Qualitative and quantitative hard-tissue MRI with portable Halbach scanners

TL;DR

This study tackles the challenge of imaging hard tissues with portable, low-field MRI by implementing a ZTE-like PETRA framework on a Halbach-based system. It couples pre-emphasis RF pulse control, an extended SPDS protocol for simultaneous $B_0$ and $B_1$ mapping, and model-based reconstruction to mitigate strong field inhomogeneities, enabling in vivo imaging of ligaments, tendons, cartilage, and cortical bone in under 15 minutes. The authors introduce VFA-PETRA to achieve quantitative $T_1$ mapping across soft and hard tissues at $B_0\approx 0.09$ T, validated in phantoms and knees, with in vivo knee/ankle demonstrations. They show that with field-PK aided reconstruction, hard-tissue visualization and $T_1$ quantification become feasible on portable Halbach systems, marking a step toward affordable, non-ionizing musculoskeletal imaging in point-of-care and remote settings. The work highlights practical limitations (SNR, resolution, partial-volume effects) and lays out a clear pathway for hardware and algorithm improvements to broaden clinical impact.

Abstract

Purpose: To demonstrate the feasibility of performing in vivo imaging and quantitative relaxation mapping of soft and hard tissues using a low-cost, portable MRI scanner, and to establish the methodological foundations for zero echo time (ZTE) imaging in systems affected by strong field inhomogeneities. Methods: A complete framework for artifact-free ZTE imaging at low field was developed, including: (i) RF pulse pre/counteremphasis calibration to minimize ring-down and electronics switching time; (ii) an extension of a recent single-point double-shot (SPDS) protocol for simultaneous B0 and B1 mapping; and (iii) a model-based reconstruction incorporating these field maps into the encoding matrix. ZTE imaging and variable flip angle (VFA) T1 mapping were performed on phantoms and in vivo human knees and ankles, and benchmarked against standard RARE and STIR acquisitions. Results: The optimized PETRA sequence produced 3D images of knees and ankles within clinically compatible times (< 15 min), revealing hard tissues such as ligaments, tendons, cartilage, and bone that are invisible in spin-echo sequences. The extended SPDS method enabled accurate field mapping, while the VFA approach provided the first in vivo T1 measurements of hard tissues at B0 < 0.1 T. Conclusions: The proposed framework broadens the range of pulse sequences feasible in portable low-field MRI and demonstrates the potential of ZTE for quantitative and structural imaging of musculoskeletal tissues in affordable Halbach-based systems.

Figures (12)

• Figure 1: The scanner used for the experiments here presented. a) Elliptical Halbach magnet after ring joining and gradients module inserted. b) RF solenoidal coil loaded with a phantom for $B_0$ and $B_1$ mapping of full volume. c) General outlook of the MRI system in the laboratory once finished the integration of all componentes. d) Patient positioning during a knee study in this system.
• Figure 2: Temporal trace of the $B_1(t)$ field generated by the RF coil after triggering the RFPA with a pulse (a) without and (b) with pre- and counter-emphasis. (c) FFTs of the $B_1(t)$ field envelopes detected by the pickup coil in both cases.
• Figure 3: Calibration of the NMR signal with the RF coil loaded by a knee for PETRA optimization in Halbach systems. (a) Rabi flop acquired from the repeated application ($\text{TR}=1$ s) of square RF pulses with pre-emphasis ($t_\text{RF}=25$ µs) and linearly increasing amplitude for flip-angle calibration. (b) FID before and after active shimming optimization. (c) Frequency spectra of the FIDs in (b), showing line narrowing to 88 ppm ($\text{FWHM}=320$ Hz at 3.64 MHz) after shimming. (d) Larmor frequency drift during PETRA and STIR studies on in vivo subjects and the corresponding variation observed in phantom measurements.
• Figure 4: Evaluation of reconstruction quality of a reticulated phantom encoded with PETRA sequences: (a) effect of the ratio between Tx/Rx switching time and acquisition window ($T_\text{dead}/T_\text{acq}$), (b) effect of radial $k$-space undersampling ($\text{US}_\text{rad}$), and (c) dependence on RF pulse length ($t_\text{RF}$).
• Figure 5: Evaluation of $B_0(\vec{r})$ and $B_1(\vec{r})$ mapping protocols and comparison between agnostic and model-based reconstruction methods. (a) Different slices along the $z$-direction of the $\Delta B_0(\vec{r})$ map. (b) Corresponding slices showing the $B_1$ efficiency map $\eta(\vec{r})$. (c) Same slices reconstructed with ART without PK of the fields (Eq. (\ref{['eq:EM']}) with $\Delta B_0(\vec{r}_j)=0$ and $\epsilon(\vec{r}_j)=1$) from a grid phantom encoded with PETRA. (d) Same slices reconstructed with ART including PK of both $B_0(\vec{r})$ and $B_1(\vec{r})$, using the same PETRA sequence as in (c).