A Method to Improve the Reliability of Pulmonary Perfusion-weighted Maps Obtained Using 3D Ultrashort Echo-time MRI

Motivation: MRI-based non-enhanced 3D perfusion-weighted (QW) maps have challenges due to unwanted signals from small vessels and registration errors. Goal(s): In this study, we propose developing a method that reliably excludes these unwanted signals to improve the accuracy of quantitative pulmonary perfusion analysis. Approach: We rescaled the difference map into 30 equal-level bins and applied a bin-based clustering technique to remove non-clustered, artifact-like signals from the QW map after image registration. Results: The proposed method effectively removes unwanted signals caused by small vessels or boundary signals during image processing, including registration and demonstrated performance in patients with COPD and ILD. Impact: Improves the reliability of the 3D non-contrast enhanced perfusion-weighted map by effectively removing unwanted signals due to registration errors combined with low SNR and undersampling during image processing.