Many biological tissues exhibit such short relaxation times that their signals decay completely by the time conventional sequences begin sampling. For this reason, many tissues, especially bone, cartilage, ligaments, and tendons of the musculoskeletal system have been difficult to study using MRI. To address this issue, ultrashort-echo-time (UTE) pulse sequences have been developed to begin sampling as closely as possible to the excitation pulse. The time delay caused by the slice selection gradient has led to the development of specialized RF pulses and acquisition schemes for 2D imaging [1]; however these methods are inherently challenging and may have limited robustness. Alternate implementations of UTE sequences are based on 3D radial acquisitions with nonselective RF pulses [2]. While relatively easy to implement and perform, 3D radial acquisition schemes are slow, often requiring several minutes to collect a full 3D volume of data [3]. Qian, et al [4] demonstrated that a 3D stack-of-spirals acquisition can achieve very short echo times by beginning each spiral readout immediately after the through-plane phase-encoding gradient waveform has completed, resulting in a variable TE in the through-plane direction. For the center of k-space where the PE gradients are small (or nonexistent), the minimum TE achievable with slab-selective excitation pulses in that study was approximately 600 µs. This variable-TE stack-of-spirals method introduces some blurring in the through-plane direction in exchange for improved scan times due to the efficiency of spiral readouts. Here, we have replaced the slab-selective excitation pulse with a short nonselective hard RF pulse [5], reducing the minimum TE, and demonstrate the sequence’s utility in visualizing bone (skull) and cartilage in the knee of a normal volunteer.A prototype 3D spoiled gradient-echo sequence was developed to support a stack-of-spirals acquisition. A 60 µs nonselective hard excitation pulse was used, reducing the minimum TE to 50 µs. Maximum TE depended on number of slices and slice resolution, and was generally in the range of 250 – 400 µs (Fig. 1). The sequence’s operation was demonstrated in two settings. In the first, a human head was scanned with parameters: TR = 10 ms; TE = 50-370 µs; flip angle 5°; matrix 96x96x64; FOV 240 mm³; 98 interleaves of 1.0 ms duration each; 67-second acquisition time. Imaging was performed using a 12-channel head RF coil. A second volumetric image was obtained with a TE of 5.1 ms (to preserve fat/water phase) to provide late-TE comparison images. The second setting was a human knee, using an extremity coil, with scan parameters adjusted slightly to achieve true 1.5x1.5x1.5 mm³ isotropic resolution. Most notably, because cartilage has a longer T2* than bone, the readout duration was extended to 2.5 ms and the required number of interleaves dropped to 70, resulting in a total acquisition time of 97 seconds. All imaging was performed on a 1.5 T scanner (MAGNETOM Avanto, Siemens Healthcare, Erlangen, Germany). Informed consent was obtained prior to imaging.Figure 2 shows whole-head spiral UTE images alongside late-echo images to illustrate the difference in contrast achievable with this sequence. Direct subtractions as well as scaled subtractions [6] are shown, highlighting the bone signal. SNR, measured in a region of the frontal bone, is 54 in the minimum-TE image, 23 in the direct-subtraction image, and 43 in the scaled-subtraction image. In Fig. 3, cartilage and the meniscus is visible in the knee joint.By utilizing non-selective RF pulses, the minimum echo time achievable by a stack-of-spirals UTE sequence was reduced from 600 to 50 µs, enabling capture of signals from rapidly decaying musculoskeletal tissues. Rapid imaging may be desirable for patients who have joint or bone pain, and the efficiency of spiral readouts supports rapid generation of 3D UTE images; achieving whole-head UTE images in 67s and whole-knee images in 97s.