Vendor‐agnostic 3D multiparametric relaxometry improves cross‐platform reproducibility

• Published in: Magnetic resonance in medicine Vol. 94; no. 3; pp. 937 - 948
• Main Authors: Fujita, Shohei, Gagoski, Borjan, Nielsen, Jon‐Fredrik, Zaitsev, Maxim, Jun, Yohan, Cho, Jaejin, Yong, Xingwang, Uhl, Quentin, Xu, Pengcheng, Milshteyn, Eugene, Shaik, Imam Ahmed, Liu, Qiang, Chen, Qingping, Afacan, Onur, Kirsch, John E., Rathi, Yogesh, Bilgic, Berkin
• Format: Journal Article
• Language: English
• Published: United States Wiley Subscription Services, Inc 01.09.2025
• Subjects: Adult; Algorithms; Brain - diagnostic imaging; Coefficient of variation; Correlation coefficients; cross‐vendor technique; data pooling; Evaluation; Humans; Image Processing, Computer-Assisted - methods; Imaging, Three-Dimensional - methods; Linearity; Magnetic Resonance Imaging - methods; Male; Multiparametric Magnetic Resonance Imaging - methods; multiparametric mapping; NMR; Nuclear magnetic resonance; Phantoms, Imaging; quantitative magnetic resonance imaging; relaxation time; relaxometry; Reproducibility; Reproducibility of Results; Scanners; Sequences; Software; Substantia grisea; data pooling; quantitative magnetic resonance imaging; relaxometry; relaxation time; multiparametric mapping; cross‐vendor technique
• ISSN: 0740-3194; 1522-2594; 1522-2594
• DOI: 10.1002/mrm.30566

Purpose To address the unmet need for a cross‐platform, multiparametric relaxometry technique to facilitate data harmonization across different sites. Methods A simultaneous T1 and T2 mapping technique, 3D quantification using an interleaved Look–Locker acquisition sequence with a T2 preparation pulse (3D‐QALAS), was implemented using the open‐source vendor‐agnostic Pulseq platform. The technique was tested on four 3 T scanners from two vendors across two sites, evaluating cross‐scanner, cross‐software version, cross‐site, and cross‐vendor variability. The cross‐vendor reproducibility was assessed using both the vendor‐native and Pulseq‐based implementations. A National Institute of Standards and Technology/International Society for Magnetic Resonance in Medicine system phantom and three human subjects were evaluated. The acquired T1 and T2 maps from the different 3D‐QALAS runs were compared using linear regression, Bland–Altman plots, coefficient of variation (CV), and intraclass correlation coefficient (ICC). Results Pulseq‐QALAS demonstrated high linearity (R2 = 0.994 for T1, R2 = 0.999 for T2) and correlation (ICC = 0.99 [0.98–0.99]) against temperature‐corrected NMR reference values in the system phantom. Compared to vendor‐native sequences, the Pulseq implementation showed significantly higher reproducibility in phantom T2 values (CV, 2.3% vs. 17%; p < 0.001), and improved T1 reproducibility (CV, 3.4% vs. 4.9%; p = 0.71, not significant). The Pulseq implementation reduced cross‐vendor variability to a level comparable to cross‐scanner (within‐vendor) variability. In vivo, Pulseq‐QALAS exhibited reduced cross‐vendor variability, particularly for T2 values in gray matter with a twofold reduction in variability (CV, 2.3 vs. 5.9%; p < 0.001). Conclusion An identical implementation across different scanners and vendors, combined with consistent reconstruction and fitting pipelines, can improve relaxometry measurement reproducibility across platforms.