Accurate segmentation of the right ventricle (RV) base in cardiac MRI is challenging due to complex anatomy and interplanar motion. This paper introduces a novel deep learning method that leverages motion tracking uncertainty ("loss-of-tracking") and refined annotations to improve RV base segmentation accuracy and reproducibility.
本稿では、心臓MRIの右室基部におけるセグメンテーションの精度と再現性を向上させるため、時間的な非一貫性を利用した新しい深層学習手法を提案しています。
This research proposes an improved 3D UNet model for automated segmentation of the left ventricle (LV) and myocardium in cardiac MRI, emphasizing the exclusion of papillary muscles for accurate clinical parameter estimation, and demonstrating its superior performance compared to existing methods.