[DAT]: AOMIC PIOP1 and PIOP2 #94

New issue

Closed

opened 2022-10-04 13:51:08 +00:00 by LeSasse · 3 comments

LeSasse commented

2022-10-04 13:51:08 +00:00

(Migrated from github.com)

Which dataset is it?

For the AOMIC dataset so far the ID1000 datagrabber is implemented by @verakye. We would like to use the PIOP1 and PIOP2 datasets as well through junifer.

Implementation

It can be implemented very similarly to Vera's AOMIC1000 datagrabber. I can implement the PIOP1 and PIOP2 datagrabbers based on that.

Dataset access restrictions

Public and open access (no registration)
Public and open access (registration required)
Restricted access (needs approuval)
Available only in specific locations (Juseless, Jureca)

Anything else to say?

No response

### Which dataset is it? For the [AOMIC](https://nilab-uva.github.io/AOMIC.github.io/) dataset so far the [ID1000 datagrabber](https://github.com/juaml/junifer/blob/main/junifer/datagrabber/aomic1000.py) is implemented by @verakye. We would like to use the [PIOP1](https://openneuro.org/datasets/ds002785) and [PIOP2](https://openneuro.org/datasets/ds002790/versions/2.0.0) datasets as well through junifer. ### Implementation It can be implemented very similarly to Vera's AOMIC1000 datagrabber. I can implement the PIOP1 and PIOP2 datagrabbers based on that. ### Dataset access restrictions - [X] Public and open access (no registration) - [ ] Public and open access (registration required) - [ ] Restricted access (needs approuval) - [ ] Available only in specific locations (Juseless, Jureca) ### Anything else to say? _No response_

LeSasse commented

2022-10-04 14:03:12 +00:00

(Migrated from github.com)

@fraimondo @synchon @verakye before i start working on these, there are two options:

Either I implement these datagrabbers as separate classes.
Or I take the existing AOMIC class and add a dataset parameter in init, so it can be initialised by name i.e. using "id1000", "piop1", "piop2", and then the individual elements and patterns depend on that initialisation.

Which option do you think is better/more consistent with junifer?

@fraimondo @synchon @verakye before i start working on these, there are two options: 1. Either I implement these datagrabbers as separate classes. 2. Or I take the existing AOMIC class and add a dataset parameter in init, so it can be initialised by name i.e. using "id1000", "piop1", "piop2", and then the individual elements and patterns depend on that initialisation. Which option do you think is better/more consistent with junifer?

fraimondo commented

2022-10-05 06:54:27 +00:00

(Migrated from github.com)

I think the best option would be option 1.

From this image, it seems that they are 3 independent datasets, each with different tasks and sequences. Having them all together will add parameters to the aomic1000 datagrabber. This is something we are trying to avoid with junifer.

Also, follow the same logic as with the HCP datagrabber. Tasks can be part of the constructor, but they are also part of the __getitem__ method.

Can you also do #95 on this one? It would be a good idea to create a package aomic under junifer.datagrabber and place the 3 aomic datasets there.

I think the best option would be option 1. From this image, it seems that they are 3 independent datasets, each with different tasks and sequences. Having them all together will add parameters to the [aomic1000](https://github.com/juaml/junifer/blob/main/junifer/datagrabber/aomic1000.py) datagrabber. This is something we are trying to avoid with junifer. Also, follow the same logic as with the HCP datagrabber. Tasks can be part of the constructor, but they are also part of the `__getitem__` method. Can you also do #95 on this one? It would be a good idea to create a package `aomic` under `junifer.datagrabber` and place the 3 aomic datasets there. ![figure_1](https://user-images.githubusercontent.com/4493699/193997932-e9126866-527d-4cbe-9bee-111f116e3124.png)

👍 1

LeSasse commented

2022-10-05 07:31:04 +00:00

(Migrated from github.com)

I think the best option would be option 1.

From this image, it seems that they are 3 independent datasets, each with different tasks and sequences. Having them all together will add parameters to the aomic1000 datagrabber. This is something we are trying to avoid with junifer.

Also, follow the same logic as with the HCP datagrabber. Tasks can be part of the constructor, but they are also part of the __getitem__ method.

Can you also do #95 on this one? It would be a good idea to create a package aomic under junifer.datagrabber and place the 3 aomic datasets there.

Yes, I agree with this after some thinking also. I will do #95 as well. Thanks for the quick response!

> I think the best option would be option 1. > > From this image, it seems that they are 3 independent datasets, each with different tasks and sequences. Having them all together will add parameters to the [aomic1000](https://github.com/juaml/junifer/blob/main/junifer/datagrabber/aomic1000.py) datagrabber. This is something we are trying to avoid with junifer. > > Also, follow the same logic as with the HCP datagrabber. Tasks can be part of the constructor, but they are also part of the `__getitem__` method. > > Can you also do #95 on this one? It would be a good idea to create a package `aomic` under `junifer.datagrabber` and place the 3 aomic datasets there. > > ![figure_1](https://user-images.githubusercontent.com/4493699/193997932-e9126866-527d-4cbe-9bee-111f116e3124.png) Yes, I agree with this after some thinking also. I will do #95 as well. Thanks for the quick response!