A technique for creating 'MR-sim-CT' images with positive signal in cortical bone by combining a pair of images acquired with a three-dimensional (3D) ultra-short echo-time (UTE) sequence has been demonstrated [1]. The acquisition method alternates application of differently-$$$T_2$$$-sensitised hard excitation pulses and encoding at staggered echo times (TE). Here, a method is presented for composing together the excitation pulses and acquiring a single set of gradient echoes that can be combined to generate similar contrast. Application of the composite pulse for imaging of bone reduces the net scan duration by half.

The previously-demonstrated method ([1]) is used as reference. It alternates between excitation by a 'fast' RF pulse that excites a broad range of $$$T_2$$$ signals and a 'slow' RF pulse that excites only longer-$$$T_2$$$ signals (Fig.1). The basic principle of setting excitation rate to modulate the response in $$$T_2$$$ is also previously demonstrated for long-$$$T_2$$$-suppression preparation pulses [2,3]. The reference method additionally staggers the echo times to match excitation rates. A magnitude subtraction of echo images create strong $$$T_2^*$$$ sensitivity specific to cortical bone and limited sensitivity to off-resonance phase accumulation. The subtraction is also normalised by the sum to mitigate proton density weighting, since bone has low water content relative to other tissues.

If the pair of excitation pulses is composed together with opposing polarity to create one excitation pulse (Fig.1), some of the $$$T_2$$$-specificity achieved by subtraction can be generated directly in excitation. During the slow 'tip-back' portion of the composite pulse, only longer-$$$T_2$$$ magnetisation is excited; during the fast 'tip-forth' duration all is excited in the opposite direction (Fig.2). For long-$$$T_2$$$ magnetisation the net area cancels, but for short-$$$T_2$$$ it does not. The combination pulse will also, however, tend to excite magnetisation precessing at a frequency different from that of the applied pulse because phase accrued during the pulse interferes with the cancellation effect (Figs.2,3).

Unintended off-resonance excitation can be addressed to some degree by collecting a second echo, which is also required to create the bone-specific image of the reference method. The high sensitivity of the excitation pulse to off-resonance is then mitigated for long-$$$T_2$$$ in the difference image (Fig.3b). The second echo can be acquired during the same repetition interval (TR), so only one excitation is required, and the net acquisition time is reduced by a factor of two.

A 3D UTE gradient echo sequence having facility for modifying the excitation pulse was used to acquire lower-leg images with a T/R extremity coil performing excitation and signal reception for a 1.5T commercially-available clinical scanner. The images were encoded with isotropic 1.2mm resolution and 29.4cm field-of-view using 250kHz bandwidth and 7.6ms TR, requiring 5.55min. scan time per volume encoding (i.e. 11.1min. total for separated-excitation scans).

Separated-excitation acquisitions applied 1.5$$$\mu$$$T or 24$$$\mu$$$T pulses for 512$$$\mu$$$s or 32$$$\mu$$$s respectively (nominal 12$$$^{\circ}$$$ tip angle) in an interleaved fashion. Combined, back-and-forth pulse acquisitions excited by application of '-1.5$$$\mu$$$T' ($$$-y$$$-direction) for 512$$$\mu$$$s and 24$$$\mu$$$T ($$$+y$$$-direction) for 32$$$\mu$$$s. Gradient echoes at 0.034ms and 2.034ms were recorded in both separated-excitation and composite excitation acquisitions.

Bone images from each acquisition were generated by magnitude difference of the different-pulse/different-echo images and normalisation by their sum. Images were masked to exclude voxels presumed to be air using region-growing classification formed from the late-echo image of each acquisition.

In phantom images acquired with the back-and-forth excitation, short-$$$T_2$$$ phantoms give bright signal (Fig.4). Excitation of the relatively long-$$$T_2$$$ resolution phantom and oil bottle is mostly avoided. However, near the edges of the phantoms, susceptibility-change-induced off-resonance confounds long-$$$T_2$$$-nulling.

Bone-specific images acquired with two differently-$$$T_2$$$-selective excitations show bright signal in cortical bones of the leg (Fig.5), a contrast previously demonstrated and verified as CT-similar [4]. The tibia and fibula are hyperintense relative to soft tissue; other structures between the muscles also show similar contrast. These may be tendons or epimysium and have been shown with hyperintense signal using other methods, such as inversion-recovery UTE [5].

Composing differently-$$$T_2$$$-selective excitation hard pulses into a back-and-forth excitation pulse and acquiring two different-TE gradient echoes allows generation of bone-specific images. The contrast is similar to that given by the reference method [1], but the acquisition time is halved. Because the reference spends only half its scan duration for each image used in the subtraction there is no image SNR penalty to use of the composite pulse, but the soft-tissue suppression is less robust to off-resonance or other confounds. There can be benefits though, such as reduced motion artefacts derived from the shorter duration.