Fetal Brain Volume Reconstruction from Motion-corrupted Stacks Based on Hybrid Convolution Neural Network and Transformer

Motivation: The irregular fetal motion during MRI acquisition result in loss of structural continuity and corrupted 3D volumetric information. Goal(s): Herein, we propose a hybrid convolutional neural network (CNN) and Transformer based method for fetal brain MRI 3D volume reconstruction. Approach: Firstly, a coarse-to-fine CNN and Transformer based method is proposed to solve arbitrary fetal motion correction of 2D slices. Secondly, an initial 3D volume is estimated by point spread function. Finally, we propose an encoder-decoder model to reconstruct high-resolution fetal brain MRI 3D volume. Results: The experimental results demonstrate that the proposed method is effective and efficient. Impact: The proposed fetal brain MRI 3D volume reconstruction method based on CNN and Transformer can solve arbitrary motion correction of 2D slices and reconstruct high-resolution fetal brain MRI 3D volumes effectively and efficiently.