Customizing Native T1 Mapping: the effects of compressed sensing, deep learning-based denoising and high-resolution on measurement of native myocardial T1.
Authors
Affiliations (6)
Affiliations (6)
- Department of Radiology, Mayo Clinic College of Medicine, Rochester, MN, USA; Department of Radiological and Hematological Sciences, Università Cattolica del Sacro Cuore, Rome, Italy.
- Department of Radiology, Mayo Clinic College of Medicine, Rochester, MN, USA; Department of Radiology, Humanitas Research Hospital, Milan, Italy.
- Department of Radiology, Mayo Clinic College of Medicine, Rochester, MN, USA.
- Department of Radiology, Mayo Clinic College of Medicine, Rochester, MN, USA; School of Medicine, Vita-Salute San Raffaele University, Milan, Italy.
- Department of Radiology, Mayo Clinic College of Medicine, Rochester, MN, USA; North American Clinical Science, Philips, Cambridge, MA, USA.
- Department of Radiology, Mayo Clinic College of Medicine, Rochester, MN, USA. Electronic address: [email protected].
Abstract
Quantitative native T1 mapping is a key component in Cardiovascular magnetic resonance (CMR) for myocardial tissue characterization; however, further improvements in acquisition efficiency and robustness are needed to optimize clinical applicability. Undersampling techniques in k-space, such as compressed sensing (CS), and deep learning-based (DL) denoising reconstructions, have improved morphological and cine-imaging, but their impact on quantitative relaxation times remains underexplored. This study evaluated image quality and native T1 quantification across ten combinations with varying CS acceleration levels, spatial resolutions, and application of DL-denoising reconstruction. In this prospective single-center study, 48 healthy volunteers underwent native T1 mapping. After quality review, 41 subjects were included. Blurring, aliasing, susceptibility artifacts, and overall image quality (IQ), were rated by three blinded readers using a 4-point Likert scale. Quantitative analysis was performed on a per-segment basis using custom software, yielding mean native T1 values for each AHA segment. Nine cases were reanalyzed by three additional readers to assess interobserver variability. Test-retest and phantom experiments were performed to assess reproducibility and to cover pathological T1 ranges, respectively. The reference T1-mapping protocol was obtained with CS3, with spatial resolution of 2.0×2.0×10mm<sup>3</sup>. Additional acquisitions were obtained with CS3, 4, 5 and spatial resolutions of 2.0×2.0×10mm<sup>3</sup>, 1.8×1.8×10mm<sup>3</sup>, and 1.6×1.6×10mm<sup>3</sup> for CS3 and 2.0×2.0×10mm<sup>3</sup> for CS4 and 5. Significant differences in IQ, blurring, and aliasing were observed among acquisition protocols (p<0.05), but not in susceptibility artifacts (p=0.66). Higher CS levels slightly reduced IQ and increased aliasing and blurring. DL-denoising and higher spatial resolution methods improved sharpness without changing overall scores. Segment-wise T1 quantitative analysis revealed only minor differences, even in the presence of high acceleration factors or increased spatial resolution. In most segments, two one-sided testing confirmed equivalence with low bias (-21 to +22 ms). Scan-rescan experiments confirmed repeatability in healthy volunteers, while phantom experiments extended repeatability and inter-method stability to pathological T1 ranges. Native T1 values remained stable across various CS acceleration factors, changes in spatial resolution, and application of denoising, with clinically negligible bias for tissue characterization. These findings, within a framework where total acquisition time is primarily determined by the number of inversion time (TI) images, support the use of CS to improve temporal resolution within each TI image, benefiting patients with arrhythmias or elevated heart rates, and also support the use of denoising and higher spatial resolution to improve image details. Overall, these results can enable patient-tailored T1 mapping protocols, improving the efficiency and diagnostic utility of quantitative CMR.