Search

Brain disorders comprise several psychiatric and neurological disorders which can be characterized by impaired cognition, mood alteration, psychosis, depressive episodes, and neurodegeneration. Clinical diagnoses primarily rely on a combination of life history information and questionnaires, with a distinct lack of discriminative biomarkers in use for psychiatric disorders. Symptoms across brain conditions are associated with functional alterations of cognitive and emotional processes, which can correlate with anatomical variation; structural magnetic resonance imaging (MRI) data of the brain are therefore an important focus of research, particularly for predictive modelling. With the advent of large MRI data consortia (such as the Alzheimer's Disease Neuroimaging Initiative) facilitating a greater number of MRI‐based classification studies, convolutional neural networks (CNNs)—deep learning models well suited to image processing tasks—have become increasingly popular for research into brain conditions. This has resulted in a myriad of studies reporting impressive predictive performances, demonstrating the potential clinical value of deep learning systems. However, methodologies can vary widely across studies, making them difficult to compare and/or reproduce, potentially limiting their clinical application. Here, we conduct a qualitative systematic literature review of 55 studies carrying out CNN‐based predictive modelling of brain disorders using MRI data and evaluate them based on three principles—modelling practices, transparency, and interpretability. We propose several recommendations to enhance the potential for the integration of CNNs into clinical care. [ABSTRACT FROM AUTHOR]

Aims: The aim of this study was to determine microvascular function in the acute phase of Takotsubo syndrome (TTS) and to identify inflammatory mediators that could reflect TTS‐induced pathology. Methods and results: The study included 20 females [median age 65 years; interquarile range (IQR) = 58–70 years] with TTS according to the Mayo diagnostic criteria. During heart catheterization, we determined the index of microvascular resistance (IMR) and drew blood samples almost simultaneously from the aorta and coronary sinus. Cardiac magnetic resonance imaging (MRI) was done in the acute phase. We present descriptive coronary physiology and cardiac MRI data and compare inflammatory biomarkers between samples from the aorta, coronary sinus, and venous samples after 3 months using the Wilcoxon signed‐rank test. For comparison, we also analysed the actual biomarkers in venous blood from 15 healthy female controls. A supplementary analysis explored Spearman's rank correlation between the inflammatory biomarkers, IMR, MRI data, and cardiac biomarkers. The median IMR was 16.5 mmHg·s (IQR = 10.5–28.2 mmHg·s), which was only slightly higher than that in the reference populations. Seven (35%) of the study subjects had IMR > 25 mmHg·s, suggesting a microvascular dysfunction. IMR was not affected by time from symptom onset. According to MRI, the apical region of the left ventricle was affected in 65% of the subjects. The median ejection fraction was 41% (IQR = 31–48%). Biomarker analyses revealed elevation of markers for extracellular matrix remodelling and fibrosis, inflammation, immune activation, and upstream inflammation as compared with healthy controls. Only the levels of interleukin (IL)‐1 receptor antagonist and soluble T‐cell immunoglobulin mucin domain‐3 (sTIM‐3) were higher in the coronary sinus than in the aorta. No variable was significantly correlated with IMR. The IL‐6 level in the aorta was inversely correlated with the left ventricular ejection fraction. Growth differentiation factor‐15, osteoprotegerin, and von Willebrand factor levels in both aorta and coronary sinus were positively correlated with N‐terminal‐pro‐brain‐natriuretic peptide, while the correlations of IL‐6 and sTIM‐3 with N‐terminal‐pro‐brain‐natriuretic peptide were restricted to the aorta and coronary sinus, respectively. While most of the markers were within normal limits after 3 months, matrix metalloproteinase‐9 increased during follow‐up to reach levels higher than those in the healthy controls. Conclusion: The median IMR was only slightly elevated in this study, but about one‐third of the patients had values indicating microvascular dysfunction. The present study supports the involvement of several inflammatory pathways in TTS, including monocyte/macrophage activation, with sTIM‐3 as a potential novel marker. [ABSTRACT FROM AUTHOR]

