Our poster titled Neonatal Brain MRI Motion Correction using Adult MRI has been presented to ISMRM 2024.
Our poster titled Neonatal Brain MRI Motion Correction using Adult MRI has been presented to ISMRM 2024.
Abstract
Our motion correction algorithm is aimed toward neonatal brain Magnetic Resonance Imaging (MRI). This would benefit researchers and clinical practitioners in overcoming motion artifacts in neonatal scans. Neonatal brain MRIs are frequently compromised by motion artifacts, resulting in low-quality outputs and scan interruptions. While popular Deep Learning (DL) models like UNets trained on adult datasets, perform well in artifact reduction tasks, they are less effective for neonatal scans due to the significant domain shift and the scarcity of high-quality neonatal datasets. To address these challenges, we simulated motion artifacts (random head rotations and translations) on both adult T1-weighted (T1w) scans and neonatal T2-weighted (T2w) scans. This approach is feasible due to the contrast similarity between adult T1w and neonatal T2w scans. By training two instances of the same model on both datasets, we achieved comparable results, demonstrating the effectiveness of our method for motion correction in neonatal MRI scans.