Help Improving Accuracy of Vessels Segmentation

[soumen02]

Hi everyone,

I'm using MONAI Label's deepedit app to segment vessels in a dataset of 21 patients. I trained the Dynunet network for 500 iterations and also tried increasing the number of epochs, but I'm not seeing a significant improvement in the results.
As you can see in the image below, the actual label is in red and the predicted label is in green.
I would appreciate any suggestions or feedback on how to improve the segmentation results.

Thank you in advance!

[Nic-Ma]

Hi @tangy5 ,

Could you please help share some comments about this ticket?

Thanks in advance.

[tangy5]

Happy to help, I think we already had a discussion in the MONAI Slack channel. Let's catch there.
Thank you!