Automated reconstruction of 3D+time fetal cardiac MRI from stacks of Doppler-gated slices

Motivation: Visualization of the fetal cardiac dynamics is essential for prenatal diagnosis of congenital heart diseases. Recently, 3D+time MRI was reconstructed from Doppler-gated 2D acquisitions, but this reconstruction requires lengthy manual input and long processing time. Goal(s): We aim at developing a novel automatic and computationally efficient approach combining Doppler-gated cine imaging and motion-corrected slice-to-volume reconstruction (SVR). Approach: We propose an automatic deep learning segmentation module of the thorax and a new 3D+time SVR algorithm specifically adapted to Doppler-gated sequences. Results: The segmentation network produced robust predictions. The proposed reconstruction algorithm is ten times faster than the original approach while preserving image quality. Impact: This preliminary study opens new potential for the use of Doppler-gated imaging in conjunction with SVR for 3D+time fetal cardiac imaging in a clinical context through the development of a fast, reliable, and automated reconstruction pipeline.