Synthesizing Contrast-Enhanced MR Images from Noncontrast MR Images Using Deep Learning.

BACKGROUND AND PURPOSE: Recent developments in deep learning methods offer a potential solution to the need for alternative imaging methods due to concerns about the toxicity of gadolinium-based contrast agents. The purpose of the study was to synthesize virtual gadolinium contrast-enhanced T1-weighted MR images from noncontrast multiparametric MR images in patients with primary brain tumors by using deep learning. MATERIALS AND METHODS: We trained and validated a deep learning network by using MR images from 335 subjects in the Brain Tumor Segmentation Challenge 2019 training data set. A held out set of 125 subjects from the Brain Tumor Segmentation Challenge 2019 validation data set was used to test the generalization of the model. A residual inception DenseNet network, called T1c-ET, was developed and trained to simultaneously synthesize virtual contrast-enhanced T1-weighted (vT1c) images and segment the enhancing portions of the tumor. Three expert neuroradiologists independently scored the synthesized vT1c images by using a 3-point Likert scale, evaluating image quality and contrast enhancement against ground truth T1c images (1 = poor, 2 = good, 3 = excellent). RESULTS: The synthesized vT1c images achieved structural similarity index, peak signal-to-noise ratio, and normalized mean square error scores of 0.91, 64.35, and 0.03, respectively. There was moderate interobserver agreement between the 3 raters, regarding the algorithm's performance in predicting contrast enhancement, with a Fleiss kappa value of 0.61. Our model was able to accurately predict contrast enhancement in 88.8% of the cases (scores of 2 to 3 on the 3-point scale). CONCLUSIONS: We developed a novel deep learning architecture to synthesize virtual postcontrast enhancement by using only conventional noncontrast brain MR images. Our results demonstrate the potential of deep learning methods to reduce the need for gadolinium contrast in the evaluation of primary brain tumors.

Structural MRI Ratios Fail to Distinguish Progressive Supranuclear Palsy From Parkinson Disease in Individual Patients.

Background and Objectives: Parkinson disease (PD) and progressive supranuclear palsy (PSP) are often difficult to differentiate in the clinic. The MR parkinsonism index (MRPI) has been recommended to assist in making this distinction. We aimed to assess the usefulness of this tool in our real-world practice of movement disorders. Methods: We prospectively obtained MRI scans on consecutive patients with movement disorders with a clinical indication for imaging and obtained measures of MRI regions of interest (ROIs) from our neuroradiologists. The authors reviewed all MRI scans and corrected any errors in the original ROI drawings for this analysis. We retrospectively assigned diagnoses using established consensus criteria from progress notes stored in our electronic medical record. We analyzed the data using multinomial logistic regression models and receiver operating curve analysis to determine the predictive accuracy of the MRI ratios. Results: MRI measures and consensus diagnoses were available on 130 patients with PD, 54 with PSP, and 77 diagnosed as other. The out-of-sample prediction error rate of our 5 regression models ranged from 45% to 59%. The average sensitivity and specificity of the 5 models in the testing sample were 53% and 80%, respectively. The positive predictive value of an MRPI ≥13.55 (the published cutoff) in our patients was 79%. Discussion: These results indicate that MRI measures of brain structures were not effective at predicting diagnosis in individual patients. We conclude that the search for a biomarker that can differentiate PSP from PD must continue.

A Comprehensive Analysis of Cerebellar Volumes in the 22q11.2 Deletion Syndrome.

BACKGROUND: The presence of a 22q11.2 microdeletion (22q11.2 deletion syndrome [22q11DS]) ranks among the greatest known genetic risk factors for the development of psychotic disorders. There is emerging evidence that the cerebellum is important in the pathophysiology of psychosis. However, there is currently limited information on cerebellar neuroanatomy in 22q11DS specifically. METHODS: High-resolution 3T magnetic resonance imaging was acquired in 79 individuals with 22q11DS and 70 typically developing control subjects (N = 149). Lobar and lobule-level cerebellar volumes were estimated using validated automated segmentation algorithms, and subsequently group differences were compared. Hierarchical clustering, principal component analysis, and graph theoretical models were used to explore intercerebellar relationships. Cerebrocerebellar structural connectivity with cortical thickness was examined via linear regression models. RESULTS: Individuals with 22q11DS had, on average, 17.3% smaller total cerebellar volumes relative to typically developing subjects (p < .0001). The lobules of the superior posterior cerebellum (e.g., VII and VIII) were particularly affected in 22q11DS. However, all cerebellar lobules were significantly smaller, even after adjusting for total brain volumes (all cerebellar lobules p < .0002). The superior posterior lobule was disproportionately associated with cortical thickness in the frontal lobes and cingulate cortex, brain regions known be affected in 22q11DS. Exploratory analyses suggested that the superior posterior lobule, particularly Crus I, may be associated with psychotic symptoms in 22q11DS. CONCLUSIONS: The cerebellum is a critical but understudied component of the 22q11DS neuroendophenotype.

