Purpose
Respiration induces temporal variations of the main magnetic field B0 along the spinal cord. These variations are typically not compensated for in velocity quantifications using ...phase‐contrast MRI. The goal of this study was to analyze errors caused by respiration‐induced B0 variations in real‐time phase‐contrast echo planar imaging (PCEPI) of cervical cerebrospinal fluid (CSF) velocity measurements and to evaluate this effect for various sequence parameters using numerical simulations.
Methods
Real‐time B0 measurements with double gradient echo sequence and PCEPI measurements were acquired in the cervical CSF of 10 healthy subjects. Dynamic phase offsets attributed to respiration‐induced B0 variations were analyzed by quantifying amplitudes and comparing the temporal behavior with respiratory signals. In experiments and simulations, the influence of the echo time (TE) and the delay between PCEPI images (Δt) with respect to respiration on the dynamic phase offsets were investigated.
Results
A good agreement was found between phase offsets extracted from both acquisition types. Furthermore, respiratory signals qualitatively matched the temporal behavior of the measured phase offsets showing a dependency on subject‐dependent local B0 distribution and respiration physiology. Simulations revealed residual background phases in PCEPI velocity quantification varying with TE and Δt.
Conclusion
Respiration‐induced B0 variations result in dynamic background phases in real‐time PCEPI velocity quantifications of the CSF in the cervical spine. The current work underlines that these background phases need to be corrected to avoid confounding effects.
Three-dimensional time-resolved phase-contrast cardiovascular magnetic resonance (4D flow CMR) enables the quantification and visualisation of blood flow, but its clinical applicability remains ...hampered by its long scan time. The aim of this study was to evaluate the use of compressed sensing (CS) with on-line reconstruction to accelerate the acquisition and reconstruction of 4D flow CMR of the thoracic aorta.
4D flow CMR of the thoracic aorta was acquired in 20 healthy subjects using CS with acceleration factors ranging from 4 to 10. As a reference, conventional parallel imaging (SENSE) with acceleration factor 2 was used. Flow curves, net flows, peak flows and peak velocities were extracted from six contours along the aorta. To measure internal data consistency, a quantitative particle trace analysis was performed. Additionally, scan-rescan, inter- and intraobserver reproducibility were assessed. Subsequently, 4D flow CMR with CS factor 6 was acquired in 3 patients with differing aortopathies. The flow patterns resulting from particle trace visualisation were qualitatively analysed.
All collected data were successfully acquired and reconstructed on-line. The average acquisition time including respiratory navigator efficiency with CS factor 6 was 5:02 ± 2:23 min while reconstruction took approximately 9 min. For CS factors of 8 or less, mean differences in net flow, peak flow and peak velocity as compared to SENSE were below 2.2 ± 7.8 ml/cycle, 4.6 ± 25.2 ml/s and - 7.9 ± 13.0 cm/s, respectively. For a CS factor of 10 differences reached 5.4 ± 8.0 ml/cycle, 14.4 ± 28.3 ml/s and - 4.0 ± 12.2 cm/s. Scan-rescan analysis yielded mean differences in net flow of - 0.7 ± 4.9 ml/cycle for SENSE and - 0.2 ± 8.5 ml/cycle for CS factor of 6.
A six- to eightfold acceleration of 4D flow CMR using CS is feasible. Up to a CS acceleration rate of 6, no statistically significant differences in measured flow parameters could be observed with respect to the reference technique. Acquisitions in patients with aortopathies confirm the potential to integrate the proposed method in a clinical routine setting, whereby its main benefits are scan-time savings and direct on-line reconstruction.
Objectives
Evaluation of imaging performance across dual-energy CT (DECT) platforms, including dual-layer CT (DLCT), rapid-kVp-switching CT (KVSCT) and dual-source CT (DSCT).
Methods
A ...semi-anthropomorphic abdomen phantom was imaged on these DECT systems. Scans were repeated three times for CTDIvol levels of 10 mGy, 20 mGy, 30 mGy and different fat-simulating extension rings. Over the available range of virtual-monoenergetic images (VMI), noise as well as quantitative accuracy of hounsfield units (HU) and iodine concentrations were evaluated.
Results
For all VMI levels, HU values could be determined with high accuracy compared to theoretical values. For KVSCT and DSCT, a noise increase was observed towards lower VMI levels. A patient-size dependent increase in the uncertainty of quantitative iodine concentrations is observed for all platforms. For a medium patient size the iodine concentration root-mean-square deviation at 20 mGy is 0.17 mg/ml (DLCT), 0.30 mg/ml (KVSCT) and 0.77mg/ml (DSCT).
