Automatic Active Lesion Tracking in Multiple Sclerosis Using Unsupervised Machine Learning.

BACKGROUND: Identifying active lesions in magnetic resonance imaging (MRI) is crucial for the diagnosis and treatment planning of multiple sclerosis (MS). Active lesions on MRI are identified following the administration of Gadolinium-based contrast agents (GBCAs). However, recent studies have reported that repeated administration of GBCA results in the accumulation of Gd in tissues. In addition, GBCA administration increases health care costs. Thus, reducing or eliminating GBCA administration for active lesion detection is important for improved patient safety and reduced healthcare costs. Current state-of-the-art methods for identifying active lesions in brain MRI without GBCA administration utilize data-intensive deep learning methods. OBJECTIVE: To implement nonlinear dimensionality reduction (NLDR) methods, locally linear embedding (LLE) and isometric feature mapping (Isomap), which are less data-intensive, for automatically identifying active lesions on brain MRI in MS patients, without the administration of contrast agents. MATERIALS AND METHODS: Fluid-attenuated inversion recovery (FLAIR), T2-weighted, proton density-weighted, and pre- and post-contrast T1-weighted images were included in the multiparametric MRI dataset used in this study. Subtracted pre- and post-contrast T1-weighted images were labeled by experts as active lesions (ground truth). Unsupervised methods, LLE and Isomap, were used to reconstruct multiparametric brain MR images into a single embedded image. Active lesions were identified on the embedded images and compared with ground truth lesions. The performance of NLDR methods was evaluated by calculating the Dice similarity (DS) index between the observed and identified active lesions in embedded images. RESULTS: LLE and Isomap, were applied to 40 MS patients, achieving median DS scores of 0.74 ± 0.1 and 0.78 ± 0.09, respectively, outperforming current state-of-the-art methods. CONCLUSIONS: NLDR methods, Isomap and LLE, are viable options for the identification of active MS lesions on non-contrast images, and potentially could be used as a clinical decision tool.

Dynamic Imaging of Blood Coagulation Within the Hematoma of Patients With Acute Hemorrhagic Stroke.

BACKGROUND: The dynamics of blood clot (combination of Hb [hemoglobin], fibrin, and a higher concentration of aggregated red blood cells) formation within the hematoma of an intracerebral hemorrhage is not well understood. A quantitative neuroimaging method of localized coagulated blood volume/distribution within the hematoma might improve clinical decision-making. METHODS: The deoxyhemoglobin of aggregated red blood cells within extravasated blood exhibits a higher magnetic susceptibility due to unpaired heme iron electrons. We propose that coagulated blood, with higher aggregated red blood cell content, will exhibit (1) a higher positive susceptibility than noncoagulated blood and (2) increase in fibrin polymerization-restricted localized diffusion, which can be measured noninvasively using quantitative susceptibility mapping and diffusion tensor imaging. In this serial magnetic resonance imaging study, we enrolled 24 patients with acute intracerebral hemorrhage between October 2021 to May 2022 at a stroke center. Patients were 30 to 70 years of age and had a hematoma volume >15 cm3 and National Institutes of Health Stroke Scale score >1. The patients underwent imaging 3×: within 12 to 24 (T1), 36 to 48 (T2), and 60 to 72 (T3) hours of last seen well on a 3T magnetic resonance imaging system. Three-dimensional anatomic, multigradient echo and 2-dimensional diffusion tensor images were obtained. Hematoma and edema volumes were calculated, and the distribution of coagulation was measured by dynamic changes in the susceptibilities and fractional anisotropy within the hematoma. RESULTS: Using a coagulated blood phantom, we demonstrated a linear relationship between the percentage coagulation and susceptibility (R2=0.91) with a positive red blood cell stain of the clot. The quantitative susceptibility maps showed a significant increase in hematoma susceptibility (T1, 0.29±0.04 parts per millions; T2, 0.36±0.04 parts per millions; T3, 0.45±0.04 parts per millions; P<0.0001). A concomitant increase in fractional anisotropy was also observed with time (T1, 0.40±0.02; T2, 0.45±0.02; T3, 0.47±0.02; P<0.05). CONCLUSIONS: This quantitative neuroimaging study of coagulation within the hematoma has the potential to improve patient management, such as safe resumption of anticoagulants, the need for reversal agents, the administration of alteplase to resolve the clot, and the need for surgery.

Cryopreserved human umbilical cord as a meningeal patch during in utero spina bifida repair in a modified ovine model.

