Thank you for providing the code. I have some questions regarding DisQ.

[celi7]

1. What preprocessing was applied to the training and testing data? Was affine registration applied before DisQ?
2. What do the variables swap_img, theta_a, theta_b, mu_a, mu_b, amu_a, amu_b, beta_a, and beta_b mean in the python file eval_disen_t1_affine.py?
3. How to save images with different contrasts after motion correction by DisQ in the file eval_disen_t1_affine.py?(I checked the swap_img and it appears to be images with the same contrast after motion correction.) Looking forward to your reply, thank you!

[Changchun-Yang]

1. We did not perform affine preprocessing because our task is focused on sequence MRI data. We remove bad data points and record the maximum value of a sequence during normalization, which allows us to map back to absolute values for later quantification.
2. theta_a, theta_b, mu_a, mu_b, amu_a, amu_b, beta_a, and beta_b mean contrast and anatomy of A,B and corresponding mean of them. swap_img means generated images by difference contrast and anatomy, such contrast from A, anatomy from B.
3. We are aiming to disentangle the images to generate new with same contrast, which will reduce the adverse effects caused by contrast on registration. We can register images with different contrasts using the displacement vector field obtained from images with the same contrast, because the anatomy is the same.

[celi7]

1. We did not perform affine preprocessing because our task is focused on sequence MRI data. We remove bad data points and record the maximum value of a sequence during normalization, which allows us to map back to absolute values for later quantification.
2. theta_a, theta_b, mu_a, mu_b, amu_a, amu_b, beta_a, and beta_b mean contrast and anatomy of A,B and corresponding mean of them. swap_img means generated images by difference contrast and anatomy, such contrast from A, anatomy from B.
3. We are aiming to disentangle the images to generate new with same contrast, which will reduce the adverse effects caused by contrast on registration. We can register images with different contrasts using the displacement vector field obtained from images with the same contrast, because the anatomy is the same.

Thank you for your response！ With regard to your response 3, do you mean that swap_img is merely a set of images with consistent contrast but lacking strict alignment, thus requiring further registration for motion correction? If so, may I inquire as to what registration method you utilized in the article for achieving additional alignment?