Assessing axonal morphology in vivo opens new avenues for the combined study of brain structure and function. A novel approach has recently been introduced to estimate the morphology of axonal fibers from the combination of magnetic resonance imaging (MRI) data and electroencephalography (EEG) measures of the interhemispheric transfer time (IHTT). In the original study, the IHTT measures were computed from EEG data averaged across a group, leading to bias of the axonal morphology estimates. Here, we seek to estimate axonal morphology from individual measures of IHTT, obtained from EEG data acquired in a visual evoked potential experiment. Subject‐specific IHTTs are computed in a data‐driven framework with minimal a priori constraints, based on the maximal peak of neural responses to visual stimuli within periods of statistically significant evoked activity in the inverse solution space. The subject‐specific IHTT estimates ranged from 8 to 29 ms except for one participant and the between‐session variability was comparable to between‐subject variability. The mean radius of the axonal radius distribution, computed from the IHTT estimates and the MRI data, ranged from 0 to 1.09 μm across subjects. The change in axonal g‐ratio with axonal radius ranged from 0.62 to 0.81 μm−α. The single‐subject measurement of the IHTT yields estimates of axonal morphology that are consistent with histological values. However, improvement of the repeatability of the IHTT estimates is required to improve the specificity of the single‐subject axonal morphology estimates. [ABSTRACT FROM AUTHOR]

Purpose: To develop a convolutional neural network (CNN) for the robust and fast correction of velocity aliasing in 4D‐flow MRI. Methods: This study included 667 adult subjects with aortic 4D‐flow MRI data with existing velocity aliasing (n = 362) and no velocity aliasing (n = 305). Additionally, 10 controls received back‐to‐back 4D‐flow scans with systemically varied velocity‐encoding sensitivity (vencs) at 60, 100, and 175 cm/s. The no‐aliasing data sets were used to simulate velocity aliasing by reducing the venc to 40%–70% of the original, alongside a ground truth locating all aliased voxels (153 training, 152 testing). The 152 simulated and 362 existing aliasing data sets were used for testing and compared with a conventional velocity antialiasing algorithm. Dice scores were calculated to quantify CNN performance. For controls, the venc 175‐cm/s scans were used as the ground truth and compared with the CNN‐corrected venc 60 and 100 cm/s data sets Results: The CNN required 176 ± 30 s to perform compared with 162 ± 14 s for the conventional algorithm. The CNN showed excellent performance for the simulated data compared with the conventional algorithm (median range of Dice scores CNN: [0.89–0.99], conventional algorithm: [0.84–0.94], p < 0.001, across all simulated vencs) and detected more aliased voxels in existing velocity aliasing data sets (median detected CNN: 159 voxels [31–605], conventional algorithm: 65 [7–417], p < 0.001). For controls, the CNN showed Dice scores of 0.98 [0.95–0.99] and 0.96 [0.87–0.99] for venc = 60 cm/s and 100 cm/s, respectively, while flow comparisons showed moderate‐excellent agreement. Conclusion: Deep learning enabled fast and robust velocity anti‐aliasing in 4D‐flow MRI. [ABSTRACT FROM AUTHOR]