OBJECTIVE: Despite in utero spina bifida (SB) repair, more than two-thirds of patients with SB are unable to ambulate independently, and 1 in 4 children need surgery for tethered cord by school age. The objective of this study was to test the cryopreserved human umbilical cord (HUC) as an antiscarring material to reduce tethering and improve function in a modified in utero SB repair model. METHODS: An SB defect (L2-6 levels) without myelotomy was created in fetuses of timed-pregnant ewes at gestational day (GD) 75. On GD 96, the fetal defect was exposed, and the arachnoid layer was removed to disrupt the barrier and expose the spinal cord to simulate human in utero SB repair. The fetuses were randomly assigned to two groups according to the method used to cover the spinal cord: the conventional repair (CR) group, for which myofascial closure was used (n = 7), and the HUC meningeal patch group, for which HUC was used as a meningeal patch (n = 6), followed by primary skin closure. The lambs were delivered at GD 140. Blinded clinical assessment of spinal cord function was performed using the Texas Spinal Cord Injury Scale (TSCIS). Histology of the spine was performed for quantitative assessment of spinal cord tethering, inflammatory markers, and arachnoid layer regeneration. RESULTS: The TSCIS scores were significantly lower in the CR than the HUC meningeal patch group (p = 0.0015) and the controls (p = 0.0018). The loss of spinal cord function in the CR group was mainly due to ataxia and loss of proprioception (p = 0.01 and 0.005 vs control and HUC, respectively). The histology at the repair site showed higher rates of spinal cord tethering in the CR lambs than the HUC lambs at all levels of the repair site (p = 0.01 and 0.02 vs control and HUC, respectively). In the CR with tethering compared with the HUC repair, there was a lower arachnoid layer covering at the repair site (p = 0.001). There was greater astrocyte activation in the posterior column in the CR than in the HUC repair group (p = 0.01). CONCLUSIONS: In a modified ovine SB model, the HUC as a meningeal patch allows regeneration of the arachnoid layer, prevents spinal cord tethering, and improves spinal cord function after in utero SB repair.

Functional MRI Correlates of Carbon Dioxide Chemosensing in Persons With Epilepsy.

Objectives: Sudden unexpected death in epilepsy (SUDEP) is a catastrophic epilepsy outcome for which there are no reliable premortem imaging biomarkers of risk. Percival respiratory depression is seen in monitored SUDEP and near SUDEP cases, and abnormal chemosensing of raised blood carbon dioxide (CO2) is thought to contribute. Damage to brainstem respiratory control and chemosensing structures has been demonstrated in structural imaging and neuropathological studies of SUDEP. We hypothesized that functional MRI (fMRI) correlates of abnormal chemosensing are detectable in brainstems of persons with epilepsy (PWE) and are different from healthy controls (HC). Methods: We analyzed fMRI BOLD activation and brain connectivity in 10 PWE and 10 age- and sex-matched HCs during precisely metered iso-oxic, hypercapnic breathing challenges. Segmented brainstem responses were of particular interest, along with characterization of functional connectivity metrics between these structures. Regional BOLD activations during hypercapnic challenges were convolved with hemodynamic responses, and the resulting activation maps were passed on to group-level analyses. For the functional connectivity analysis, significant clusters from BOLD results were used as seeds. Each individual seed time-series activation map was extracted for bivariate correlation coefficient analyses to study changes in brain connectivity between PWE and HCs. Results: (1) Greater brainstem BOLD activations in PWE were observed compared to HC during hypercapnic challenges in several structures with respiratory/chemosensing properties. Group comparison between PWE vs. HC showed significantly greater activation in the dorsal raphe among PWE (p < 0.05) compared to HCs. (2) PWE had significantly greater seed-seed connectivity and recruited more structures during hypercapnia compared to HC. Significance: The results of this study show that BOLD responses to hypercapnia in human brainstem are detectable and different in PWE compared to HC. Increased dorsal raphe BOLD activation in PWE and increased seed-seed connectivity between brainstem and adjacent subcortical areas may indicate abnormal chemosensing in these individuals. Imaging investigation of brainstem respiratory centers involved in respiratory regulation in PWE is an important step toward identifying suspected dysfunction of brainstem breathing control that culminates in SUDEP and deserve further study as potential imaging SUDEP biomarkers.

Allografts for Skin Closure during In Utero Spina Bifida Repair in a Sheep Model.

