Skip to content

Latest commit

 

History

History
31 lines (21 loc) · 970 Bytes

README.md

File metadata and controls

31 lines (21 loc) · 970 Bytes

Spleen-Segmentation

3D & 2D Segmentation for Medical Decathlon Spleen Dataset

Target: Spleen
Modality: CT
Size: 61 3D volumes (41 Training + 20 Testing)
Source: Memorial Sloan Kettering Cancer Center
Challenge: Large ranging foreground size
Link to Dataset : https://goo.gl/QzVZcm

Results :
2D Segmentation (Best mean dice) : 96.03 %
3D Segmentation (Best mean dice for each fold of 5-fold cross validation) :
Fold 1 : 95.05 %
Fold 2 : 93.08 %
Fold 3 : 94.91 %
Fold 4 : 95.12 %
Fold 5 : 94.57 %

Final results Visualisation :

merge-from-ofoct


in collaboration with AIMedic studio start-up, Tehran, Iran