-
-
Notifications
You must be signed in to change notification settings - Fork 101
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
New function: sct_deepseg_lesion #1850
Conversation
@charleygros looks like Travis is failing due to install problem, do you want to fix it before i review? |
@charleygros Logs (sct_download_data):
|
@jcohenadad : It should be fixed now. |
error with this commit fba8c86
probably because the model is not installed during SCT installation. It would need to be added. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
fba8c86
to
3796362
Compare
(The models should now be available, except if I missed something else) |
@abtahizadeh could you please create a test for this function before we merge it? in the same vein as the other testing functions that already exist. No need for integrity testing at this point (but we can open an issue to address it in the future). Thanks! |
@jcohenadad okay, please allow me some time to understand the function. |
* first implementation of sct_deepseg_lesion * add t2_ax contrast + threshold * models on OSF and neuropoly * add comma in sct_download_data * add deep_lesion_models to the list to download * installer: added download of deepseg_lesion models * TEST: added deepseg_lesion Former-commit-id: f4d917a
TODO
sct_deepseg_lesion
GH labelDescription of the Change
AVAILABLE CONTRASTS:
{t2,t2s,t2_ax}
-c t2_ax
--> was train on ~200 3D T2-w volumes with anisotropic resolution, axial orientation.-c t2
--> was train on 3D T2-w volumes with anisotropic resolution, both axial orientation and sagittal orientation.CENTERLINE DETECTION: 2 available ways to do it --> via the flag
-ctr
-ctr svm
--> OptiC initialized with SVM+HoG maps (former version, currently the one integrated in master). Avantages: faster, more robust to exotic / new contrasts.-ctr cnn
--> OptiC initialized with CNN_1 maps. Avantages: more robust to compressed cord and/or sagittal images.DEALING WITH THE PRESENCE OF THE BRAIN:
-brain 0
--> The algorithm will go through all the axial slices. Currently the version running in master. Note that this option is always set with -ctr svm.-brain 1
--> If brain is detected (using the CNN1 previously mentioned), the segmentation algorithm will ignore axial slices above the brainstem. PROS: (i) speed up the segmentation (especially for sagittal images), (ii) Less false positive 'blobs' in the brain. CONS: The segmentation could sometimes stop below the brainstem.Steps and Constraints
/!\ Please re-install the toolbox: data folder