OBJECTIVES: Use of off-label tissue graft materials, such as acellular dermal matrix (ADM), for in utero repair of severe spina bifida (SB), where primary skin layer closure is not possible, is associated with poor neurological outcomes. The cryopreserved human umbilical cord (HUC) patch has regenerative, anti-inflammatory, and anti-scarring properties, and provides watertight SB repair. We tested the hypothesis that the HUC is a superior skin patch to ADM for reducing inflammation at the repair site and preserving spinal cord function. METHODS: In timed-pregnant ewes with twins, on gestational day (GD) 75, spina bifida was created without a myelotomy (functional model). On GD 95, repair was performed using HUC vs. ADM patches (randomly assigned) by suturing them to the skin edges. Additionally, full thickness skin closure as a primary skin closure (PSC) served as a positive control. Delivery was performed on GD 140, followed by blinded to treatment neurological assessments of the lambs using the Texas Spinal Cord Injury Scale (TSCIS) for gait, proprioception, and nociception. Lambs without spina bifida were used as controls (CTL). Ex vivo magnetic resonance imaging of spines at the repair site were performed, followed by quantitative pathological assessments. Histological assessments (blinded) included Masson's trichrome, and immunofluorescence for myeloperoxidase (MPO; neutrophils) and for reactive astrocytes (inflammation) by co-staining vimentin and GFAP. RESULTS: The combined hind limbs' TSCIS was significantly higher in the HUC group than in ADM and PSC groups, p = 0.007. Both ADM and PSC groups exhibited loss of proprioception and mild to moderate ataxia compared to controls. MRI showed increased pathological findings in the PSC group when compared to the HUC group, p = 0.045. Histologically, the meningeal layer was thickened (inflammation) by 2-3 fold in ADM and PSC groups when compared to HUC and CTL groups, p = 0.01. There was lower MPO positive cells in the HUC group than in the ADM group, p = 0.018. Posterior column astrocyte activation was increased in ADM and PSC lambs compared to HUC lambs, p = 0.03. CONCLUSION: The HUC as a skin patch for in utero spina bifida repair preserves spinal cord function by reducing underlying inflammation when compared to ADM.

Targeting white matter neuroprotection as a relapse prevention strategy for treatment of cocaine use disorder: Design of a mechanism-focused randomized clinical trial.

Cocaine use continues to be a significant public health problem with limited treatment options and no approved pharmacotherapies. Cognitive-behavioral therapy (CBT) remains the mainstay treatment for preventing relapse, however, people with chronic cocaine use display cognitive impairments that are associated with poor response to CBT. Emerging evidence in animal and human studies suggests that the peroxisome proliferator-activated receptor-gamma (PPAR- γ) agonist, pioglitazone, improves white matter integrity that is essential for cognitive function. This project will determine whether adjunctive use of pioglitazone enhances the effect of CBT in preventing relapse during the early phase of recovery from cocaine use disorder. This paper describes the design of a mechanism-focused phase 2 randomized clinical trial that aims first to evaluate the effects of pioglitazone on targeted mechanisms related to white matter integrity, cognitive function, and cocaine craving; and second, to evaluate the extent to which improvements on target mechanisms predict CBT response. Positive results will support pioglitazone as a potential cognitive enhancing agent to advance to later stage medication development research.

In vitro and in vivo comparative performance studies of gadolinium-loaded zeolites and Gd-DOTA as contrast agents for MRI applications.

Gadolinium-based contrast agents (CAs) were synthesized using faujasite zeolite (NaX) and zeolite beta (BEA) and their performances in vitro and in vivo were compared to the widely used commercial CA, gadoteric acid (Gd-DOTA). Magnetic resonance imaging (MRI) relaxometry studies (considering longitudinal [T1 ] and transverse [T2 ] relaxation times) were performed using Gd-DOTA and the zeolitic materials loaded with Gd3+ . The Gd-loaded NaX, which presented large pores and cavities (7.35 and 11.24 Å, respectively), exhibited relaxivity values of around 52 mM-1 s-1 , while BEA, which presented smaller pore and cavity diameters (5.95 and 6.68 Å, respectively) showed lower relaxivity values of ~4.8 mM-1 s-1 . The effect of the Gd-loaded NaX as MRI CA was tested in vivo in Sprague-Dawley rats, employing a 7 T scanner, with comparison to Gd-DOTA MRI angiography. The relaxivity measurements showed that the Gd-loaded NaX (50 mM-1 s-1 ) provided better image contrast than Gd-DOTA (5.1 mM-1 s-1 ). Clearance studies of the CAs using urine and blood showed that both Gd-loaded NaX and Gd-DOTA were eliminated from the body after 2 days, demonstrating the potential of Gd-loaded NaX for use as an MRI CA.