MRI demonstration of gadolinium deposition in bone after monthly triple-dose gadopentetate dimeglumine and correlation with frequency of hypophosphatemia.

OBJECTIVES: We retrospectively analyzed data of the BECOME trial to investigate whether serial administration of triple-dose (3-dose) gadopentetate dimeglumine would result in the development of T1 signal-to-noise (S/N) changes in the cranial diploic space and whether S/N changes correlated with on-study hypophosphatemia. METHODS: Signal intensity analysis was performed on the first year's data of the BECOME trial using 3-dose Gd (14 months, maximum number of doses, 39, mean: 36). Routine blood and urine tests were obtained each month for safety monitoring. Linear mixed regression modeling with random intercept was used to analyze monthly signal-to-noise ratio (S/N = Bone/Air) using an ROI of the diploic space created from T2W images and overlaid on T1FS (T1 fat-saturated) images. Incidence of phosphate abnormalities was analyzed using the general estimation equation; correlation of phosphate and S/N change was achieved with type 3 test of fixed effects. RESULTS: Cranial diploic space T1FS S/N increased over 14 months: S/N = 0.039 mean monthly increase (S.E. 0.008; p < 0.0001). Subjects with consistently normal phosphate levels (n = 32) experienced more of a S/N increase than patients with at least one episode of hypophosphatemia (n = 35) (0.057 vs. 0.023, respectively, p = 0.037). Those with moderate hypophosphatemia demonstrated no significant S/N increase. CONCLUSION: Monthly administration of 3-dose gadopentetate dimeglumine is associated with development of increased S/N on T1FS imaging in the cranial diploic space, suggesting Gd retention in bone. Our data suggests MRI could be used as a noninvasive method of tracking Gd retention in bone, which was more pronounced in patients with normal phosphate levels.

Gray Matter Nucleus Hyperintensity After Monthly Triple-Dose Gadopentetate Dimeglumine With Long-term Magnetic Resonance Imaging.

OBJECTIVES: Gadolinium deposition is widely believed to occur, but questions regarding accumulation pattern and permanence remain. We conducted a retrospective study of intracranial signal changes on monthly triple-dose contrast-enhanced magnetic resonance imaging (MRI) examinations from the previously published Betaseron vs. Copaxone in Multiple Sclerosis With Triple-Dose Gadolinium and 3-Tesla MRI Endpoints Trial (N = 67) to characterize the dynamics of gadolinium deposition in several deep brain nuclei and track persistence versus washout of gadolinium deposition on long-term follow-up (LTFU) examinations (N = 28) obtained approximately 10 years after enrollment in the Betaseron vs. Copaxone in Multiple Sclerosis With Triple-Dose Gadolinium and 3-Tesla MRI Endpoints Trial. MATERIALS AND METHODS: Using T2 and proton density images and using image analysis software (ITK-SNAP), manual regions of interest were created ascribing boundaries of the caudate nucleus, dentate nucleus, globus pallidus, pulvinar, putamen, white matter, and air. Intensity analysis was conducted on T1-weighted fat-saturated (fat-sat) images using the FSL package. A linear rigid-body transform was calculated from the fat-sat image at each target time point to the region of interest segmentation reference time point fat-sat image. Serial MRI signal was analyzed using linear mixed regression modeling with random intercept. Annual MRI signal changes including LTFU scans were assessed with t test. RESULTS: During monthly scanning, all gray matter structures demonstrated a significant (P < 0.0001) increase in contrast-to-noise ratio. Yearly changes in deposition showed distinctive patterns for the specific nucleus: globus pallidus showed complete retention, pulvinar showed partial washout, while dentate, caudate, and putamen returned to baseline (ie, complete washout). CONCLUSIONS: Monthly increased contrast-to-noise ratio in gray matter nuclei is consistent with gadolinium deposition over time. The study also suggests that some deep gray matter nuclei permanently retain gadolinium, whereas others demonstrate washout of soluble gadolinium.