukey test of the intensity of bright pixels of certain binary frame. IQR_index = 1.5
- Tukey test of the
intensity_volume
of certain volume. IQR_index = 3 - Tukey test of the diff or ratio. IQR_index = 3
-
How to deal with outliers?
- make them to be nan.
Answer: Use opening.
super-parameter
- sense
- disk_r
note:
- Usually,
$sense = 0.2$ is OK. In other words, you don't need to fine tune this super-parameter. -
$disk_r$ is needed to changed. You must tune it to split the soma and the neurite of the template channel.$disk_r \in {3,4,5,6,7,8}$
Requirements
- for the soma template: the soma and the neurite are easy enough to be split.
- for the neurite template: the neurite is brighter than the other channel.
- for the all template: the soma and the neurite are brighter than the other channel.
For example, in our taxis project in 2023/11, the green channel is brighter than the red channel, so
- the red channel is used as the soma template.
- the green channel is used as the neurite template.
- the green channel is used as the all template.
Use opening to remove the halo of the soma after digging out it from the original BW.
First, we must have a template for the whole neuron, otherwise, the affect of noise will be much more severe.
I use this criteria: in certain frame, for certain worm, choose the channel which has more bright pixels!