Correction of Motion Artifacts Using a Multiscale Fully Convolutional Neural Network

BACKGROUND AND PURPOSE

Motion artifacts are a frequent source of image degradation in the clinical application of MR imaging (MRI). Here we implement and validate an MRI motion-artifact correction method using a multiscale fully convolutional neural network.

MATERIALS AND METHODS

The network was trained to identify motion artifacts in axial T2-weighted spin-echo images of the brain. Using an extensive data augmentation scheme and a motion artifact simulation pipeline, we created a synthetic training dataset of 93,600 images based on only 16 artifact-free clinical MRI cases. A blinded reader study using a unique test dataset of 28 additional clinical MRI cases with real patient motion was conducted to evaluate the performance of the network.

RESULTS

Application of the network resulted in notably improved image quality without the loss of morphologic information. For synthetic test data, the average reduction in mean squared error was 41.84%. The blinded reader study on the real-world test data resulted in significant reduction in mean artifact scores across all cases (P < .03).

CONCLUSIONS

Retrospective correction of motion artifacts using a multiscale fully convolutional network is promising and may mitigate the substantial motion-related problems in the clinical MRI workflow.

Spiral T1 Spin-Echo for Routine Postcontrast Brain MRI Exams: A Multicenter Multireader Clinical Evaluation

BACKGROUND AND PURPOSE

Spiral MR imaging has several advantages compared with Cartesian MR imaging that can be leveraged for added clinical value. A multicenter multireader study was designed to compare spiral with standard-of-care Cartesian postcontrast structural brain MR imaging on the basis of relative performance in 10 metrics of image quality, artifact prevalence, and diagnostic benefit.

MATERIALS AND METHODS

Seven clinical sites acquired 88 total subjects. For each subject, sites acquired 2 postcontrast MR imaging scans: a spiral 2D T1 spin-echo, and 1 of 4 routine Cartesian 2D T1 spin-echo/TSE scans (fully sampled spin-echo at 3T, 1.5T, partial Fourier, TSE). The spiral acquisition matched the Cartesian scan for scan time, geometry, and contrast. Nine neuroradiologists independently reviewed each subject, with the matching pair of spiral and Cartesian scans compared side-by-side, and scored on 10 image-quality metrics (5-point Likert scale) focused on intracranial assessment. The Wilcoxon signed rank test evaluated relative performance of spiral versus Cartesian, while the Kruskal-Wallis test assessed interprotocol differences.

RESULTS

Spiral was superior to Cartesian in 7 of 10 metrics (flow artifact mitigation, SNR, GM/WM contrast, image sharpness, lesion conspicuity, preference for diagnosing abnormal enhancement, and overall intracranial image quality), comparable in 1 of 10 metrics (motion artifacts), and inferior in 2 of 10 metrics (susceptibility artifacts, overall extracranial image quality) related to magnetic susceptibility (P < .05). Interprotocol comparison confirmed relatively higher SNR and GM/WM contrast for partial Fourier and TSE protocol groups, respectively (P < .05).

CONCLUSIONS

Spiral 2D T1 spin-echo for routine structural brain MR imaging is feasible in the clinic with conventional scanners and was preferred by neuroradiologists for overall postcontrast intracranial evaluation.

Optimal Fat Suppression in Head and Neck MRI: Comparison of Multipoint Dixon with 2 Different Fat-Suppression Techniques, Spectral Presaturation and Inversion Recovery, and STIR

BACKGROUND AND PURPOSE

Uniform complete fat suppression is essential for identification and characterization of most head and pathology. Our aim was to compare the multipoint Dixon turbo spin-echo fat-suppression technique with 2 different fat-suppression techniques, including a hybrid spectral presaturation with inversion recovery technique and an inversion recovery STIR technique, in head and neck fat-suppression MR imaging.

MATERIALS AND METHODS

Head and neck MR imaging datasets of 72 consecutive patients were retrospectively reviewed. All patients were divided into 2 groups based on the type of fat-suppression techniques used (group A: STIR and spectral presaturation with inversion recovery gadolinium-T1WI; group B: multipoint Dixon T2 TSE and multipoint Dixon gadolinium-T1WI TSE). Objective and subjective image quality and scan acquisition times were assessed and compared between multipoint Dixon T2 TSE versus STIR and multipoint Dixon gadolinium-T1WI TSE versus spectral presaturation with inversion recovery gadolinium-T1WI using the Mann-Whitney U test.

