Zero-TE MRI: principles and applications in the head and neck

Three contrasts in 3 min: Rapid, high-resolution, and bone-selective UTE MRI for craniofacial imaging with automated deep-learning skull segmentation

PURPOSE

Ultrashort echo time (UTE) MRI can be a radiation-free alternative to CT for craniofacial imaging of pediatric patients. However, unlike CT, bone-specific MR imaging is limited by long scan times, relatively low spatial resolution, and a time-consuming bone segmentation workflow.

METHODS

A rapid, high-resolution UTE technique for brain and skull imaging in conjunction with an automatic segmentation pipeline was developed. A dual-RF, dual-echo UTE sequence was optimized for rapid scan time (3 min) and smaller voxel size (0.65 mm3). A weighted least-squares conjugate gradient method for computing the bone-selective image improves bone specificity while retaining bone sensitivity. Additionally, a deep-learning U-Net model was trained to automatically segment the skull from the bone-selective images. Ten healthy adult volunteers (six male, age 31.5 ± 10 years) and three pediatric patients (two male, ages 12 to 15 years) were scanned at 3 T. Clinical CT for the three patients were obtained for validation. Similarities in 3D skull reconstructions relative to clinical standard CT were evaluated based on the Dice similarity coefficient and Hausdorff distance. Craniometric measurements were used to assess geometric accuracy of the 3D skull renderings.

RESULTS

The weighted least-squares method produces images with enhanced bone specificity, suppression of soft tissue, and separation from air at the sinuses when validated against CT in pediatric patients. Dice similarity coefficient overlap was 0.86 ± 0.05, and the 95th percentile Hausdorff distance was 1.77 ± 0.49 mm between the full-skull binary masks of the optimized UTE and CT in the testing dataset.

CONCLUSION

An optimized MRI acquisition, reconstruction, and segmentation workflow for craniofacial imaging was developed.

Enhanced bone assessment of the shoulder using zero-echo time MRI with deep-learning image reconstruction

OBJECTIVES

To assess a deep learning-based reconstruction algorithm (DLRecon) in zero echo-time (ZTE) MRI of the shoulder at 1.5 Tesla for improved delineation of osseous findings.

METHODS

In this retrospective study, 63 consecutive exams of 52 patients (28 female) undergoing shoulder MRI at 1.5 Tesla in clinical routine were included. Coronal 3D isotropic radial ZTE pulse sequences were acquired in the standard MR shoulder protocol. In addition to standard-of-care (SOC) image reconstruction, the same raw data was reconstructed with a vendor-supplied prototype DLRecon algorithm. Exams were classified into three subgroups: no pathological findings, degenerative changes, and posttraumatic changes, respectively. Two blinded readers performed bone assessment on a 4-point scale (0-poor, 3-perfect) by qualitatively grading image quality features and delineation of osseous pathologies including diagnostic confidence in the respective subgroups. Quantitatively, signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) of bone were measured. Qualitative variables were compared using the Wilcoxon signed-rank test for ordinal data and the McNemar test for dichotomous variables; quantitative measures were compared with Student's t-testing.

RESULTS

DLRecon scored significantly higher than SOC in all visual metrics of image quality (all, p < 0.03), except in the artifact category (p = 0.37). DLRecon also received superior qualitative scores for delineation of osseous pathologies and diagnostic confidence (p ≤ 0.03). Quantitatively, DLRecon achieved superior CNR (95 CI [1.4-3.1]) and SNR (95 CI [15.3-21.5]) of bone than SOC (p < 0.001).

CONCLUSION

DLRecon enhanced image quality in ZTE MRI and improved delineation of osseous pathologies, allowing for increased diagnostic confidence in bone assessment.

Craniofacial Imaging of Pediatric Patients by Ultrashort Echo-Time Bone-Selective MRI in Comparison to CT

RATIONALE AND OBJECTIVES

The emergence of low-dose protocols for CT imaging has mitigated pediatric radiation exposure, yet ionizing radiation remains a concern for children with complex craniofacial conditions requiring repeated radiologic monitoring. In this work, the clinical feasibility of an ultrashort echo time (UTE) MRI sequence was investigated in pediatric patients.