Deep learning segmentation of gadolinium-enhancing lesions in multiple sclerosis.

OBJECTIVE: The aim of this study is to assess the performance of deep learning convolutional neural networks (CNNs) in segmenting gadolinium-enhancing lesions using a large cohort of multiple sclerosis (MS) patients. METHODS: A three-dimensional (3D) CNN model was trained for segmentation of gadolinium-enhancing lesions using multispectral magnetic resonance imaging data (MRI) from 1006 relapsing-remitting MS patients. The network performance was evaluated for three combinations of multispectral MRI used as input: (U5) fluid-attenuated inversion recovery (FLAIR), T2-weighted, proton density-weighted, and pre- and post-contrast T1-weighted images; (U2) pre- and post-contrast T1-weighted images; and (U1) only post-contrast T1-weighted images. Segmentation performance was evaluated using the Dice similarity coefficient (DSC) and lesion-wise true-positive (TPR) and false-positive (FPR) rates. Performance was also evaluated as a function of enhancing lesion volume. RESULTS: The DSC/TPR/FPR values averaged over all the enhancing lesion sizes were 0.77/0.90/0.23 using the U5 model. These values for the largest enhancement volumes (>500 mm3) were 0.81/0.97/0.04. For U2, the average DSC/TPR/FPR values were 0.72/0.86/0.31. Comparable performance was observed with U1. For all types of input, the network performance degraded with decreased enhancement size. CONCLUSION: Excellent segmentation of enhancing lesions was observed for enhancement volume ⩾70 mm3. The best performance was achieved when the input included all five multispectral image sets.

A meta-analysis of tract-based spatial statistics studies examining white matter integrity in cocaine use disorder.

Tract-based spatial statistics (TBSS) of diffusion tensor imaging (DTI) studies have consistently shown diminished white matter (WM) integrity for individuals with cocaine use disorder (CUD). The present study used seed-based d mapping (SDM) to determine the extent to which a systematic difference in the WM integrity of cocaine users may exist (as compared with that of healthy controls). Articles from 2006 (when TBSS was first developed) to present were reviewed, with eight selected for inclusion. Meta-analysis found lower fractional anisotropy (FA) in the genu of the corpus callosum for cocaine users, with a small-to-moderate peak effect size (Hedge's g = -0.331). Sensitivity analyses mostly supported the robustness of the obtained difference. Differences detected at exploratory thresholds for significance suggested insult to WM integrity extending beyond the corpus callosum. The present results compliment a previous region-of-interest (ROI)-based meta-analysis of DTI studies in individuals with CUD. These findings have significant implications for the potential role of neuroprotective agents in the treatment of CUD and merit additional iteration as more studies accrue in the literature.

Quantitative Assessment of In Vivo Human Anterior Cruciate Ligament Autograft Remodeling: A 3-Dimensional UTE-T2* Imaging Study.

