Rapid cine cardiovascular magnetic resonance with compressed sensing – where is the limit? (RF_SA_445)

Cine imaging is a staple cardiovascular magnetic resonance (CMR) sequence, however cine scans are long and require extensive breath-holds. Compressed sensing (CS) was introduced commercially in 2017, and while this method improves image quality compared to sensitivity encoding, CS has yet to be widely adopted. Typical clinical acceleration factors for cine exams are 2-3. In this study we evaluated a range of CS factors 2-8 and investigated the effect of CS on image quality and quantitative measures of ejection fraction.

Methods:

Thirty healthy volunteers (16 F; 41.5±11.1 years; 81.4±15.7 kg) provided informed consent and underwent an MRI scan including short-axis cine accelerated with CS factors 2, 3, 4, 5, 6, and 8. Breath-holds were kept between 10 and 16 seconds by increasing slices acquired per breath-hold. Cine scans were 2D balanced Turbo Field Echo acquisitions with echo train length = 8, TR/TE = 2.8/1.4 ms, and voxel size 2.1x2.2x8 mm. Breath-holds for left ventricle coverage were 15 for CS2, 8 for CS3 and CS4, 5 for CS5 and CS6, and 4 for CS8. System was a 3.0T Philips Elition X with 28-channel torso receive array.

Image quality was defined as clinicians’ comfort in determining pathology from the image and image usefulness for defining contours – scores were 1=excellent, 2=good, 3=fair, 4=non-diagnostic. One reviewer, blinded to CS acceleration, graded all short axis stacks (n=180). Population-wise scores were evaluated using a Wilcoxon rank-sum test.

In a subset of eight randomly selected subjects, volumetric parameters and ejection fraction were determined for each CS factor, a total of 40 short axis cine acquisitions, using commercially available software.

Results:

Images for a central slice on cine with CS factors 2-8 are shown in Figure 1, where image quality stays at least good up to a CS factor of 5. The scores for 30 volunteers are shown in Figure 2, where all exams up to CS factor 6 were diagnostic. P-values comparing the image quality metric to CS2 were CS2 vs CS3 p=0.657, CS2 vs CS4 p=0.626, CS2 vs CS5 p=0.626, CS2 vs CS6 p=0.207, and CS2 vs CS8 p=0.002. These findings indicate that CS5 has very similar perceived image quality compared to CS2 in this group of 30 volunteers, and no CS2-CS5 images had non-diagnostic quality.

Quantitative measurements in a subset of volunteers (Figure 3) found that CS factors 3-8 did not yield significantly different ejection fraction measures compared to the CS2 acquired image. Up to CS2 vs CS5, p-values were ≥0.842 for left ventricular ejection fraction, and ≥0.917 for right ventricular ejection fraction.

Conclusion:

Cine scans acquired on modern 3.0 T CMR equipment provide diagnostic-quality images for compressed sense factors 5-6, an additional two-fold acceleration compared to typical clinical acceleration. Higher acceleration factors would result in improved patient compliance and smoother workflows. One limitation of this study is that only healthy volunteers were included. Further work in a clinical population is needed.