Background: Investigation of the factors influencing dilated cardiomyopathy (DCM) prognosis is important as it could facilitate risk stratification and guide clinical decision‐making. Purpose: To assess the prognostic value of magnetic resonance imaging (MRI) radiomics analysis of native T1 mapping in DCM. Study Type: Prospective. Subjects: Three hundred and thirty consecutive patients with non‐ischemic DCM (mean age 48.42 ± 14.20 years, 247 males). Field Strength/Sequence: Balanced steady‐state free precession and modified Look‐Locker inversion recovery T1 mapping sequences at 3 T. Assessment: Clinical characteristics, conventional MRI parameters (ventricular volumes, function, and mass), native myocardial T1, and radiomics features extracted from native T1 mapping were obtained. The study endpoint was defined as all‐cause mortality or heart transplantation. Models were developed based on 1) clinical data; 2) radiomics data based on T1 mapping; 3) clinical and conventional MRI data; 4) clinical, conventional MRI, and native T1 data; and 5) clinical, conventional MRI, and radiomics T1 mapping data. Each prediction model was trained according to follow‐up results with AdaBoost, random forest, and logistic regression classifiers. Statistical Tests: The predictive performance was evaluated using the area under the receiver operating characteristic curve (AUC) and F1 score by 5‐fold cross‐validation. Results: During a median follow‐up of 53.5 months (interquartile range, 41.6–69.5 months), 77 patients with DCM experienced all‐cause mortality or heart transplantation. The random forest model based on radiomics combined with clinical and conventional MRI parameters achieved the best performance, with AUC and F1 score of 0.95 and 0.89, respectively. Data Conclusion: A machine‐learning framework based on radiomics analysis of T1 mapping prognosis prediction in DCM. Level of Evidence: 1 Technical Efficacy: Stage 2 [ABSTRACT FROM AUTHOR]

Background: Magnetic resonance fingerprinting (MRF) is a method to speed up acquisition of quantitative MRI data. However, MRF does not usually produce contrast‐weighted images that are required by radiologists, limiting reachable total scan time improvement. Contrast synthesis from MRF could significantly decrease the imaging time. Purpose: To improve clinical utility of MRF by synthesizing contrast‐weighted MR images from the quantitative data provided by MRF, using U‐nets that were trained for the synthesis task utilizing L1‐ and perceptual loss functions, and their combinations. Study Type: Retrospective. Population: Knee joint MRI data from 184 subjects from Northern Finland 1986 Birth Cohort (ages 33–35, gender distribution not available). Field Strength and Sequence: A 3 T, multislice‐MRF, proton density (PD)‐weighted 3D‐SPACE (sampling perfection with application optimized contrasts using different flip angle evolution), fat‐saturated T2‐weighted 3D‐space, water‐excited double echo steady state (DESS). Assessment: Data were divided into training, validation, test, and radiologist's assessment sets in the following way: 136 subjects to training, 3 for validation, 3 for testing, and 42 for radiologist's assessment. The synthetic and target images were evaluated using 5‐point Likert scale by two musculoskeletal radiologists blinded and with quantitative error metrics. Statistical Tests: Friedman's test accompanied with post hoc Wilcoxon signed‐rank test and intraclass correlation coefficient. The statistical cutoff P <0.05 adjusted by Bonferroni correction as necessary was utilized. Results: The networks trained in the study could synthesize conventional images with high image quality (Likert scores 3–4 on a 5‐point scale). Qualitatively, the best synthetic images were produced with combination of L1‐ and perceptual loss functions and perceptual loss alone, while L1‐loss alone led to significantly poorer image quality (Likert scores below 3). The interreader and intrareader agreement were high (0.80 and 0.92, respectively) and significant. However, quantitative image quality metrics indicated best performance for the pure L1‐loss. Data Conclusion: Synthesizing high‐quality contrast‐weighted images from MRF data using deep learning is feasible. However, more studies are needed to validate the diagnostic accuracy of these synthetic images. Evidence Level: 4. Technical Efficacy: Stage 1. [ABSTRACT FROM AUTHOR]

The logopenic variant of primary progressive aphasia (lvPPA) is a neurodegenerative syndrome characterized linguistically by gradual loss of repetition and naming skills resulting from left posterior temporal and inferior parietal atrophy. Here, we sought to identify which specific cortical loci are initially targeted by the disease (epicenters) and investigate whether atrophy spreads through predetermined networks. First, we used cross‐sectional structural MRI data from individuals with lvPPA to define putative disease epicenters using a surface‐based approach paired with an anatomically fine‐grained parcellation of the cortical surface (i.e., HCP‐MMP1.0 atlas). Second, we combined cross‐sectional functional MRI data from healthy controls and longitudinal structural MRI data from individuals with lvPPA to derive the epicenter‐seeded resting‐state networks most relevant to lvPPA symptomatology and ascertain whether functional connectivity in these networks predicts longitudinal atrophy spread in lvPPA. Our results show that two partially distinct brain networks anchored to the left anterior angular and posterior superior temporal gyri epicenters were preferentially associated with sentence repetition and naming skills in lvPPA. Critically, the strength of connectivity within these two networks in the neurologically‐intact brain significantly predicted longitudinal atrophy progression in lvPPA. Taken together, our findings indicate that atrophy progression in lvPPA, starting from inferior parietal and temporoparietal junction regions, predominantly follows at least two partially nonoverlapping pathways, which may influence the heterogeneity in clinical presentation and prognosis. [ABSTRACT FROM AUTHOR]