BACKGROUND: The timing of return to play after anterior cruciate ligament (ACL) reconstruction is still controversial due to uncertainty of true ACL graft state at the time of RTP. Recent work utilizing ultra-short echo T2* (UTE-T2*) magnetic resonance imaging (MRI) as a scanner-independent method to objectively and non-invasively assess the status of in vivo ACL graft remodeling has produced promising results. PURPOSE/HYPOTHESIS: The purpose of this study was to prospectively and noninvasively investigate longitudinal changes in T2* within ACL autografts at incremental time points up to 12 months after primary ACL reconstruction in human patients. We hypothesized that (1) T2* would increase from baseline and initially exceed that of the intact contralateral ACL, followed by a gradual decline as the graft undergoes remodeling, and (2) remodeling would occur in a region-dependent manner. STUDY DESIGN: Case series; Level of evidence, 4. METHODS: Twelve patients (age range, 14-45 years) who underwent primary ACL reconstruction with semitendinosus tendon or bone-patellar tendon-bone autograft (with or without meniscal repair) were enrolled. Patients with a history of previous injury or surgery to either knee were excluded. Patients returned for UTE MRI at 1, 3, 6, 9, and 12 months after ACL reconstruction. Imaging at 1 month included the contralateral knee. MRI pulse sequences included high-resolution 3-dimensional gradient echo sequence and a 4-echo T2-UTE sequence (slice thickness, 1 mm; repetition time, 20 ms; echo time, 0.3, 3.3, 6.3, and 9.3 ms). All slices containing the intra-articular ACL were segmented from high-resolution sequences to generate volumetric regions of interest (ROIs). ROIs were divided into proximal/distal and core/peripheral sub-ROIs using standardized methods, followed by voxel-to-voxel registration to generate T2* maps at each time point. This process was repeated by a second reviewer for interobserver reliability. Statistical differences in mean T2* values and mean ratios of T2*inj/T2*intact (ie, injured knee to intact knee) among the ROIs and sub-ROIs were assessed using repeated measures and one-way analyses of variance. P < .05 represented statistical significance. RESULTS: Twelve patients enrolled in this prospective study, 2 withdrew, and ultimately 10 patients were included in the analysis (n = 7, semitendinosus tendon; n = 3, bone-patellar tendon-bone). Interobserver reliability for T2* values was good to excellent (intraclass correlation coefficient, 0.84; 95% CI, 0.59-0.94; P < .001). T2* values increased from 5.5 ± 2.1 ms (mean ± SD) at 1 month to 10.0 ± 2.9 ms at 6 months (P = .001), followed by a decline to 8.1 ± 2.0 ms at 12 months (P = .129, vs 1 month; P = .094, vs 6 months). Similarly, mean T2*inj/T2*intact ratios increased from 62.8% ± 22.9% at 1 month to 111.1% ± 23.9% at 6 months (P = .001), followed by a decline to 92.8% ± 29.8% at 12 months (P = .110, vs 1 month; P = .086, vs 6 months). Sub-ROIs exhibited similar increases in T2* until reaching a peak at 6 months, followed by a gradual decline until the 12-month time point. There were no statistically significant differences among the sub-ROIs (P > .05). CONCLUSION: In this preliminary study, T2* values for ACL autografts exhibited a statistically significant increase of 82% between 1 and 6 months, followed by an approximate 19% decline in T2* values between 6 and 12 months. In the future, UTE-T2* MRI may provide unique insights into the condition of remodeling ACL grafts and may improve our ability to noninvasively assess graft maturity before return to play.

Brain and lesion segmentation in multiple sclerosis using fully convolutional neural networks: A large-scale study.

OBJECTIVE: To investigate the performance of deep learning (DL) based on fully convolutional neural network (FCNN) in segmenting brain tissues in a large cohort of multiple sclerosis (MS) patients. METHODS: We developed a FCNN model to segment brain tissues, including T2-hyperintense MS lesions. The training, validation, and testing of FCNN were based on ~1000 magnetic resonance imaging (MRI) datasets acquired on relapsing-remitting MS patients, as a part of a phase 3 randomized clinical trial. Multimodal MRI data (dual-echo, FLAIR, and T1-weighted images) served as input to the network. Expert validated segmentation was used as the target for training the FCNN. We cross-validated our results using the leave-one-center-out approach. RESULTS: We observed a high average (95% confidence limits) Dice similarity coefficient for all the segmented tissues: 0.95 (0.92-0.98) for white matter, 0.96 (0.93-0.98) for gray matter, 0.99 (0.98-0.99) for cerebrospinal fluid, and 0.82 (0.63-1.0) for T2 lesions. High correlations between the DL segmented tissue volumes and ground truth were observed (R2 > 0.92 for all tissues). The cross validation showed consistent results across the centers for all tissues. CONCLUSION: The results from this large-scale study suggest that deep FCNN can automatically segment MS brain tissues, including lesions, with high accuracy.

Deep Learning for Predicting Enhancing Lesions in Multiple Sclerosis from Noncontrast MRI.

