Fast quantitative MRI: Spiral Acquisition Matching-Based Algorithm (SAMBA) for Robust T1 and T2 Mapping

Conventional diagnostic images from Magnetic Resonance Imaging (MRI) are typically qualitative and require subjective interpretation. Alternatively, quantitative MRI (qMRI) methods have become more prevalent in recent years with multiple clinical and preclinical imaging applications. Quantitative MRI studies on preclinical MRI scanners are being used to objectively assess tissues and pathologies in animal models and to evaluate new molecular MRI contrast agents. Low-field preclinical MRI scanners (≤3.0T) are particularly important in terms of evaluating these new MRI contrast agents at human MRI field strengths. Unfortunately, these low-field preclinical qMRI methods are challenged by long acquisition times, intrinsically low MRI signal levels, and susceptibility to motion artifacts. In this study, we present a new rapid qMRI method for a preclinical 3.0T MRI scanner that combines a Spiral Acquisition with a Matching-Based Algorithm (SAMBA) to rapidly and quantitatively evaluate MRI contrast agents. In this initial development, we compared SAMBA with gold-standard Spin Echo MRI methods using Least Squares Fitting (SELSF) in vitro phantoms and demonstrated shorter scan times without compromising measurement accuracy or repeatability. These initial results will pave the way for future in vivo qMRI studies using state-of-the-art chemical probes.


Introduction
Magnetic Resonance Imaging (MRI) is a noninvasive imaging modality of interest for a wide range of clinical applications due to its high spatial resolution and soft-tissue contrast [1].While traditional MRI relies on qualitative interpretation by radiologists [2], quantitative MRI (qMRI) provides an objective assessment of tissues and pathologies, crucial for understanding disease mechanisms and evaluating therapies.qMRI is particularly valuable in preclinical studies using animal models to develop new MRI probes and sensors targeting specific disease molecular processes.The magnetic field strength in preclinical MRI scanners typically range from 1.0T to 15.2T, presenting a tradeoff the higher the magnetic strength between improved resolution and reduced translational capabilities when exploring newer molecular MRI contrast agents.The application of qMRI, especially in low-field preclinical scanners (≤3.0T), is hindered by (1) long scan times, particularly with conventional spin-echo MRI protocols to estimate T 1 and T 2 relaxation times [3]; and (2) susceptibility to respiratory motion artifacts, which require gating that can further extend the acquisition time.Conventional faster MRI approaches, such as echo-planar imaging (EPI) and balanced Steady-State Free Precession (bSSFP) come at the cost of increased eddy current and banding artifacts, respectively [4,5].Therefore, there is a critical need for new, rapid, motion-resistant qMRI methods for low-field preclinical MRI scanners, particularly for evaluating new molecular MRI probes and sensors accurately and effectively with sufficient spatial and temporal resolution.
Herein, we propose a rapid preclinical qMRI method "SA-MBA" (<5 min/scan) to objectively evaluate contrast agents at 3.0T based on T 1 and T 2 maps.Inspired by the increasing developments in MR Fingeprinting [6] techniques, SAMBA combines a spiral acquisition "SA" (accelerates data sampling, decreasing scan durations) [5] and a dictionary matching-based algorithm "MBA" [1,6] (with inherent resistance to motion artifacts [7] in comparison to conventional least squared error fitting methods [3]).While the majority of components have been described previously, this is the first implementation of these spiral and MBA methods that we are aware of on a preclinical 3.0T MRI scanner [8].In this study, the performance of SAMBA against gold-standard spin echo MRI methods has been evaluated with in vitro phantoms to pave the path for future development of fast, quantitative acquisitions for low-field MRI scanners.
All images were exported for offline analysis and calculation of T 1 and T 2 maps in MATLAB (MathWorks, Natick, MA).Established exponential decay models were applied to the magnitude images [12,13], and regions-of-interest (ROI) were defined on each image to obtain the average values.Spin Echo T 1 and T 2 maps were obtained with conventional linear least squared error fitting.SAMBA T 1 and T 2 maps were derived from a dedicated dictionary as previously described [14,15] (Supplemental Figure 1).
Statistical analysis was performed via GraphPad Prism (8.4.3).Statistical significance was defined at p< 0.05.Linear regression analysis, Bland-Altman plots, and two-tailed Student's t-tests were performed to compare the T 1 and T 2 estimates between SAMBA and SELSF methods.The repeatability for each method was determined by calculating the coefficient of variation (CV% = standard deviation / mean * 100 %) for the 8 consecutive scans [16].The CV% * scan time (in hours) to account for the different acquisition times.

