Rapid Fire Abstracts
Karin Pola, MD
MD, PhD student
Lund University, Skåne University Hospital Lund, Department of Clinical Sciences Lund, Clinical Physiology, Lund, Sweden, Sweden
Karin Pola, MD
MD, PhD student
Lund University, Skåne University Hospital Lund, Department of Clinical Sciences Lund, Clinical Physiology, Lund, Sweden, Sweden
Johannes Töger, PhD
PhD, Assoc Professor, Adj senior lecturer
Lund University and Skåne University Hospital, Lund, Sweden, Sweden
Per M Arvidsson, MD, PhD
Postdoctoral researcher
Lund University, Sweden
Cardiovascular magnetic resonance with three-dimensional, time-resolved (4D) flow enables advanced analysis of intracardiac blood flow patterns, including hemodynamic forces (HDF) and kinetic energy (KE) which are attracting increasing clinical attention. The accuracy and precision of 4D flow CMR is subject to image noise, arising primarily from the tissue of the subject in the scanner, and residual phase offsets, which are commonly mitigated using offline postprocessing. Here we evaluate how varying levels of velocity noise and different phase background correction (BGC) strategies impact HDF and KE.
Methods:
From our database we retrospectively included 12 datasets (6 healthy subjects, 6 heart failure patients) acquired on three different platforms (1.5T and 3T Achieva, Philips, the Netherlands, or 1.5T MAGNETOM Aera, Siemens Healthineers, Germany). 4D flow was acquired using sequences with Cartesian readout accelerated with parallel imaging (SENSE factor 2 or GRAPPA factor 2x2) and retrospective ECG gating. Spatial resolution was 3x3x3 mm and acquired temporal resolution typically 50 ms. Concomitant gradient terms were compensated by the scanner software. After contouring the LV cavity over the cardiac cycle, each dataset was analyzed for HDF and KE at uncorrected baseline, with 0th, 1st, or 2nd order background correction applied, or with added random noise with a 2, 4, 8, or 16 cm/s velocity standard deviation in 3D space (Figure 1), using the software Segment based on Matlab. Repeated-measures ANOVA was conducted to test for systematic effects on HDF and KE.
Results:
Figure 2 shows an example of apical-basal hemodynamic force and kinetic energy in a healthy subject. BGC strategy left HDF and KE largely unaffected (±1% compared to baseline for systole, diastole, force ratio, and peak KE, p >0.05 for all). While increasing noise levels had no significant effect on HDF peaks (at most ±3%, p >0.05) or diastolic force ratio (at most +10%, p >0.05), the systolic force ratio was affected at noise levels exceeding 4 cm/s (at most +12%, p< 0.0001). Most importantly, peak KE was dramatically affected by increasing noise, reaching 251% of baseline at 16 cm/s added noise (p< 0.01).
Conclusion:
Analysis of HDF is unaffected by the choice of time-constant background phase correction and relatively insensitive to random velocity noise, as it relies heavily upon the first derivative of the velocity field. In contrast, KE analysis is significantly affected by noise levels that may be seen in clinical scans, as it amplifies errors in velocity.