Conclusion
Noticeable performance differences are observed between investigated DECT systems. Iodine concentrations and VMI HUs are accurately determined across all DECT systems. KVSCT and DLCT deliver slightly more accurate iodine concentration values than DSCT for investigated scenarios. In DLCT, low-noise and high-image contrast at low VMI levels may help to increase diagnostic information in abdominal CT.
Key Points
• Current dual-energy CT platforms provide accurate, reliable quantitative information.
• Dual-energy CT cross-platform evaluation revealed noticeable performance differences between different systems.
• Dual-layer CT offers constant noise levels over the complete energy range.
In aneurysmal subarachnoid hemorrhage (aSAH), accurate diagnosis of aneurysm is essential for subsequent treatment to prevent rebleeding. However, aneurysm detection proves to be challenging and ...time-consuming. The purpose of this study was to develop and evaluate a deep learning model (DLM) to automatically detect and segment aneurysms in patients with aSAH on computed tomography angiography. In this retrospective single-center study, three different DLMs were trained on 68 patients with 79 aneurysms treated for aSAH (2016-2017) using five-fold-cross-validation. Their outputs were combined to a single DLM via ensemble-learning. The DLM was evaluated on an independent test set consisting of 185 patients with 215 aneurysms (2010-2015). Independent manual segmentations of aneurysms in a 3D voxel-wise manner by two readers (neurosurgeon, radiologist) provided the reference standard. For aneurysms > 30 mm
(mean diameter of ~ 4 mm) on the test set, the DLM provided a detection sensitivity of 87% with false positives (FPs)/scan of 0.42. Automatic segmentations achieved a median dice similarity coefficient (DSC) of 0.80 compared to the reference standard. Aneurysm location (anterior vs. posterior circulation; P = .07) and bleeding severity (Fisher grade ≤ 3 vs. 4; P = .33) did not impede detection sensitivity or segmentation performance. For aneurysms > 100 mm
(mean diameter of ~ 6 mm), a sensitivity of 96% with DSC of 0.87 and FPs/scan of 0.14 were obtained. In the present study, we demonstrate that the proposed DLM detects and segments aneurysms > 30 mm
in patients with aSAH with high sensitivity independent of cerebral circulation and bleeding severity while producing FP findings of less than one per scan. Hence, the DLM can potentially assist treating physicians in aSAH by providing automated detection and segmentations of aneurysms.
•Obesity and body composition determined in initial LDCT is a risk factor for SARS‐CoV‐2 infected patients.•An initial LDCT can be used to screen opportunistically for obese COVID-19 ...patients.•Unfavorable body composition is associated with increased risk for the need of intensive care treatment.
Low-dose computed tomography (LDCT) of the chest is a recommended diagnostic tool in early stage of COVID-19 pneumonia. High age, several comorbidities as well as poor physical fitness can negatively influence the outcome within COVID-19 infection. We investigated whether the ratio of fat to muscle area, measured in initial LDCT, can predict severe progression of COVID-19 in the follow-up period.
We analyzed 58 individuals with confirmed COVID-19 infection that underwent an initial LDCT in one of two included centers due to COVID-19 infection. Using the ratio of waist circumference per paravertebral muscle circumference (FMR), the body composition was estimated. Patient outcomes were rated on an ordinal scale with higher numbers representing more severe progression or disease associated complications (hospitalization/ intensive care unit (ICU)/ tracheal intubation/ death) within a follow-up period of 22 days after initial LDCT.
In the initial LDCT a significantly higher FMR was found in patients requiring intensive care treatment within the follow-up period. In multivariate logistic regression analysis, FMR (p < .001) in addition to age (p < .01), was found to be a significant predictor of the necessity for ICU treatment of COVID-19 patients.
FMR as potential surrogate of body composition and obesity can be easily determined in initial LDCT of COVID-19 patients. Within the multivariate analysis, in addition to patient age, low muscle area in proportion to high fat area represents an additional prognostic information for the patient outcome and the need of an ICU treatment during the follow-up period within the next 22 days.
This multicentric pilot study presents a method using an initial LDCT to screen opportunistically for obese patients who have an increased risk for the need of ICU treatment. While clinical capacities, such as ICU beds and ventilators, are more crucial than ever to help manage the current global corona pandemic, this work introduces an approach that can be used for a cost-effective way to help determine the amount of these rare clinical resources required in the near future.
Deep learning-based algorithms have demonstrated enormous performance in segmentation of medical images. We collected a dataset of multiparametric MRI and contour data acquired for use in ...radiosurgery, to evaluate the performance of deep convolutional neural networks (DCNN) in automatic segmentation of brain metastases (BM).