Background: Healthy articular cartilage presents structural gradients defined by distinct zonal patterns through the thickness, which may be disrupted in the pathogenesis of several disorders. Analysis of textural patterns using quantitative MRI data may identify structural gradients of healthy or degenerating tissue that correlate with early osteoarthritis (OA). Purpose: To quantify spatial gradients and patterns in MRI data, and to probe new candidate biomarkers for early severity of OA. Study Type: Retrospective study. Subjects: Fourteen volunteers receiving total knee replacement surgery (eight males/two females/four unknown, average age ± standard deviation: 68.1 ± 9.6 years) and 10 patients from the OA Initiative (OAI) with radiographic OA onset (two males/eight females, average age ± standard deviation: 57.7 ± 9.4 years; initial Kellgren‐Lawrence [KL] grade: 0; final KL grade: 3 over the 10‐year study). Field Strength/Sequence: 3.0‐T and 14.1‐T, biomechanics‐based displacement‐encoded imaging, fast spin echo, multi‐slice multi‐echo T2 mapping. Assessment: We studied structure and strain in cartilage explants from volunteers receiving total knee replacement, or structure in cartilage of OAI patients with progressive OA. We calculated spatial gradients of quantitative MRI measures (eg, T2) normal to the cartilage surface to enhance zonal variations. We compared gradient values against histologically OA severity, conventional relaxometry, and/or KL grades. Statistical Tests: Multiparametric linear regression for evaluation of the relationship between residuals of the mixed effects models and histologically determined OA severity scoring, with a significance threshold at α = 0.05. Results: Gradients of individual relaxometry and biomechanics measures significantly correlated with OA severity, outperforming conventional relaxometry and strain metrics. In human explants, analysis of spatial gradients provided the strongest relationship to OA severity (R2 = 0.627). Spatial gradients of T2 from OAI data identified variations in radiographic (KL Grade 2) OA severity in single subjects, while conventional T2 alone did not. Data Conclusion: Spatial gradients of quantitative MRI data may improve the predictive power of noninvasive imaging for early‐stage degeneration. Evidence Level: 1 Technical Efficacy: Stage 1 [ABSTRACT FROM AUTHOR]

Since magnetic resonance imaging (MRI) of the brain has proved to be the most important paraclinical tool for diagnosing multiple sclerosis (MS) and monitoring its evolution, methodological and statistical issues related to the use of MRI markers in MS have been the focus of several studies in the past 10 years. While many of these methodological issues have been addressed using standard procedures available from other areas of application of medical statistics, in some cases statistical procedures that are not standard have been developed specifically for MRI variables in MS. Two of the major achievements in the statistical methods applied to the use of MRI variables in MS in the past 10 years were the identification of a parametric model to describe the distribution of MRI lesion counts across patients and the study of the relationships between MRI markers and clinical variables, with the aim to validate MRI parameters as surrogates for clinical progression. [ABSTRACT FROM AUTHOR]