RESULTS

A total of 64 patients were enrolled in the study (group A, n = 33 and group B, n = 31). Signal intensity ratios were significantly higher for multipoint Dixon T2 and gadolinium-T1WI techniques compared with STIR (P < .001) and spectral presaturation with inversion recovery gadolinium-T1WI (P < .001), respectively. Two independent blinded readers revealed that multipoint Dixon T2 and gadolinium-T1WI techniques had significantly higher overall image quality (P = .022 and P < .001) and fat-suppression grades (P < .013 and P < .001 across 3 different regions) than STIR and spectral presaturation with inversion recovery gadolinium-T1WI, respectively. The scan acquisition time was relatively short for the multipoint Dixon technique (2 minutes versus 4 minutes 56 seconds for the T2-weighted sequence and 2 minutes versus 3 minutes for the gadolinium-T1WI sequence).

CONCLUSIONS

The multipoint Dixon technique offers better image quality and uniform fat suppression at a shorter scan time compared with STIR and spectral presaturation with inversion recovery gadolinium-T1WI techniques.

Medial Occipital Lobe Hyperperfusion Identified by Arterial Spin-Labeling: A Poor Prognostic Sign in Patients with Hypoxic-Ischemic Encephalopathy

Hypoxic-ischemic encephalopathy carries an uncertain prognosis. We sought to retrospectively assess the prognostic value of arterial spin-labeling MR imaging in 22 adult patients diagnosed with hypoxic-ischemic encephalopathy. Quantitative CBF maps were generated from the M0 map, and arterial spin-labeling data on a per-voxel basis were regionally interrogated via visual inspection and ROI placement. Hyperperfusion was defined as regional increases in CBF of >20% (relative to global CBF) and/or >100 mL/100 g/min. Eleven of 22 patients had prominent bilateral medial occipital lobe hyperperfusion, all of whom died before hospital discharge. One patient who had nondistinct arterial spin-labeling hyperperfusion and restricted diffusion survived. Medial occipital lobe hyperperfusion is a distinctive pattern that merits prospective investigation in a cohort of patients with moderate hypoxic-ischemic encephalopathy to determine its predictive ability in patients with a higher likelihood of survival.

Impact of model-based iterative reconstruction on image quality of contrast-enhanced neck CT

BACKGROUND AND PURPOSE

Improved image quality is clinically desired for contrast-enhanced CT of the neck. We compared 30% adaptive statistical iterative reconstruction and model-based iterative reconstruction algorithms for the assessment of image quality of contrast-enhanced CT of the neck.

MATERIALS AND METHODS

Neck contrast-enhanced CT data from 64 consecutive patients were reconstructed retrospectively by using 30% adaptive statistical iterative reconstruction and model-based iterative reconstruction. Objective image quality was assessed by comparing SNR, contrast-to-noise ratio, and background noise at levels 1 (mandible) and 2 (superior mediastinum). Two independent blinded readers subjectively graded the image quality on a scale of 1-5, (grade 5 = excellent image quality without artifacts and grade 1 = nondiagnostic image quality with significant artifacts). The percentage of agreement and disagreement between the 2 readers was assessed.

RESULTS

Compared with 30% adaptive statistical iterative reconstruction, model-based iterative reconstruction significantly improved the SNR and contrast-to-noise ratio at levels 1 and 2. Model-based iterative reconstruction also decreased background noise at level 1 (P = .016), though there was no difference at level 2 (P = .61). Model-based iterative reconstruction was scored higher than 30% adaptive statistical iterative reconstruction by both reviewers at the nasopharynx (P < .001) and oropharynx (P < .001) and for overall image quality (P < .001) and was scored lower at the vocal cords (P < .001) and sternoclavicular junction (P < .001), due to artifacts related to thyroid shielding that were specific for model-based iterative reconstruction.

CONCLUSIONS

Model-based iterative reconstruction offers improved subjective and objective image quality as evidenced by a higher SNR and contrast-to-noise ratio and lower background noise within the same dataset for contrast-enhanced neck CT. Model-based iterative reconstruction has the potential to reduce the radiation dose while maintaining the image quality, with a minor downside being prominent artifacts related to thyroid shield use on model-based iterative reconstruction.

Diffusion-weighted imaging with dual-echo echo-planar imaging for better sensitivity to acute stroke

BACKGROUND AND PURPOSE

Parallel imaging facilitates the acquisition of echo-planar images with a reduced TE, enabling the incorporation of an additional image at a later TE. Here we investigated the use of a parallel imaging-enhanced dual-echo EPI sequence to improve lesion conspicuity in diffusion-weighted imaging.

MATERIALS AND METHODS

Parallel imaging-enhanced dual-echo DWI data were acquired in 50 consecutive patients suspected of stroke at 1.5T. The dual-echo acquisition included 2 EPI for 1 diffusion-preparation period (echo 1 [TE = 48 ms] and echo 2 [TE = 105 ms]). Three neuroradiologists independently reviewed the 2 echoes by using the routine DWI of our institution as a reference. Images were graded on lesion conspicuity, diagnostic confidence, and image quality. The apparent diffusion coefficient map from echo 1 was used to validate the presence of acute infarction. Relaxivity maps calculated from the 2 echoes were evaluated for potential complementary information.

