Filter images by plane resolution

[peristeri]

Feature

Be able to filter either through the admin or the REST api images by resolution.

The filter is broken down into 3 parts:

• Isotropic Resolution
  The filter is set in a range min and max
• Anisotropic Resolution with Axial Orientation
  In the RPI orientation, the in-plane and slice thickness if the 3rd value is larger of the other 2 values
• Anisotropic Resolution with Sagittal Orientation
  In the RPI orientation, the in-plane and slice thickness if the 1st value is larger of the other 2 values

[peristeri]

@jcohenadad to properly populate the resolution field, there must be a standard format. I suggest using the the one proposed by @charleygros:

Take the 4th, 5th and 6th of the image dimension in the `RPI` orientation.  ex: 

x, y, z = img.dim[3:6]
data.orientation = "{}x{}x{}".format(x, y, z)

[peristeri]

For the case of image grenoble_ofsep_02065SYST/t2_sag/t2_sag.nii.gz, the resolution is 0.74x0.74x4.40 and the orientation is sag. What would be the isotropic resolution, thickness and in-plane?

[jcohenadad]

"isotropic" means pX=pY=pZ. E.g. 1x1x1mm.

in sagittal data with 0.74x0.74x4.40mm resolution, the thickness is 4.4mm and the in-plane is 0.74x0.74.

what we call thickness is always the dimension that refers to the slice.

but: do we need that information from your web interface? don't you just report "orientation" and "resolution" in XxYxZ, with X, Y and Z defined by the slice orientation (not the physical coordinates of the scanner). I.e., in this example, the "Z" refers to the R-L direction.

[peristeri]

Filtering by plane resolution using thickness and in-plane has too many edge cases that is starting to complicate the query structure. Therefore, for this iteration, the filter will be the following:

To filter by resolution, the query will be structure: resolution={dx}x{dy}x{dz} where the values dx, dy and dz are the values of image.dim[4,6] when the image is at the RPI position.

To filter by sagittal from 1.0 to 5.0: sag=1.0-5.0
To filter by isotropic images: isotropic=true

The filter will not have any concept of thickness and in-plane

[peristeri]

@benjamindeleener :
The filtering by a range is possible. The user need to provide a upper and lower bound in the form of [lower]-[upper]. I use sagittal instead of x because the expected results will differ to one image to the other, since they might be saved in different orientation. Having combinations as you present them can make for a very complicated query to the database and with the limited time, I don't think the time to implement. An alternative is to make multiple queries for each combination and mage the results.

Finding datasets with sagittal and axial resolution less then 1.0mm and not isotropic ex: ?sagittal=0.0-1.0&axial=0.0-1.0&isotropic=False