A conventional U-Net (cU-Net), a modified U-Net (moU-Net) and a U-Net trained only on BM smaller than 0.4 ml (sU-Net) were implemented. Performance was assessed on a separate test set employing sensitivity, specificity, average false positive rate (AFPR), the dice similarity coefficient (DSC), Bland-Altman analysis and the concordance correlation coefficient (CCC).
A dataset of 509 patients (1223 BM) was split into a training set (469 pts) and a test set (40 pts). A combination of all trained networks was the most sensitive (0.82) while maintaining a specificity 0.83. The same model achieved a sensitivity of 0.97 and a specificity of 0.94 when considering only lesions larger than 0.06 ml (75% of all lesions). Type of primary cancer had no significant influence on the mean DSC per lesion (p = 0.60). Agreement between manually and automatically assessed tumor volumes as quantified by a CCC of 0.87 (95% CI, 0.77-0.93), was excellent.
Using a dataset which properly captured the variation in imaging appearance observed in clinical practice, we were able to conclude that DCNNs reach clinically relevant performance for most lesions. Clinical applicability is currently limited by the size of the target lesion. Further studies should address if small targets are accurately represented in the test data.
Abstract Objective The aim of the study was to investigate the performance and diagnostic value of metal artifact reduction in virtual monoenergetic images generated from dual-layer computed ...tomography (DLCT). Methods 35 patients that received a DLCT at the University Hospital Cologne and had an orthopedic implant in the examined region were included in this study. For each DLCT virtual monoenergetic images of different energy levels (64 keV, 70 keV, 105 keV, 140 keV, 200 keV and an optimized photon energy) were reconstructed and analyzed by three blinded observers. Images were analyzed with regard to subjective criteria (extent of artifacts, diagnostic image quality) and objective criteria (width and density of artifacts). Results 21 patients had implants in the spine, 8 in the pelvis and 6 patients in the extremities. Diagnostic image quality improved significantly at high photon energies from a Likert-score of 4.3 (± 0.83) to 2.3 (± 1.02) and artifacts decreased significantly from a score of 4.3 (± 0.66) to 2.6 (± 2.57). The average optimized photon energy was 149.2 ± 39.4 keV. The density as well as the width of the most pronounced artifacts decreased from -374.6 ± 251.89 HU to -12.5 ± 205.84 HU and from 14.5 ± 8.74 mm to 6.4 ± 10.76 mm, respectively. Conclusion Using virtual monoenergetic images valuable improvements of diagnostic image quality can be achieved by reduction of artifacts associated with metal implants. As preset for virtual monoenergetic images, 140 keV appear to provide optimal artifact reduction. In 20% of the patients, individually optimized keV can lead to a further improvement of image quality compared to 140 keV.
Background
Decreasing MRI scan time is a key factor to increase patient comfort and compliance as well as the productivity of MRI scanners.
Purpose/Hypothesis
Compressed sensing (CS) should ...significantly accelerate 3D scans. This study evaluated the clinical application and cost effectiveness of accelerated 3D T2 sequences of the lumbar spine.
Study Type
Prospective, cross‐sectional, observational.
Population
Twenty healthy volunteers and 10 patients.
Field Strength/Sequence
A 3D T2 TSE sequence, identical 3D sequences with three different parallel imaging and CS accelerating factors, and 2D TSE sequences as a clinical reference were obtained on a 3T scanner.
Assessment
Three readers evaluated the sequences for delineation of anatomical structures and image quality. A quantitative analysis consisting of root mean square error, structural similarity index, signal‐to‐noise ratio, and contrast‐to‐noise ratio were performed. The scan times were used to calculate cost differences for each sequence.
Statistical Tests
An analysis of variance with repeated measurements and the Friedman test were used to test for potential differences between the sequences. Post‐hoc analysis was made with the chi‐squared and Tukey–Kramer test.
Results
CS with factor 4.5 results in unchanged image quality compared to the T2 TSE for volunteers and patients (overall image impression: 4.75 vs. 4.20 P = 0.73 and 4.90 vs. 4.47 P = 0.44). The CS 4.5 scan is 167 seconds (–39%) faster than the 3D and 216.5 seconds (–45%) faster than the 2D sequences. No significant differences was found for the diagnostic certainty in the volunteers and patients between 2D TSE and 3D CS 4.5 (P = 0.89 and P = 0.43). A reduction of scan time to 148 seconds (CS 8) was still rated acceptable for most diagnosis.
Data Conclusion
CS accelerates the 3D T2 without compromising image quality. The 3D sequences offer comparable diagnostic quality to the clinical 2D standard with less scan time (–45%), potentially increasing the productivity of MRI scanners.
Level of Evidence: 1
Technical Efficacy: Stage 6 J. Magn. Reson. Imaging 2019;49:e164–e175.