RESULTS

Echo 1 and 2 DWIs were rated as better than the reference DWI. While echo 1 had better image quality overall, echo 2 was unanimously favored over both echo 1 and the reference DWI for its high sensitivity in detecting acute infarcts.

CONCLUSIONS

Parallel imaging-enhanced dual-echo diffusion-weighted EPI is a useful method for evaluating lesions with reduced diffusivity. The long TE of echo 2 produced DWIs that exhibited superior lesion conspicuity compared with images acquired at a shorter TE. Echo 1 provided higher SNR ADC maps for specificity to acute infarction. The relaxivity maps may serve to complement information regarding blood products and mineralization.

Improved T2* imaging without increase in scan time: SWI processing of 2D gradient echo

BACKGROUND AND PURPOSE

2D gradient-echo imaging is sensitive to T2* lesions (hemorrhages, mineralization, and vascular lesions), and susceptibility-weighted imaging is even more sensitive, but at the cost of additional scan time (SWI: 5-10 minutes; 2D gradient-echo: 2 minutes). The long acquisition time of SWI may pose challenges in motion-prone patients. We hypothesized that 2D SWI/phase unwrapped images processed from 2D gradient-echo imaging could improve T2* lesion detection.

MATERIALS AND METHODS

2D gradient-echo brain images of 50 consecutive pediatric patients (mean age, 8 years) acquired at 3T were retrospectively processed to generate 2D SWI/phase unwrapped images. The 2D gradient-echo and 2D SWI/phase unwrapped images were compared for various imaging parameters and were scored in a blinded fashion.

RESULTS

Of 50 patients, 2D gradient-echo imaging detected T2* lesions in 29 patients and had normal findings in 21 patients. 2D SWI was more sensitive than standard 2D gradient-echo imaging in detecting T2* lesions (P < .0001). 2D SWI/phase unwrapped imaging also improved delineation of normal venous structures and nonpathologic calcifications and helped distinguish calcifications from hemorrhage. A few pitfalls of 2D SWI/phase unwrapped imaging were noted, including worsened motion and dental artifacts and challenges in detecting T2* lesions adjacent to calvaria or robust deoxygenated veins.

CONCLUSIONS

2D SWI and associated phase unwrapped images processed from standard 2D gradient-echo images were more sensitive in detecting T2* lesions and delineating normal venous structures and nonpathologic mineralization, and they also helped distinguish calcification at no additional scan time. SWI processing of 2D gradient-echo images may be a useful adjunct in cases in which longer scan times of 3D SWI are difficult to implement.

Clinical assessment of standard and generalized autocalibrating partially parallel acquisition diffusion imaging: effects of reduction factor and spatial resolution

BACKGROUND AND PURPOSE

PI improves routine EPI-based DWI by enabling higher spatial resolution and reducing geometric distortion, though it remains unclear which of these is most important. We evaluated the relative contribution of these factors and assessed their ability to increase lesion conspicuity and diagnostic confidence by using a GRAPPA technique.

MATERIALS AND METHODS

Four separate DWI scans were obtained at 1.5T in 48 patients with independent variation of in-plane spatial resolution (1.88 mm(2) versus 1.25 mm(2)) and/or reduction factor (R = 1 versus R = 3). A neuroradiologist with access to clinical history and additional imaging sequences provided a reference standard diagnosis for each case. Three blinded neuroradiologists assessed scans for abnormalities and also evaluated multiple imaging-quality metrics by using a 5-point ordinal scale. Logistic regression was used to determine the impact of each factor on subjective image quality and confidence.

RESULTS

Reference standard diagnoses in the patient cohort were acute ischemic stroke (n = 30), ischemic stroke with hemorrhagic conversion (n = 4), intraparenchymal hemorrhage (n = 9), or no acute lesion (n = 5). While readers preferred both a higher reduction factor and a higher spatial resolution, the largest effect was due to an increased reduction factor (odds ratio, 47 ± 16). Small lesions were more confidently discriminated from artifacts on R = 3 images. The diagnosis changed in 5 of 48 scans, always toward the reference standard reading and exclusively for posterior fossa lesions.

CONCLUSIONS

PI improves DWI primarily by reducing geometric distortion rather than by increasing spatial resolution. This outcome leads to a more accurate and confident diagnosis of small lesions.

Clinical evaluation of reduced field-of-view diffusion-weighted imaging of the cervical and thoracic spine and spinal cord

BACKGROUND AND PURPOSE

DWI has the potential to improve the detection and evaluation of spine and spinal cord pathologies. This study assessed whether a recently described method (rFOV DWI) adds diagnostic value in clinical patients.