MATERIALS AND METHODS

Twelve pediatric patients (6 female, age range 8 to 18 years) with various imaging conditions were scanned at 3T using a dual-radiofrequency, dual-echo UTE MRI sequence. Bright-bone images were generated using a weighted least-squares conjugate gradient method to enhance bone specificity. The overlap of the binary skull masks was quantified using the Dice similarity coefficient (DSC) and the 95th percentile Hausdorff distance (HD95) to evaluate the similarity between MRI and CT. To assess the anatomic accuracy of 3D skull reconstructions, six craniometric distances were recorded and the agreement between MRI- and CT-derived measurements was evaluated using Lin's concordance correlation coefficient (ρc).

RESULTS

The bright-bone images from UTE MRI demonstrated high bone-contrast, suppression of soft tissue, and separation from air at the sinuses. The DSC and HD95 between MRI and CT had medians of 0.81 ± 0.10 and 1.87 ± 0.32 mm, respectively. There was good agreement between MRI and CT for all craniometric distances (ρc ranging from 0.90 to 0.99) with a mean absolute difference in measurements of < 2 mm.

CONCLUSION

The clinical feasibility of the UTE MRI sequence for craniofacial imaging was demonstrated in a cohort of pediatric patients, showing good agreement with CT in resolving thin bone structures and craniometry.

MRI-based Zero Echo Time and Black Bone Pseudo-CT Compared with Whole-Body CT to Detect Osteolytic Lesions in Multiple Myeloma

Background MRI is highly sensitive for assessing bone marrow involvement in multiple myeloma (MM) but does not enable detection of osteolysis. Purpose To assess the diagnostic accuracy, repeatability, and reproducibility of pseudo-CT MRI sequences (zero echo time [ZTE], gradient-echo black bone [BB]) in detecting osteolytic lesions in MM using whole-body CT as the reference standard. Materials and Methods In this prospective study, consecutive patients were enrolled in our academic hospital between June 2021 and December 2022. Inclusion criteria were newly diagnosed MM, monoclonal gammopathy of undetermined significance at high risk for MM, or suspicion of progressive MM. Participants underwent ZTE and BB sequences covering the lumbar spine, pelvis, and proximal femurs as part of 3-T whole-body MRI examinations, as well as clinically indicated fluorine 18 fluorodeoxyglucose PET/CT examination within 1 month that included optimized whole-body CT. Ten bone regions and two scores (categorical score = presence/absence of osteolytic lesion; semiquantitative score = osteolytic lesion count) were assessed by three radiologists (two experienced and one unfamiliar with pseudo-CT reading) on the ZTE, BB, and whole-body CT images. The accuracy, repeatability, and reproducibility of categorical scores (according to Gwet agreement coefficients AC1 and AC2) and differences in semiquantitative scores were assessed at the per-sequence, per-region, and per-patient levels. Results A total of 47 participants (mean age, 67 years ± 11 [SD]; 27 male) were included. In experienced readers, BB and ZTE had the same high accuracy (98%) in the per-patient analysis, while BB accuracy ranged 83%-100% and ZTE accuracy ranged 74%-94% in the per-region analysis. An increase of false-negative (FN) findings in the spine ranging from +17% up to +23%, according to the lumbar vertebra, was observed using ZTE (P < .013). Regardless of the region (except coxal bones), differences in the BB score minus the ZTE score were positively skewed (P < .021). Regardless of the sequence or region, repeatability was very good (AC1 ≥0.87 for all), while reproducibility was at least good (AC2 ≥0.63 for all). Conclusion Both MRI-based ZTE and BB pseudo-CT sequences of the lumbar spine, pelvis, and femurs demonstrated high diagnostic accuracy in detecting osteolytic lesions in MM. Compared with BB, the ZTE sequence yielded more FN findings in the spine. ClinicalTrials.gov Identifier: NCT05381077 Published under a CC BY 4.0 license. Supplemental material is available for this article.