Discussion
This study introduces a fast, quantitative MRI sequence [13] for T 1 and T 2 Spiral Acquisition with a Matching-Based Algorithm (SAMBA) implemented in a preclinical 3.0T scanner [6].Developing successful fast qMRI approaches for in vivo [10] preclinical studies at low-field is crucial to better understand molecular processes [16] evaluate pathophysiology, and develop novel chemical probes or sensors.
We, therefore, present SAMBA as a fast, accurate, and precise T 1 and T 2 mapping approach at low-field preclinical scanners, paving the path for easier translation of findings to clinical settings of similar field strengths [16].
SAMBA produced T 1 and T 2 maps significantly faster but strongly correlated with goldstandard Spin Echo Least Squares Fitting "SELSF" approaches (linear associations for T 1 : R 2 = 0.999, slope = 1.007, y-intercept = 11.65 ms; T 2 : R 2 = 0.998, slope = 1.062, y-intercept = −0.466ms; Fig. 1 and Supplemental Figures 2 & 3).T 1 SAMBA and SELSF results had a stronger correlation than those for T 2 , which has been previously attributed to a heavier influence of B 1 inhomogeneities for T 2 sequences implementing similar matching-based approaches [16].Although B 1 map correction [18], slice profile, and B1 + compensation [19] approaches are beyond the scope of this study, their implementation could serve to increase the accuracy of T 2 estimates in future work [20][21][22].Both SAMBA and SELSF had comparable repeatability (T 1 : 0.30 ± 0.10 % SAMBA vs. 0.26 ± 0.14 % SELSF; T 2 : 0.95 ± 0.32 % SAMBA vs. 1.63 ± 0.85 % SELSF).SAMBA has the most positive impact when analyzing the decrease in variation CV% per scan time (Fig. 2), in comparison to SELSF.SAMBA performance demonstrated against SELSF defines a pathway towards the analysis of MRI sensors in vitro with high accuracy and repeatability even when using low volumes and concentrations [23], therefore advancing the field of MRI for chemical imaging [24,25].Future studies should investigate the performance of SAMBA as a preclinical in vivo imaging platform [2] to reduce sensitivity to motion artifacts [26].In vivo SAMBA imaging could serve to quantify biomarkers, characterize treatment responses, and monitor chemical signals (e.g., neurotransmitter Acetylcholine) [27].
At the sequence level, future projects could optimize SAMBA by implementing parallel imaging 27 or undersampling methods to improve temporal performance for dynamic contrast-enhanced MRI studies, although they lead to signal-deprived images.Increasing coverage via 3D acquisitions could compensate for this by increasing the signal, although at the cost of extended scan times that may limit future in vivo dynamic MRI studies using molecular MRI contrast agents.In addition, the sequence could also characterize other parameters (diffusion, non-contrast perfusion, magnetization transfer [28], and chemical exchange saturation transfer [29]) by adjusting the preparation schema [8].Lastly, SAMBA measurements could be analyzed via semi-automatic approaches based on artificial intelligence frameworks [30], providing further potential improvements in time efficiency, cost, and accuracy of existing practice.
In summary, we have reported and evaluated a novel quantitative MRI sequence specifically designed for a preclinical 3.0T MRI scanner based on a spiral trajectory scheme and a matching-based algorithm (SAMBA), resulting in much shorter scan times without compromising the accuracy or repeatability in T 1 and T 2 estimates in comparison to conventional spin echo MRI [16].This SAMBA platform could pave the way for more advanced in vivo non-invasive molecular quantitative methods, facilitating the development and evaluation of state-of-the-art chemical probes and serving as a foundation to further advance the study of biochemical pathways and quantitative biomarkers non-invasively [13], and ultimately unlock the full potential of MRI for chemical imaging of living systems.Future lines of research could explore multi-site, multi-strength, and multi-vendor quantifications in efforts to standardize the measurements for quantitative in vivo MRI [2].Linear regression and representative colormaps of T 1 (left) & T 2 measurements (right) obtained via conventional approach (Spin Echo Least Squares Fitting "SELSF" x axes) vs. proposed method (Spiral Acquisition Matching-Based Algorithm "SAMBA", y axes).Dashed line represents the unity, error bars represent standard deviation between 8 measurements.Analysis of measurement repeatability from T 1 & T 2 phantoms (left column): comparison of Coefficient of Variation (CV%) and CV% per time (middle columns), and ratio between conventional and proposed approaches (right column).