MATERIALS AND METHODS

Consecutive patients undergoing clinically indicated cervical and/or thoracic spine imaging received standard anatomic sequences supplemented with sagittal rFOV DWI by using a b-value of 500 s/mm(2). Two neuroradiologists blinded to clinical history evaluated the standard anatomic sequences only for pathology and provided their level of confidence in their diagnosis. These readers then rescored the examinations after reviewing the rFOV DWI study and indicated whether this sequence altered findings or confidence levels.

RESULTS

Two hundred twenty-three patients were included in this study. One hundred eighty patient scans (80.7%) demonstrated at least 1 pathologic finding. Interobserver agreement for identifying pathology (κ = 0.77) and in assessing the added value of the rFOV DWI sequence (κ = 0.77) was high. In pathologic cases, the rFOV DWI sequence added clinical utility in 33% of cases (P < .00001, Fisher exact test). The rFOV DWI sequence was found to be helpful in the evaluation of acute infarction, demyelination, infection, neoplasm, and intradural and epidural collections (P < .001, χ(2) test) and provided a significant increase in clinical confidence in the evaluation of 11 of the 15 pathologic subtypes assessed (P < .05, 1-sided paired Wilcoxon test).

CONCLUSIONS

rFOV diffusion-weighted imaging of the cervical and thoracic spine is feasible in a clinical population and increases clinical confidence in the diagnosis of numerous common spinal pathologies.

Reduced field-of-view diffusion imaging of the human spinal cord: comparison with conventional single-shot echo-planar imaging

BACKGROUND AND PURPOSE

DWI of the spinal cord is challenging because of its small size and artifacts associated with the most commonly used clinical imaging method, SS-EPI. We evaluated the performance of rFOV spinal cord DWI and compared it with the routine fFOV SS-EPI in a clinical population.

MATERIALS AND METHODS

Thirty-six clinical patients underwent 1.5T MR imaging examination that included rFOV SS-EPI DWI of the cervical spinal cord as well as 2 comparison diffusion sequences: fFOV SS-EPI DWI normalized for either image readout time (low-resolution fFOV) or spatial resolution (high-resolution fFOV). ADC maps were created and compared between the methods by using single-factor analysis of variance. Two neuroradiologists blinded to sequence type rated the 3 DWI methods, based on susceptibility artifacts, perceived spatial resolution, signal intensity-to-noise ratio, anatomic detail, and clinical utility.

RESULTS

ADC values for the rFOV and both fFOV sequences were not statistically different (rFOV: 1.01 ± 0.18 × 10(-3) mm(2)/s; low-resolution fFOV: 1.12 ± 0.22 × 10(-3) mm(2)/s; high-resolution fFOV: 1.10 ± 0.21 × 10(-3) mm(2)/s; F = 2.747, P > .05). The neuroradiologist reviewers rated the rFOV diffusion images superior in terms of all assessed measures (P < 0.0001). Particular improvements were noted in patients with metal hardware, degenerative disease, or both.

CONCLUSIONS

rFOV DWI of the spinal cord overcomes many of the problems associated with conventional fFOV SS-EPI and is feasible in a clinical population. From a clinical standpoint, images were deemed superior to those created by using standard fFOV methods.

Peritumoral apparent diffusion coefficient as a metric of response in patients with recurrent glioblastoma multiforme treated with bevacizumab and irinotecan

Bevacizumab and irinotecan have shown promising results in patients with recurrent glioblastoma multiforme (GBM), which traditionally carries a poor prognosis after first-line therapies have been exhausted. Retrospectively documenting the short-term effects of this chemotherapeutic regimen on recurrent GBM, as evidenced by comparative magnetic resonance images obtained two weeks prior to, and one-month following initiation of treatment, we hypothesize that peritumoral apparent diffusion coefficient (ADC) values will decrease on post-treatment scans. Brain MR data were collected from August 2005 to December 2006, in which post-contrast T1-weighted images demonstrated measurable enhancement or GBM tumor mass. Pre- and post-treatment MR images for ten consecutive patients were collected, each having failed temozolomide and radiation therapy. Pre- and post-treatment recurrent GBM bulk tumor and peritumoral T2 signal abnormality were measured in three dimensions. Diffusion of peritumoral T2 signal abnormality was evaluated on pre- and post-treatment ADC. All patients witnessed a significant decrease in tumor bulk ranging from 15.3% to 96.7% with a mean reduction of 48.2%, having received an average of two cycles of chemotherapy. FLAIR images demonstrated a mean volumetric reduction in peritumoral T2 signal abnormality of 44.3%. ADC measurements demonstrated an average reduction in peritumoral ADC of 20.6%, which was statistically significant (p-value < .005). Recurrent GBM tumor bulk demonstrated a 48.2% mean reduction, with corresponding decrease in peritumoral ADC values of 20.6%, suggesting that ADC may represent a valuable metric in the evaluation of the chemotherapeutic response of recurrent GBM, when treated with bevacizumab and irinotecan.