EEG-fMRI in awake rat and whole-brain simulations show decreased brain responsiveness to sensory stimulations during absence seizures

In patients suffering absence epilepsy, recurring seizures can significantly decrease their quality of life and lead to yet untreatable comorbidities. Absence seizures are characterized by spike-and-wave discharges on the electroencephalogram associated with a transient alteration of consciousness. However, it is still unknown how the brain responds to external stimuli during and outside of seizures. This study aimed to investigate responsiveness to visual and somatosensory stimulation in Genetic Absence Epilepsy Rats from Strasbourg (GAERS), a well-established rat model for absence epilepsy. Animals were imaged under non-curarized awake state using a quiet, zero echo time, functional magnetic resonance imaging (fMRI) sequence. Sensory stimulations were applied during interictal and ictal periods. Whole-brain hemodynamic responses were compared between these two states. Additionally, a mean-field simulation model was used to explain the changes of neural responsiveness to visual stimulation between states. During a seizure, whole-brain responses to both sensory stimulations were suppressed and spatially hindered. In the cortex, hemodynamic responses were negatively polarized during seizures, despite the application of a stimulus. The mean-field simulation revealed restricted propagation of activity due to stimulation and agreed well with fMRI findings. Results suggest that sensory processing is hindered or even suppressed by the occurrence of an absence seizure, potentially contributing to decreased responsiveness during this absence epileptic process.

Large Numbers for Small Children-Up to What Age Do Infants Benefit from a Longer Echo Time in Cerebral T2 MRI Sequences?

In newborns, white matter shows a high T2-weighted (T2w) signal in MRI with poor grey-white matter contrast. To increase this contrast, an extremely long echo time (TE) is used in the examination of children. It is not known up to what age this long TE should be used. The purpose of this study was to find up to what age a long TE should be used in infants. In the prospective study, 101 infants (0-18 months) underwent cranial MRI at 3 Tesla. T2-weighted Fast Spin Echo sequences with long TE (200 ms) and medium TE (100 ms) were used. The signal intensities of the cortex and white matter were measured and the grey-white matter contrast (MC) was calculated. A cut-off age was determined. The T2w sequences with long TE had a statistically significantly higher MC until the age of six months (medium TE: 0.1 ± 0.05, Long TE: 0.19 ± 0.07; p < 0.001). After the tenth month, the T2w sequence with medium TE provided significantly better MC (Medium TE: 0.1 ± 0.05; long TE: 0.05 ± 0.4; p < 0.001). The use of a long TE is only helpful in the first six months of life. After the tenth month of life, a medium TE should be favored as is used in adult brain MRI.

Robot-Assisted Lumbar Pedicle Screw Placement Based on 3D Magnetic Resonance Imaging

STUDY DESIGN

Human Cadaveric Study.

OBJECTIVE

This study aims to explore the feasibility of using preoperative magnetic resonance imaging (MRI), zero-time-echo (ZTE) and spoiled gradient echo (SPGR), as source data for robotic-assisted spine surgery and assess the accuracy of pedicle screws.

METHODS

Zero-time-echo and SPGR MRI scans were conducted on a human cadaver. These images were manually post-processed, producing a computed tomography (CT)-like contrast. The Mazor X robot was used for lumbar pedicle screw-place navigating of MRI. The cadaver underwent a postoperative CT scan to determine the actual position of the navigated screws.

RESULTS

Ten lumbar pedicle screws were robotically navigated of MRI (4 ZTE; 6 SPGR). All MR-navigated screws were graded A on the Gertzbein-Robbins scale. Comparing preoperative robotic planning to postoperative CT scan trajectories: The screws showed a median deviation of overall 0.25 mm (0.0; 1.3), in the axial plane 0.27 mm (0.0; 1.3), and in the sagittal plane 0.24 mm (0.0; 0.7).

