In conventional Magnetic Resonance Imaging (MRI), certain tissues are notoriously invisible. Cortical bone, deep articular cartilage, and the insulating myelin sheaths of neurons do not appear as structured anatomy; instead, they show up as dark, empty "signal voids."
This invisibility is not due to a lack of hydrogen protons, but rather a race against the clock. To understand why these tissues disappear—and how modern sequences can recover them—we must look at the physics of transverse relaxation and the timing of data acquisition.
The Problem of the Invisible Signal
When an MRI machine delivers a radiofrequency (RF) pulse, it tips the magnetic moments of water protons into a transverse plane, where they begin to precess in unison. This synchronized precession creates the detectable MR signal. However, this synchronization immediately begins to decay at a rate dictated by the tissue's transverse relaxation time ($T_2$ or $T_2^*$).
In soft tissues like the brain or liver, protons move relatively freely, and their $T_2$ relaxation times are long enough (tens to hundreds of milliseconds) for the scanner to comfortably record the signal. But in rigid or highly structured environments, the story changes completely:
- Cortical Bone: Protons are tightly bound within a dense mineral matrix, causing extreme local magnetic field variations that dephase the protons almost instantly, as noted by Dharshan Chandramohan et al. in their work on bone material analogues.
- Myelin Sheaths: Water trapped within the lipid bilayers of myelin experiences a similarly restricted environment, leading to rapid signal decay, as discussed by Erick J. Canales‐Rodríguez et al. in A diffusion MRI model for random walks confined on cylindrical surfaces.
- Protein-Bound Molecules: Even small molecules or drugs that are perfectly visible in a liquid solvent can experience a drastic reduction in $T_2$ time by several orders of magnitude the moment they bind to biological proteins in vivo, a phenomenon detailed by Ludger Starke et al. in First in vivo fluorine-19 magnetic resonance imaging of the multiple sclerosis drug siponimod.
In all these cases, the $T_2$ or $T_2^*$ relaxation time drops to microseconds. Because conventional MRI sequences require a modest delay—the Echo Time ($T_E$)—to apply spatial encoding gradients before listening to the signal, the magnetization in these tissues completely vanishes before data collection even begins. The scanner records nothing, resulting in a signal void.
Conventional MRI:
RF Pulse ───► [ Spatial Encoding Delay (TE) ] ───► Data Acquisition (Signal is GONE)
UTE / ZTE MRI:
RF Pulse ───► Data Acquisition (Begins Instantly)
Beating the Clock with UTE and ZTE
To capture these rapidly decaying signals, researchers utilize specialized acquisition strategies: Ultrashort Echo Time (UTE) and Zero Echo Time (ZTE) imaging. The fundamental intuition behind both techniques is to minimize or entirely eliminate the delay between RF excitation and data readout.
Instead of using traditional Cartesian grid-based sampling—which requires waiting for phase-encoding gradients to turn on and off—UTE sequences employ non-slice-selective or short-pulse RF excitations followed immediately by a radial k-space readout. As Ming Zhang et al. demonstrate in Simultaneous Quantitative Susceptibility Mapping of Articular Cartilage and Cortical Bone, sampling along radial trajectories allows data collection to start from the very center of k-space outward, mere microseconds after the RF pulse.
ZTE sequences push this concept to its logical limit. In ZTE, the spatial encoding gradients are turned on before the RF pulse is even delivered. Because the gradients are already stable, data acquisition can begin the exact moment the RF pulse ends, effectively achieving a "zero" echo time. This allows the scanner to catch the transient magnetization of short-$T_2$ structures while they are still coherent, transforming unobservable voids into measurable data points.
Resolving the Bone-Air Ambiguity in PET/MRI
One of the most impactful applications of this mechanism is solving the "missing link" in hybrid PET/MRI scanners. To produce accurate, quantitative Positron Emission Tomography (PET) images, the system must correct for how different tissues attenuate gamma rays. This requires an attenuation map, which is traditionally derived from the electron density maps provided by a CT scan.
In a standalone MRI scanner, creating this map is exceptionally difficult. As reviewed by Matthias Hofmann et al. in Towards quantitative PET/MRI, both cortical bone (which highly attenuates radiation) and air cavities (which do not attenuate radiation at all) appear as identical black voids on standard MR images.
Standard MRI: [Bone = Black Void] ↔ [Air = Black Void] (Ambiguous)
UTE/ZTE MRI: [Bone = Bright Signal] ↔ [Air = Black Void] (Differentiated)
By introducing UTE or ZTE sequences, the system can exploit the physical differences between these environments. Air contains no protons and remains dark, but bone contains rapidly decaying protons that UTE/ZTE can capture as a bright signal.
This capability allows researchers to cleanly segment bone from air. For instance, Snehashis Roy et al. utilized UTE data to generate accurate synthetic CT scans via patch-based synthesis (PET Attenuation Correction Using Synthetic CT from Ultrashort Echo-Time MR Imaging), while Jakub Baran et al. demonstrated highly accurate hybrid template-based attenuation masks (Accurate hybrid template–based and MR-based attenuation correction using UTE images).
Furthermore, the integration of ZTE sequences into deep learning models, such as the 3D U-Net framework proposed by Paul Blanc‐Durand et al., provides the precise structural data needed to generate synthetic CT maps that virtually eliminate metabolic measurement errors in brain PET/MR imaging (Attenuation correction using 3D deep convolutional neural network).
Beyond Bone: Vascular and Molecular Frontiers
The ability to capture rapid transverse relaxation has opened diagnostic avenues far beyond skeletal imaging. In neurovascular applications, blood flow and complex vessel orientations frequently cause phase dispersion and signal dropouts in standard sequences. To circumvent this, Joshua Leaston et al. utilized a quantitative UTE contrast-enhanced (QUTE-CE) MRI approach, leveraging 3D non-slice-selective pulses to minimize flow artifacts and map intricate vascular networks at high resolution (Neurovascular imaging with QUTE-CE MRI in APOE4 rats).
Similarly, in musculoskeletal radiology, Benedikt J. Schwaiger et al. demonstrated that combining UTE sequences with T1 spoiled gradient-echo imaging can yield "CT-like" structural clarity of the spine, enabling the radiation-free assessment of vertebral fractures and degenerative bone changes (CT-like images based on T1 spoiled gradient-echo and ultra-short echo time MRI sequences).
By fundamentally altering the timing of the MR experiment, ultrashort and zero echo time imaging bypass the physical limitations of rapid $T_2$ decay, turning what was once invisible noise into a rich source of quantitative clinical data.
Key Foundational Work
Towards quantitative PET/MRI: a review of MR-based attenuation correction techniques (2008) — contributed 1 concept elements to this direction.
Towards hyperpolarized 13C-succinate imaging of brain cancer (2007) — contributed 1 concept elements to this direction.