Abstract
We propose multiMap, a single scan that can generate several quantitative maps simultaneously. The sequence acquires multiple images in a time-efficient manner, which can be modeled for $T_2$, $T_2^*$, main- and transmit-field inhomogeneity, $T_1$:equilibrium magnetization, and water and fat content. The sequence is constructed so that cross-dependencies between parameters are isolated. Thus, each parameter can be estimated independently. Estimates of all parameters are shown on bottle phantoms, the brain, and the knee. The results are compared to estimates from established techniques.
Introduction
In clinical MRI, qualitative "weighted" images (e.g., $T_1$-weighted or $T_2$-weighted) remain the standard. However, quantitative MRI (qMRI) is increasingly recognized as the "holy grail" for objective disease tracking, as it provides absolute physical values. Despite its benefits, qMRI is often too slow for routine clinical use. Traditionally, measuring parameters like $B_1$ (transmit field), $B_0$ (static field), $T_1$, $T_2$, $T_2^*$, and fat fraction requires six separate scans, leading to long exam times and susceptibility to patient motion. While MR Fingerprinting (MRF) offers a simultaneous approach, it relies on computationally intensive dictionary matching. We introduce multiMap, a gradient-spoiled sequence that efficiently captures these six maps in a single scan by isolating parameter dependencies.
Methods
The Sequence Design
The multiMap sequence is a gradient-spoiled design building upon the Saturated Double Angle Method (SDAM). It utilizes a series of seven RF pulses and eleven distinct acquisition windows ($I_1$ through $I_{11}$) per repetition time (TR). The scan is divided into two segments using different imaging flip angles (60° and 120°) to facilitate $B_1$ estimation.
A key innovation is "signal recycling." While standard SDAM ignores the transverse magnetization from the initial saturation pulse, multiMap uses this signal to estimate $T_2^*$, $B_0$, and fat fraction. The subsequent longitudinal recovery period probes $T_1$, and the final spin-echoes provide the data for $T_2$ mapping.
The 4-Step Quantitation Algorithm
To avoid complex, non-convex optimization, the authors decoupled the parameter estimation into four sequential steps:
- Estimate $B_1$: Calculated using the ratio of images from the two segments at acquisition $I_8$.
- Estimate $T_2$: Derived from the late spin-echo acquisitions ($I_9–I_{11}$).
- Estimate $B_0$, Fat Fraction, and $T_2^*$: Extracted from the early gradient-echo acquisitions ($I_1–I_5$).
- Estimate $T_1/M_0$: Determined using the recovery-curve acquisitions ($I_6–I_8$).
This decoupling ensures that errors in one parameter (like $T_1$) do not propagate into others (like $B_1$), a common issue in multi-parametric mapping. Furthermore, the model incorporates slice-profile effects using a piecewise constant approximation of the RF pulse and a lookup table for correction.
Datasets and Evaluation
The performance of multiMap was validated using:
- Bottle Phantoms: Six bottles containing varying concentrations of manganese chloride, copper sulfate, and emulsified peanut oil to simulate different tissue properties and fat content.
- In-vivo Human Imaging: Sagittal knee and axial brain slices were acquired to demonstrate clinical feasibility in musculoskeletal and neural tissues.
Key Results
The quantitative maps generated by multiMap showed strong agreement with gold-standard individual scans for $B_1$, $B_0$, and $T_1/M_0$. In phantom experiments, the fat fraction estimates were highly accurate, with signal intensity ratios within 5% of the actual volume ratios.
The total scan time for a 128x128 phantom map was approximately 10.5 minutes. While longer than a standard weighted scan, it represents a significant speedup over six individual quantitative sequences. In brain imaging, the $M_0$ (proton density) map clearly distinguished between parenchyma and ventricles, and the fat fraction map accurately localized subcutaneous fat.
Discussion and Future Work
While multiMap successfully demonstrates simultaneous multi-parametric mapping, several areas for improvement remain. The current scan time of ~10 minutes per slice is a hurdle for full clinical adoption; however, the integration of parallel imaging or compressed sensing could substantially reduce this. Additionally, the accuracy for $T_2$ and $T_2^*$ is currently constrained by the fixed echo spacing. Future iterations could extend this method to 3D volumes and incorporate motion compensation techniques for abdominal imaging.
By cleverly decoupling parameter dependencies, multiMap provides a robust and efficient alternative to dictionary-based methods for comprehensive quantitative MRI.
Summary updated