About the visual display of Figure 3

[Halleyawoo]

Hello, how do you get the results of the whole heart segmentation task compared with other methods in the SIFA paper (Figure 3)? Re-run other people's methods? I wonder if you can provide other methods and the visual display of SIFA experimental results? Many thanks!

[cchen-cc]

The trained model of SIFA can be downloaded from
https://www.dropbox.com/sh/787kmmuhvh3e3yb/AAC4qxBJTWwQ1UMN5psrN96ja?dl=0.
With the trained model, you can get the predictions of SIFA for the visual display.

Note that for adaptation from cardiac MR to CT, set USE_newstat=False to get the same results as in the paper. This is because we re-calculate the data statistics before releasing the dataset. Using new data statics to train can get better results than our reported models, but to evaluate our trained model, we need to use the previously calculated one.