Background Enhancing lesions on MRI scans obtained after contrast material administration are commonly thought to represent disease activity in multiple sclerosis (MS); it is desirable to develop methods that can predict enhancing lesions without the use of contrast material. Purpose To evaluate whether deep learning can predict enhancing lesions on MRI scans obtained without the use of contrast material. Materials and Methods This study involved prospective analysis of existing MRI data. A convolutional neural network was used for classification of enhancing lesions on unenhanced MRI scans. This classification was performed for each slice, and the slice scores were combined by using a fully connected network to produce participant-wise predictions. The network input consisted of 1970 multiparametric MRI scans from 1008 patients recruited from 2005 to 2009. Enhanced lesions on postcontrast T1-weighted images served as the ground truth. The network performance was assessed by using fivefold cross-validation. Statistical analysis of the network performance included calculation of lesion detection rates and areas under the receiver operating characteristic curve (AUCs). Results MRI scans from 1008 participants (mean age, 37.7 years ± 9.7; 730 women) were analyzed. At least one enhancing lesion was observed in 519 participants. The sensitivity and specificity averaged across the five test sets were 78% ± 4.3 and 73% ± 2.7, respectively, for slice-wise prediction. The corresponding participant-wise values were 72% ± 9.0 and 70% ± 6.3. The diagnostic performances (AUCs) were 0.82 ± 0.02 and 0.75 ± 0.03 for slice-wise and participant-wise enhancement prediction, respectively. Conclusion Deep learning used with conventional MRI identified enhanced lesions in multiple sclerosis from images from unenhanced multiparametric MRI with moderate to high accuracy. © RSNA, 2019.

Deep-Learning-Based Neural Tissue Segmentation of MRI in Multiple Sclerosis: Effect of Training Set Size.

BACKGROUND: The dependence of deep-learning (DL)-based segmentation accuracy of brain MRI on the training size is not known. PURPOSE: To determine the required training size for a desired accuracy in brain MRI segmentation in multiple sclerosis (MS) using DL. STUDY TYPE: Retrospective analysis of MRI data acquired as part of a multicenter clinical trial. STUDY POPULATION: In all, 1008 patients with clinically definite MS. FIELD STRENGTH/SEQUENCE: MRIs were acquired at 1.5T and 3T scanners manufactured by GE, Philips, and Siemens with dual turbo spin echo, FLAIR, and T1 -weighted turbo spin echo sequences. ASSESSMENT: Segmentation results using an automated analysis pipeline and validated by two neuroimaging experts served as the ground truth. A DL model, based on a fully convolutional neural network, was trained separately using 16 different training sizes. The segmentation accuracy as a function of the training size was determined. These data were fitted to the learning curve for estimating the required training size for desired accuracy. STATISTICAL TESTS: The performance of the network was evaluated by calculating the Dice similarity coefficient (DSC), and lesion true-positive and false-positive rates. RESULTS: The DSC for lesions showed much stronger dependency on the sample size than gray matter (GM), white matter (WM), and cerebrospinal fluid (CSF). When the training size was increased from 10 to 800 the DSC values varied from 0.00 to 0.86 ± 0.016 for T2 lesions, 0.87 ± 009 to 0.94 ± 0.004 for GM, 0.86 ± 0.08 to 0.94 ± 0.005 for WM, and 0.91 ± 0.009 to 0.96 ± 0.003 for CSF. DATA CONCLUSION: Excellent segmentation was achieved with a training size as small as 10 image volumes for GM, WM, and CSF. In contrast, a training size of at least 50 image volumes was necessary for adequate lesion segmentation. LEVEL OF EVIDENCE: 1 Technical Efficacy Stage: 1 J. Magn. Reson. Imaging 2020;51:1487-1496.

Are multi-contrast magnetic resonance images necessary for segmenting multiple sclerosis brains? A large cohort study based on deep learning.

BACKGROUND: Magnetic resonance images with multiple contrasts or sequences are commonly used for segmenting brain tissues, including lesions, in multiple sclerosis (MS). However, acquisition of images with multiple contrasts increases the scan time and complexity of the analysis, possibly introducing factors that could compromise segmentation quality. OBJECTIVE: To investigate the effect of various combinations of multi-contrast images as input on the segmented volumes of gray (GM) and white matter (WM), cerebrospinal fluid (CSF), and lesions using a deep neural network. METHODS: U-net, a fully convolutional neural network was used to automatically segment GM, WM, CSF, and lesions in 1000 MS patients. The input to the network consisted of 15 combinations of FLAIR, T1-, T2-, and proton density-weighted images. The Dice similarity coefficient (DSC) was evaluated to assess the segmentation performance. For lesions, true positive rate (TPR) and false positive rate (FPR) were also evaluated. In addition, the effect of lesion size on lesion segmentation was investigated. RESULTS: Highest DSC was observed for all the tissue volumes, including lesions, when the input was combination of all four image contrasts. All other input combinations that included FLAIR also provided high DSC for all tissue classes. However, the quality of lesion segmentation showed strong dependence on the input images. The DSC and TPR values for inputs with the four contrast combination and FLAIR alone were very similar, but FLAIR showed a moderately higher FPR for lesion size <100 µl. For lesions smaller than 20 µl all image combinations resulted in poor performance. The segmentation quality improved with lesion size. CONCLUSIONS: Best performance for segmented tissue volumes was obtained with all four image contrasts as the input, and comparable performance was attainable with FLAIR only as the input, albeit with a moderate increase in FPR for small lesions. This implies that acquisition of only FLAIR images provides satisfactory tissue segmentation. Lesion segmentation was poor for very small lesions and improved rapidly with lesion size.

