You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Experiments into segmentation of transversal slices of cardiac MR images
3
+
4
+
## Initial model
5
+
Trained using fastai v1 on 138 images (90 training, 48 validation) from 6 volunteers. Classes to predict were left ventricular cavity and left ventricular myocardium.
6
+
7
+
The provided model file (see release page) `b0_transversal_5_5.pth` is the base model from
8
+
> A Deep Learning Based Cardiac Cine Segmentation Framework for Clinicians - Transfer Learning Application to 7T Markus J Ankenbrand, David Lohr, Wiebke Schlötelburg, Theresa Reiter, Tobias Wech, Laura Maria Schreiber medRxiv 2020.06.15.20131656; doi: https://doi.org/10.1101/2020.06.15.20131656
9
+
10
+
trained for 5 epochs frozen (max_lr=1e-3) and 5 epochs unfrozen (max_lr=1e-4)
0 commit comments