Selective in Vivo Bone Imaging with Long-T2 suppressed PETRA MRI
Cheng Li1, Jeremy F. Magland1, Xia Zhao1, Alan C. Seifert2, and Felix W. Wehrli1

1Radiology, University of Pennsylvania, Philadelphia, PA, United States, 2Translational and Molecular Imaging Institute, Icahn School of Medicine at Mount Sinai, New York, NY, United States


An IR-based long-T2 suppressed PETRA sequence was designed and optimized to image sub-millisecond-T2 tissue components, e.g. collagen-bound bone water. To minimize scan time signal was sampled repeatedly after each inversion with individual excitation flip-angle designed to yield constant short-T2 signal amplitude. A fast non-iterative reconstruction algorithm combined with phase-modulated excitation pulse was applied to minimize image artifacts due to non-uniform excitation profile, allowing for increased flip-angle and higher SNR. Optimized long-T2 suppressed PETRA allows imaging of bone matrix water, opening up new possibilities for anatomic bone imaging at isotropic resolution and quantification in clinically practical scan times.


With recent advances in MRI technology and hardware, several techniques have emerged for imaging short-T2 constituents, including ultra-short echo time (UTE) imaging (1), zero echo-time (ZTE) imaging and its variants, e.g. PETRA, WASPI (2-4), and SWIFT (5). Although short-T2 protons become detectable, their signals are typically obscured by the much more intense resonances from surrounding long-T2 components. To enhance visualization of the short-T2 protons, long-T2 suppression techniques are commonly employed. Our previous work investigated the performance of various long-T2 suppression schemes in 2D UTE imaging and showed that adiabatic inversion recovery (IR) provides highly uniform short-T2 contrast achieving optimal long-T2 suppression with near immunity to B1 inhomogeneity (6). However, application of IR-based long-T2 suppression schemes to ZTE imaging performed on clinical scanners poses particular challenges. Since ZTE is a 3D encoding technique, a large number of projections are needed. IR-preparation significantly prolongs TR and B1 peak power constraint imposes limits on the achievable flip angle. Here, the IR-based long-T2 suppression approach has been extended to PETRA imaging (a ZTE embodiment on a 3T clinical scanner) and the method was evaluated in vivo at the tibia shaft and foot.


Pulse sequence: The pulse sequence accommodates a hyperbolic secant (HS) adiabatic pulse set to the resonance frequency of water for long-T2 magnetization inversion (Fig.1). To shorten scan time, seven ZTE spokes were acquired after each inversion.

Determination of optimal flip angle set: In order to achieve optimal response, i.e. maximal transverse magnetization, and equal amplitude of the short-T2 signal in each of the seven successive views a binary search algorithm was devised for fast determination of the flip angles. The algorithm considers in-pulse relaxation effects and magnetization recovery between each excitation under scanner hardware constraints, e.g. maximum allowable B1 peak power. Assuming bound bone water T1 and T2 of 145 ms and 400 ms, the optimal flip angles of the seven successive excitations are 22.5, 23.8, 25.4, 27.5, 30.2, 34.0, 40.0 degrees.

Enhanced flip angle using quadratic phase-modulated RF excitation: A constraint of the parent ZTE-based approach is that the duration of the excitation pulse be short enough to uniformly excite the spins across the FOV, thereby limiting the achievable flip angle imposed by available B1 peak power and SAR limits. Therefore, in order to attain optimal flip angles without incurring image artifacts resulting from non-uniform excitation across the FOV, we applied a phase-modulated excitation RF pulse. Images were reconstructed with a fast non-iterative reconstruction algorithm to correct the excitation profile effect (7).

In vivo imaging: The mid-tibia of a 30-year-old healthy male subject was imaged using an eight-channel transmit/receive knee coil at 3T with the following imaging parameters: FOV = 2503 mm3, TR = 200 ms, T/R switch time=50 μs, TI=75 ms, 12,000 half-projections, scan time = 6.5 min, image matrix = 2563. An HS adiabatic inversion pulse of 5 kHz bandwidth and 5 ms pulse duration was used. Seven ZTE spokes were sampled at 2 ms interval, pulse duration = 40 μs, dwell time = 10 μs. The foot of a 28-year-old healthy male volunteer was also scanned to evaluate the method’s performance of delineating the complex anatomy of the foot involving both cortical and trabecular bone. For comparison of the anatomy, images were also acquired with a gradient-echo sequence at the same FOV and spatial resolution.


Fig.2 displays representative mid-tibia images in the three orthogonal planes along with the matching slice images from a conventional gradient-echo sequence. Tibia and fibula are well visualized and the surrounding long-T2 tissue components are effectively suppressed. Figs.3a-c show axial, coronal and sagittal images of the foot. Maximum-intensity projections (MIP) from the same dataset are displayed in Figs.3d-f chosen to match standard radiographic projections (i.e. frontal, oblique and lateral). The bones of the forefoot, including metatarsals and phalanges are well-depicted. The volume rendition of the imaged foot is displayed Fig.3g.


Optimized long-T2 suppressed PETRA is able to selectively map sub-millisecond-T2 tissues at isotropic resolution in clinically practical scan times, by combining multiple readouts after each inversion, along with phase-modulated RF excitation and fast non-iterative reconstruction.


NIH grants RO1 AR50068, R01-AG038693, R21-NS082953. Howard Hughes Medical Institute (HHMI) International Student Research Fellowship.


1. Robson MD et al. JCAT2003:825-846; 2. Weiger M et al. MRM2011:379-389; 3. Grodzki DM et al. MRM2012:510-518; 4. Wu Y et al. MRM2007:554-67; 5. Idiyatullin D et al. JMR2006:342-349. 6. Li C et al. MRM2012:680-689; 7. Li C et al. ISMRM2015, p3733.


Pulse sequence diagram of IR-ZTE

Tibia images in the axial, coronal and sagittal planes (a-c), along with the matching images (d-f) obtained with a conventional gradient-echo sequence.

(a) axial, (b) coronal and (c) sagittal images of the forefoot obtained at isotropic voxel size of ~1 mm3. Maximum-intensity projections (MIP) from the same dataset are displayed to match standard radiographic projections: frontal (d), oblique (e), lateral (f), along with volume rendition (g). Note the two small sesamoid bones (arrows in (e) and (f))at the distal end of the first metatarsal bone.

Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)