Automated Segmentation of Tissues Using CT and MRI: A Systematic Review.

RATIONALE AND OBJECTIVES: The automated segmentation of organs and tissues throughout the body using computed tomography and magnetic resonance imaging has been rapidly increasing. Research into many medical conditions has benefited greatly from these approaches by allowing the development of more rapid and reproducible quantitative imaging markers. These markers have been used to help diagnose disease, determine prognosis, select patients for therapy, and follow responses to therapy. Because some of these tools are now transitioning from research environments to clinical practice, it is important for radiologists to become familiar with various methods used for automated segmentation. MATERIALS AND METHODS: The Radiology Research Alliance of the Association of University Radiologists convened an Automated Segmentation Task Force to conduct a systematic review of the peer-reviewed literature on this topic. RESULTS: The systematic review presented here includes 408 studies and discusses various approaches to automated segmentation using computed tomography and magnetic resonance imaging for neurologic, thoracic, abdominal, musculoskeletal, and breast imaging applications. CONCLUSION: These insights should help prepare radiologists to better evaluate automated segmentation tools and apply them not only to research, but eventually to clinical practice.

Regional differences in white matter integrity in stimulant use disorders: A meta-analysis of diffusion tensor imaging studies.

BACKGROUND: Converging lines of evidence from diffusion tensor imaging (DTI) studies reveal significant alterations in white matter (WM) microstructure in the prefrontal cortex of chronic stimulant users compared to controls, suggesting compromised axonal microstructure and/or myelin. METHODS: A meta-analysis of DTI-based WM integrity was conducted for white matter regions across the corpus callosum and association fibers. Articles were sourced and selected using PRISMA guidelines for systematic review and meta-analysis. Inclusion and exclusion criteria were determined by the authors in order to best capture WM integrity among individuals with primary stimulant use in comparison to healthy control subjects. RESULTS: Eleven studies that focused on region-of-interest (ROI)-based analysis of WM integrity were extracted from an initial pool of 113 independent studies. Analysis across ROIs indicated significantly lower fractional anisotropy (FA) values in stimulant use groups compared to controls with a small to moderate overall effect (Hedges' g = -0.37, 95% CI [-0.54, -0.20]). Eigenvalues were also analyzed, revealing a significant effect for radial diffusivity (RD; Hedges' g = 0.24, 95% CI [0.01, 0.47]) but not axial diffusivity (AD; Hedges' g = 0.05, 95% CI [-0.20, 0.29]) or mean diffusivity (MD; Hedges' g = 0.20, 95% CI [-0.01, 0.41]). Subgroup analyses based on specific ROIs, primary substance use, poly-substance use, and imaging technology were also explored. CONCLUSION: Results of the present study suggest a consistent effect of compromised WM integrity for individuals with stimulant use disorders. Furthermore, no significant differences were found between cocaine and methamphetamine-based groups.

MRI acoustic noise-modulated computer animations for patient distraction and entertainment with application in pediatric psychiatric patients.

PURPOSE: To reduce patient anxiety caused by the MRI scanner acoustic noise. MATERIAL AND METHODS: We developed a simple and low-cost system for patient distraction using visual computer animations that were synchronized to the MRI scanner's acoustic noise during the MRI exam. The system was implemented on a 3T MRI system and tested in 28 pediatric patients with bipolar disorder. The patients were randomized to receive noise-synchronized animations in the form of abstract animations in addition to music (n = 13, F/M = 6/7, age = 10.9 ±â€¯2.5 years) or, as a control, receive only music (n = 15, F/M = 7/8, age = 11.6 ±â€¯2.3 years). After completion of the scans, all subjects answered a questionnaire about their scan experience and the perceived scan duration. RESULTS: The scan duration with multisensory input (animations and music) was perceived to be ~15% shorter than in the control group (43 min vs. 50 min, P < 0.05). However, the overall scan experience was scored less favorably (3.9 vs. 4.6 in the control group, P < 0.04). CONCLUSIONS: This simple system provided patient distraction and entertainment leading to perceived shorter scan times, but the provided visualization with abstract animations was not favored by this patient cohort.

