SCMR 22nd Annual Scientific Sessions
In PSIR  a single inversion time (TI) provides good contrast over a range of T1 values making it attractive for navigated 3D scans during which T1 increases. However, separate navigation of readout (RO) and PSIR reference (REF, Fig 1) prolongs acquisition and can create ghosting. Therefore, RO and REF are often triggered by the RO navigator although REF occurs in a different beat, degrading spatial registration and image quality (IQ). We recently developed a single-cycle (SC) PSIR  method for better registration. Here we add 3D navigation and compressed sensing (CS) for shorter acquisition. We report initial findings on IQ, infarct detection, and relative contrast of 3D prototype compared to 2D.
On a MAGNETOM Skyra 3T (Siemens Healthineers) the prototype (Fig1) acquired 3D magnitude (MAG) and PSIR images in 11 patients covering the heart in short-axis (2) or axial (9) orientation. 2D short-axis images were taken. Hyperenhancement transmurality was assessed by 17 segment model (score: 0 0%, 1 1-25%, 2 26-59%, 3 51-75%, 4 76-100%). Number of infarct segments per patient was compared between 2D and 3D PSIR (paired t-test). IQ, fat suppression (FS, 0-non-diagnostic, 1-poor, 2-good, 3-excellent), and respiratory artifacts (RA, 0-none, 1-weak, 2-strong) were graded in one 2D and matched 3D slice per patient. MAG SNR was measured in infarct I, myocardium M, and blood B. In PSIR relative contrasts I-M, I-B, and B-M were calculated as (I-M)/M, (I-B)/B and (B-M)/M. SNR was compared between 2D and 3D in MAG, relative contrast in PSIR.
2D and 3D PSIR detected the same number of infarct segments (36) with transmurality score >0. IQ (mean±stdev, 2D 2.27±0.65, 3D 2.45±0.52) and FS (2D 2.45±0.69, 3D 2.73±0.47) were slightly better for 3D PSIR, but statistically identical to 2D (p>0.1). RA (2D 0.55±0.52, 3D 0.72±0.47) were not significantly (p>0.1) worse in 3D. Figure 2 shows typical 2D and acquired/reformatted 3D images. For MAG, SNR was identical between 2D and 3D in I (26.1±4.1 vs 31.4±22.7, p>0.05) and M (7.7±2.5 vs 6.0±2.4, p>0.05), but different in B (30.5±11.4 vs 22.8±9.8, p<0.1). Relative I-M contrast was not different (0.09±0.04 vs 0.22±0.18, p>0.1), nor was I-B (0.00±0.04 vs 0.06±0.17, p>0.1). B-M was larger for 3D (0.09±0.03 vs 0.14±0.05, p<0.05). Acquisitions took 12.4±5.2 minutes for 84.4±9 slices (average efficiency 75±19%, RR = 839±149ms).
The prototype obtains large high-resolution 3D PSIR data sets in reasonable scan time. It offers IQ comparable to 2D at much thinner slices. We expect 3D PSIR to provide better conspicuity of small infarcts, but require more data. The specific 3D protocol provides I and M SNR as in 2D, despite 16-fold CS acceleration and thinner slices. Reduction of B SNR in 3D is likely due to blood saturation by the thick 3D slab.