A novel approach for quantitative T1 measurements: validation in a longitudinal pediatric sample

1890 

Abstract Submission 

Zihan Zhou^1,2, Maya Yablonski^2,3, Xiaozhi Cao^1,4, Mengze Gao¹, Congyu Liao^1,4, Kawin Setsompop^1,4, Jason Yeatman^2,3,5

¹Department of Radiology, Stanford University, Stanford, United States, ²Graduate School of Education, Stanford University, Stanford, United States, ³Department of Pediatrics, Stanford University School of Medicine, Stanford, United States, ⁴Department of Electrical Engineering, Stanford University, Stanford, United States, ⁵Department of Psychology, Stanford University, Stanford, United States

Zihan Zhou  
Department of Radiology, Stanford University|Graduate School of Education, Stanford University
Stanford, United States|Stanford, United States

Maya Yablonski  
Graduate School of Education, Stanford University|Department of Pediatrics, Stanford University School of Medicine
Stanford, United States|Stanford, United States

Xiaozhi Cao  
Department of Radiology, Stanford University|Department of Electrical Engineering, Stanford University
Stanford, United States|Stanford, United States

Mengze Gao  
Department of Radiology, Stanford University
Stanford, United States

Congyu Liao  
Department of Radiology, Stanford University|Department of Electrical Engineering, Stanford University
Stanford, United States|Stanford, United States

Kawin Setsompop  
Department of Radiology, Stanford University|Department of Electrical Engineering, Stanford University
Stanford, United States|Stanford, United States

Jason Yeatman  
Graduate School of Education, Stanford University|Department of Pediatrics, Stanford University School of Medicine|Department of Psychology, Stanford University
Stanford, United States|Stanford, United States|Stanford, United States

Developmental cognitive neuroscience aims to shed light on evolving relationships between brain structure and cognitive development. To this end, quantitative methods that reliably measure individual differences are fundamental. Qualitative MR (e.g., T1 weighted images) are influenced by multiple biological factors, along with scan parameters and biases from hardware. In contrast, quantitative T1 mapping measures specific tissue properties [1,2]. Conventional quantitative T1 mapping methods suffer from long scan times, low resolution and sensitivity to field inhomogeneity and motion. To address these limitations, we adopted a novel MR fingerprinting (MRF) protocol for rapid 1-mm whole-brain T1 maps in 2 minutes which is robust to B1-inhomogeneity [3,4]. This method has yet to be evaluated for cognitive neuroscience research. In this study, we examine the reliability and validity of MRF-based T1 measurements in children scanned longitudinally.

49 children aged 8-13y (mean 10y ±1.4) completed two scanning sessions 2-4 months apart. In each session, two multi-axis spiral-projection 3D-MRFs were collected with complementary acquisition trajectories [5]. A separate calibration scan ("PhySiCal") [6] was used to measure B0 inhomogeneity. The total scan time was 4.5 minutes. A subspace recon with locally low-rank constraint was used for MRF reconstruction, followed by pattern matching to generate 1-mm resolution T1 maps. We examined SNR and B0 correction's impact on reliability, comparing single 2-min and combined two 2-min scans, both with and without B0 correction (Figure 1). We used two complementary approaches to evaluate test-retest reliability of T1 values. (A) We compared whole-brain voxel-based T1 values in white matter (WM), using two session data coregistered to a midpoint to avoid bias. (B) We ran an ROI-based comparison using WM and gray matter (GM) segmentations from Freesurfer (Freesurfer was run on a MRF-derived synthetic MPRAGE from each session). For WM, we focused on five corpus callosum (CC) segmentations. For GM analysis, we focused on 6 subcortical ROIs: the caudate, putamen, and thalamus, bilaterally.

Four subjects were excluded due to excessive motion. Figure 1 shows the qualitative effect of the different pipelines on image quality. MRF-derived mean T1 values in the CC range from 795-878 ms, showing the typical inverted-U shape pattern (matching known histology) along the anterior-posterior axis. GM T1 values were 1001-1360ms, in line with [7]. Voxel-based reliability analysis showed that combining two 2-min MRF scans yielded higher reliability (r=0.80) than single 2-min scans (r=0.75). Surprisingly, B0-correction did not significantly improve reliability (r=0.80 vs 0.75 for 4-min and 2-min). ROI-based reliability analysis revealed a similar pattern: reliability was highest when combining the two 2-min scans, with minimal B0 correction impact for both WM (Figure 2A) and GM (Figure 2B). CC ROIs showed r=0.83±0.06 (4-min) vs 0.72±0.09 (2-min) for uncorrected data and r=0.82±0.08 (4-min) vs 0.69±0.11 (2-min) for corrected data. Lastly, age was correlated with T1 values in subcortical ROIs and in the anterior CC (p<0.05, fdr corrected; Figure 2C-D), in line with [8].

This study assesses the repeatability and reproducibility of MRF-derived quantitative T1 metrics in children scanned longitudinally. We evaluated four MRF reconstruction strategies on data quality and found that longer acquisitions (which result in higher SNR) improve T1 repeatability. Surprisingly, reliability was high even without a B0-correction and did not improve substantially after correction. This highlights the feasibility of rapidly collecting quantitative MRI measures without requiring additional calibrations. In sum, MRF provides a promising methodology for deriving reliable quantitative metrics of brain tissue structure in children and patient populations where scan time and motion are of particular concern.

Normal Brain Development: Fetus to Adolescence

Methods Development ¹

Anatomical MRI ²

Acquisition

Data analysis

Design and Analysis

Development

MRI

PEDIATRIC

STRUCTURAL MRI

Other - MR fingerprinting, Quantitative MRI