Testing cervical model on dcm-brno dataset

[KaterinaKrejci231054]

This issue describes the application of the r20240523 model (for dorsal and ventral rootlets) on T2*w axial images (0.35 x 0.35 x 2.5mm) of C spine from the dcm-brno dataset (to be shared soon - #326). The model has been tested on a few subjects from this dataset.

Here are examples of its use on a healthy subject and a subject with compression:

Example of healthy subject:

• In this subject, the model segmented the rootlets quite well, but the spinal level order doesn't match - see below:

Example of a subject with compression:

• In this subject, the model does not preform well (spinal level does not match) - see below:

[valosekj]

Thanks for testing the model @KaterinaKrejci231054! The healthy subject does not look too bad at first glance! (especially considering the fact that the model was trained on T2w isotropic images (and not T2*w anisotropic ones)).

If we decide to extend (fine-tune?) the model for T2*w anisotropic images, we already have a lot of datasets to work with.