Algorithm 1: MR imaging modality transformation algorithm with CT as input.

Input: CT-MR dataset with corresponding relationship, initial model weight

Output: model weight W * after training

1: for i = 1 to n//n indicates the number of training sessions

2: Randomly read data from the corresponding CT-MR dataset $\{I_{CT}^i,I_{MR}^i\}~p_{data}(I_{CT},I_{MR})$

3: if i%3==0:

4: Update the parameters $D_{MR}$ and $D_{CT}$ of the discriminator in the model through the objective optimization function $L(G_{MR},G_{CT},D_{MR},D_{CT})$

5: End if

6: Update the parameters $G_{MR}$ of the generator and $G_{CT}$ of the inverse generator in the model through the objective optimization function $L(G_{MR},G_{CT},D_{MR},D_{CT})$

7: End for

8: Return generator $G_{MR}$