Automatic Detection of Breathing Motion and Bolus Timing in DCE-MRI of the Lung to Improve Perfusion Quantification Results

Motivation: Quantification of DCE-MRI in the lungs is vulnerable to respiratory motion. In previous work, timepoints with residual motion were selected manually. Goal(s): To develop automatic methods for breathing detection and bolus characterization in lung DCE-MRI. Approach: We developed a breathing motion detection CNN and algorithms for bolus classification in lung DCE-MRI. Quantitative perfusion parameters were calculated with and without removing frames with breathing motion. Results: Breathing motion was robustly detected independent of contrast agent presence. Also, artefacts in PBF maps were reduced. Impact: The developed CNN for breathing detection in DCE-MRI allows the robust, automatic identification of breathing motion. After identification, breathing correction methods can be applied to the affected frames, which might improve the comparability of quantitative perfusion metrics.