Skip to content

Latest commit

 

History

History
13 lines (7 loc) · 677 Bytes

File metadata and controls

13 lines (7 loc) · 677 Bytes

Medical-Segmentation-decathlon

Medical Segmentation Decathlon: Effective VNet-based 3D Segmentation Model of the Liver

Project overview;

(1)Developed a 3D segmentation algorithm pipeline based on MONAI with VNet model and generalised sieve loss function

(2)Optimised the loss function by combining focus loss to achieve test time enhancement to improve the segmentation performance of small tumours

(3)Combine UNet and HighResNet to construct a segmentation consensus model to achieve a final segmentation Dice coefficient of 0.81 for more accurate and stable prediction

Note: Associated PDF files for the essay on the results of the competition tasks