The morphology of the diaphragm is an important factor in the consideration of dyspnoea and treatment of respiratory diseases. The acquisition of images with commonly used methods is limited by the patient position or duration of the procedure. We present the first images of the diaphragm acquired in an upright MR scanner, and estimate repeatability and differences in morphology depending on posture.
Lung diseases such as chronic obstructive pulmonary disease (COPD), affecting > 1M people in the UK, are characterised by airflow limitation and hyperinflation of the lungs. Compromise to either diaphragm excursion or to its overall morphology further contributes to the symptoms by changing respiratory mechanics. Consequently, the characterisation of diaphragm shape is important for precision medicine and the consideration of interventions. Traditionally, the 3D geometry of the diaphragm is assessed using supine MR or CT. This puts an additional burden on the patient and is limited in its ability to investigate function in other positions and postures. A method capable of 3D reconstruction in various positions was proposed1, using ultrasound (US) and motion tracking, but required a long 30s breath-hold and was of limited utility since parts of the diaphragm may not be accessible to US.
Aim: to acquire preliminary MR images of the diaphragm in upright seated position and estimate the impact of subject position on its morphology.
Two healthy subjects (20 and 23 yo) were repeatedly scanned on a 0.5T Paramed (Genoa, Italy) MR Open scanner with a torso coil at full inspiration. We used a 2D spin echo sequence to acquire 4 sets of 3 scans each, comprising sagittal (SAG, TE/TR=12/416 ms, FA=90°, FOV=36 cm, matrix=256 x 256, resolution=1.4 x 1.4 mm, 18 slices, slice thickness=16 mm, interslice gap=10 mm) and coronal scans (COR, TE/TR=12/350 ms, FA=90°, FOV=36 cm, matrix=256 x 256, resolution=1.4 x 1.4 mm, 18 slices, slice thickness=16 mm, interslice gap=10 mm) in supine (SUP) and seated (SEAT) position. Each scan took approximately 20 seconds.
A 3D surface of the diaphragm was semi-automatically reconstructed from each individual scan. The processing pipeline comprised rigid registration of supine and seated images to a common space, the manual selection of points on the diaphragm on each slice, and the 3D reconstruction of a gridded surface by interpolating between these points.
Repeatability was assessed by estimating the variability in displacement for each of the 4 sets, around the average ($$$\bar{S}_{COR,SUP}$$$, $$$\bar{S}_{COR,SEAT}$$$, $$$\bar{S}_{SAG,SUP}$$$ and $$$\bar{S}_{SAG,SEAT}$$$) at every grid point. The impact of the patient position on the diaphragm geometry was demonstrated by comparing the mean surfaces.