Serial Cerebral Metabolic Changes in Patients With Ischemic Stroke Treated With Autologous Bone Marrow Derived Mononuclear Cells.

Purpose: Cell-based therapy offers new opportunities for the development of novel treatments to promote tissue repair, functional restoration, and cerebral metabolic balance. N-acetylasperate (NAA), Choline (Cho), and Creatine (Cr) are three major metabolites seen on proton magnetic resonance spectroscopy (MRS) that play a vital role in balancing the biochemical processes and are suggested as markers of recovery. In this preliminary study, we serially monitored changes in these metabolites in ischemic stroke patients who were treated with autologous bone marrow-derived mononuclear cells (MNCs) using non-invasive MRS. Materials and Methods: A sub-group of nine patients (3 male, 6 female) participated in a serial MRS study, as part of a clinical trial on autologous bone marrow cell therapy in acute ischemic stroke. Seven to ten million mononuclear cells were isolated from the patient's bone marrow and administered intravenously within 72 h of onset of injury. MRS data were obtained at 1, 3, and 6 months using a whole-body 3.0T MRI. Single voxel point-resolved spectroscopy (PRESS) was obtained within the lesion and contralesional gray matter. Spectral analysis was done using TARQUIN software and absolute concentration of NAA, Cho, and Cr was determined. National Institute of Health Stroke Scale (NIHSS) was serially recoreded. Two-way analysis of variance was performed and p < 0.05 considered statistically significant. Results: All metabolites showed statistically significant or clear trends toward lower ipsilesional concentrations compared to the contralesional side at all time points. Statistically significant reductions were found in ipsilesional NAA at 1M and 3M, Cho at 6M, and Cr at 1M and 6M (p < 0.03), compared to the contralesional side. Temporally, ipsilesional NAA increased between 3M and 6M (p < 0.01). On the other hand, ipsilesional Cho showed continued decline till 6M (p < 0.01). Ipsilesional Cr was stable over time. Contralesional metabolites were relatively stable over time, with only Cr showing a reduction 3M (p < 0.02). There was a significant (p < 0.03) correlation between ipsilesional NAA and NIHSS at 3M follow-up. Conclusion: Serial changes in metabolites suggest that MRS can be applied to monitor therapeutic changes. Post-treatment increasing trends of NAA concentration and significant correlation with NIHSS support a potential therapeutic effect.

Automated image quality evaluation of structural brain MRI using an ensemble of deep learning networks.

BACKGROUND: Deep learning (DL) is a promising methodology for automatic detection of abnormalities in brain MRI. PURPOSE: To automatically evaluate the quality of multicenter structural brain MRI images using an ensemble DL model based on deep convolutional neural networks (DCNNs). STUDY TYPE: Retrospective. POPULATION: The study included 1064 brain images of autism patients and healthy controls from the Autism Brain Imaging Data Exchange (ABIDE) database. MRI data from 110 multiple sclerosis patients from the CombiRx study were included for independent testing. SEQUENCE: T1 -weighted MR brain images acquired at 3T. ASSESSMENT: The ABIDE data were separated into training (60%), validation (20%), and testing (20%) sets. The ensemble DL model combined the results from three cascaded networks trained separately on the three MRI image planes (axial, coronal, and sagittal). Each cascaded network consists of a DCNN followed by a fully connected network. The quality of image slices from each plane was evaluated by the DCNN and the resultant image scores were combined into a volumewise quality rating using the fully connected network. The DL predicted ratings were compared with manual quality evaluation by two experts. STATISTICAL TESTS: Receiver operating characteristic (ROC) curve, area under ROC curve (AUC), sensitivity, specificity, accuracy, and positive (PPV) and negative (NPV) predictive values. RESULTS: The AUC, sensitivity, specificity, accuracy, PPV, and NPV for image quality evaluation of the ABIDE test set using the ensemble model were 0.90, 0.77, 0.85, 0.84, 0.42, and 0.96, respectively. On the CombiRx set the same model achieved performance of 0.71, 0.41, 0.84, 0.73, 0.48, and 0.80. DATA CONCLUSION: This study demonstrated the high accuracy of DL in evaluating image quality of structural brain MRI in multicenter studies. LEVEL OF EVIDENCE: 3 Technical Efficacy: Stage 2 J. Magn. Reson. Imaging 2019;50:1260-1267.