How can I use this tool if I already have a segmentation file?

[jcohenadad]

I already have a segmentation file. Yet, the README only explains what to do if starting from an image.

Moreover, the project is called "Segmentation to MR Properties Converter", so from the title, one would expect to already have a segmentation. I strongly recommend refactoring the README, to adapt it to the case where ppl have a segmentation.

Then, create another section that explains how to start from an image.

Also, it would be important to indicate what value each class of the segmentation corresponds to each tissue (fat, liver, etc.). I suggest adding it on https://github.com/shimming-toolbox/tissue-to-MRproperty?tab=readme-ov-file#relaxation-values--susceptibility-v10

[sriosq]

I will create a table where I show the relationship label id - label name and replace the need for a segmentation tool to be specified. This way any segmentation file would work. I'll take out all the unnecessary code.
Thank you

[jcohenadad]

I will create a table where I show the relationship label id - label name

Why not simply use the existing table and simply add a column, instead of creating another table?

Also, please do it in a PR so somebody can review the code before pushing to master

[sriosq]

Yes. Another column on the already existing table.
I will create a branch and request a PR.

[sriosq]

Hello prof. @jcohenadad, given that different segmentation tools (eg. moose, totalsegCT, totalsegMRI, etc) have different id#-label relationship (i.e., one tool may have label 13 to be heart when another has it as brain) I think the current code adapts to this scenario. I can update the readme file to explain this in a much clearer way.
If we take out we would need to state in the readme file that anyone using this tool would need to manually edit their segmentation file, on the other hand, I can state in the readme that anytime a new tool would like to be added its as simple as giving us the segmentation tool id#-label list.
Let me know what suits better for this repository,

Thank you

[jcohenadad]

You need to impose a UNIQUE numerical ID for fat, heart, etc., and make it clear in the table what this ID is.

Separately (in this repos, or in another repos), you can have conversion tool to convert labels from segmentation software X to labels for this repos.

[sriosq]

I can separate the current code into 2 different clis: the first recieves the segmentation tool used, if it finds it in its memory then it converts into our UNIQUE labels dictionary, then the second cli can be used. If the first cli fails to detect the tool then the user is prompted to create or request adding to the dictionary. This will be explained in the readme too.

[jcohenadad]

I'm not sure I understand exactly what you suggest-- do a PR and I'll review. Also tagging @mathieuboudreau for a 2nd pair of eyes

[mathieuboudreau]

DIscussing with @sriosq right now - my recommendation would be to have a feature added here where the user can input a file that associates "unique_label_name" to "int_label_value_in_user_volume", and then internally convert the user's labels to the MR prop values by using the unique label names. The table @jcohenadad linked should indeed have a new column that should be a descriptive and unique string that defines each possible labels (eg sc_wm ), and then have the new column describe it (Spinal WM).

The CLI should have three options, 1) the user provides an input file with a custom label map (tissue2MRProp unique identifier -> label value that user chose for their volume beforehand), 2) the user provides a flag if they are using a standard label dataset (eg MNI152), or if not use the default defined label mapping for this repo (which, actually should be a second column added now that I look at it).

[jcohenadad]

I'm sorry but I still don't understand-- why creating a 2nd column?

I think the main purpose of the present repos is to provide a tool that does "tissue" to "MR property", tissue being defined by a ID.

Example:

Tissue	ID	T1 (ms)	T2 (ms)	T2* (ms)	PD (%)	Susceptibility (ppm)
fat	01	401.2	129.3	64.65	20	-8.92
heart	02	1215.67	49.35	25.195	77	-9.05
liver	03	798.75	33	18.82	70	-9.05

And that's it.

Now, if ppl want to use whatever tool that generates a label file with different ID, then there should be another CLI that deals with it. That other CLI could be in this repos, but I think it should be another CLI, eg: convert_custom_label.py

[sriosq]

The second column M.B. was referring to would be used to explain what some segmentation labels have as "tags", in his example sc_wm could be hard to understand for some people, so that column would be used to explain sc_wm means spinal cord white matter.
Apologies if my comment above (#24 (comment)) was unclear, but it was in the lines of what you suggest, having 2 CLI.

[jcohenadad]

The second column M.B. was referring to would be used to explain what some segmentation labels have as "tags", in his example sc_wm could be hard to understand for some people, so that column would be used to explain sc_wm means spinal cord white matter.

why not simply using the first column for that and make it explicit instead? The purpose of the first column is to be descriptive, so let's make it descriptive