CONCLUSION

This study demonstrates the first successful registration of MRI sequences, ZTE and SPGR, in robotic spine surgery here used for intraoperative navigation of lumbar pedicle screws achieving sufficient accuracy, showcasing potential progress toward radiation-free spine surgery.

Cranial bone imaging using ultrashort echo-time bone-selective MRI as an alternative to gradient-echo based "black-bone" techniques

OBJECTIVES

CT is the clinical standard for surgical planning of craniofacial abnormalities in pediatric patients. This study evaluated three MRI cranial bone imaging techniques for their strengths and limitations as a radiation-free alternative to CT.

METHODS

Ten healthy adults were scanned at 3 T with three MRI sequences: dual-radiofrequency and dual-echo ultrashort echo time sequence (DURANDE), zero echo time (ZTE), and gradient-echo (GRE). DURANDE bright-bone images were generated by exploiting bone signal intensity dependence on RF pulse duration and echo time, while ZTE bright-bone images were obtained via logarithmic inversion. Three skull segmentations were derived, and the overlap of the binary masks was quantified using dice similarity coefficient. Craniometric distances were measured, and their agreement was quantified.

RESULTS

There was good overlap of the three masks and excellent agreement among craniometric distances. DURANDE and ZTE showed superior air-bone contrast (i.e., sinuses) and soft-tissue suppression compared to GRE.

DISCUSSIONS

ZTE has low levels of acoustic noise, however, ZTE images had lower contrast near facial bones (e.g., zygomatic) and require effective bias-field correction to separate bone from air and soft-tissue. DURANDE utilizes a dual-echo subtraction post-processing approach to yield bone-specific images, but the sequence is not currently manufacturer-supported and requires scanner-specific gradient-delay corrections.

Emergency MRI in Spine Trauma of Children and Adolescents-A Pictorial Review

Severe spinal trauma is uncommon in the pediatric population, but due to the potentially devastating consequences of missed injury, it poses a diagnostic challenge in emergency departments. Diagnostic imaging is often needed to exclude or confirm the injury and to assess its extent. Magnetic resonance imaging (MRI) offers an excellent view of both bony and soft tissue structures and their traumatic findings without exposing children to ionizing radiation. Our pictorial review aims to demonstrate the typical traumatic findings, physiological phenomena, and potential pitfalls of emergency MRI in the trauma of the growing spine.

Deep Learning for Synthetic CT from Bone MRI in the Head and Neck

BACKGROUND AND PURPOSE

Bone MR imaging techniques enable visualization of cortical bone without the need for ionizing radiation. Automated conversion of bone MR imaging to synthetic CT is highly desirable for downstream image processing and eventual clinical adoption. Given the complex anatomy and pathology of the head and neck, deep learning models are ideally suited for learning such mapping.

MATERIALS AND METHODS

This was a retrospective study of 39 pediatric and adult patients with bone MR imaging and CT examinations of the head and neck. For each patient, MR imaging and CT data sets were spatially coregistered using multiple-point affine transformation. Paired MR imaging and CT slices were generated for model training, using 4-fold cross-validation. We trained 3 different encoder-decoder models: Light_U-Net (2 million parameters) and VGG-16 U-Net (29 million parameters) without and with transfer learning. Loss functions included mean absolute error, mean squared error, and a weighted average. Performance metrics included Pearson R, mean absolute error, mean squared error, bone precision, and bone recall. We investigated model generalizability by training and validating across different conditions.

RESULTS

The Light_U-Net architecture quantitatively outperformed VGG-16 models. Mean absolute error loss resulted in higher bone precision, while mean squared error yielded higher bone recall. Performance metrics decreased when using training data captured only in a different environment but increased when local training data were augmented with those from different hospitals, vendors, or MR imaging techniques.

CONCLUSIONS

We have optimized a robust deep learning model for conversion of bone MR imaging to synthetic CT, which shows good performance and generalizability when trained on different hospitals, vendors, and MR imaging techniques. This approach shows promise for facilitating downstream image processing and adoption into clinical practice.