Magnetic resonance imaging (MRI) is playing an important role in the classification of breast tumors. MRI can be used to obtain multiparametric (mp) information, such as structural, hemodynamic, and physiological information. Quantitative analysis of mp‐MRI data has shown potential in improving the accuracy of breast tumor classification. In general, a large set of quantitative and texture features can be generated depending upon the type of methodology used. A suitable combination of selected quantitative and texture features can further improve the accuracy of tumor classification. Machine learning (ML) classifiers based upon features derived from MRI data have shown potential in tumor classification. There is a need for further research studies on selecting an appropriate combination of features and evaluating the performance of different ML classifiers for accurate classification of breast tumors. The objective of the current study was to develop and optimize an ML framework based upon mp‐MRI features for the characterization of breast tumors (malignant vs. benign and low‐ vs. high‐grade). This study included the breast mp‐MRI data of 60 female patients with histopathology results. A total of 128 features were extracted from the mp‐MRI tumor data followed by features selection. Five ML classifiers were evaluated for tumor classification using 10‐fold crossvalidation with 10 repetitions. The support vector machine (SVM) classifier based on optimum features selected using a wrapper method with an adaptive boosting (AdaBoost) technique provided the highest sensitivity (0.96 ± 0.03), specificity (0.92 ± 0.09), and accuracy (94% ± 2.91%) in the classification of malignant versus benign tumors. This method also provided the highest sensitivity (0.94 ± 0.07), specificity (0.80 ± 0.05), and accuracy (90% ± 5.48%) in the classification of low‐ versus high‐grade tumors. These findings suggest that the SVM classifier outperformed other ML methods in the binary classification of breast tumors. [ABSTRACT FROM AUTHOR]

Three-dimensional (3D) printing technologies have been increasingly utilized in medicine over the past several years and can greatly facilitate surgical planning thereby improving patient outcomes. Although still much less utilized compared to computed tomography (CT), magnetic resonance imaging (MRI) is gaining traction in medical 3D printing. The purpose of this study was two-fold: 1) to determine the prevalence in the existing literature of using MRI to create 3D printed anatomic models for surgical planning and 2) to provide image acquisition recommendations for appropriate clinical scenarios where MRI is the most suitable imaging modality. The workflow for creating 3D printed anatomic models from medical imaging data is complex and involves image segmentation of the regions of interest and conversion of that data into 3D surface meshes, which are compatible with printing technologies. CT is most commonly used to create 3D printed anatomic models due to the high image quality and relative ease of performing image segmentation from CT data. As compared to CT datasets, 3D printing using MRI data offers advantages since it provides exquisite soft tissue contrast needed for accurate organ segmentation and it does not expose patients to unnecessary ionizing radiation. MRI, however, often requires complicated imaging techniques and time-consuming postprocessing procedures to generate high-resolution 3D anatomic models needed for 3D printing. Despite these challenges, 3D modeling and printing from MRI data holds great clinical promises thanks to emerging innovations in both advanced MRI imaging and postprocessing techniques. EVIDENCE LEVEL: 2 TECHNICAL EFFICATCY: 5. [ABSTRACT FROM AUTHOR]

A scalable multichannel digital MRI receiver system was designed to achieve high bandwidth echo-planar imaging (EPI) acquisitions for applications such as BOLD-fMRI. The modular system design allows for easy extension to an arbitrary number of channels. A 16-channel receiver was developed and integrated with a General Electric (GE) Signa 3T VH/3 clinical scanner. Receiver performance was evaluated on phantoms and human volunteers using a custom-built 16-element receive-only brain surface coil array. At an output bandwidth of 1 MHz, a 100% acquisition duty cycle was achieved. Overall system noise figure and dynamic range were better than 0.85 dB and 84 dB, respectively. During repetitive EPI scanning on phantoms, the relative temporal standard deviation of the image intensity time-course was below 0.2%. As compared to the product birdcage head coil, 16-channel reception with the custom array yielded a nearly 6-fold SNR gain in the cerebral cortex and a 1.8-fold SNR gain in the center of the brain. The excellent system stability combined with the increased sensitivity and SENSE capabilities of 16-channel coils are expected to significantly benefit and enhance fMRI applications. Magn Reson Med 51:165–171, 2004. Published 2003 Wiley-Liss, Inc. [ABSTRACT FROM AUTHOR]

To validate the adaptive semiautomated self-evaluated registration technique (ASSERT) followed by factor analysis of medical image sequence (FAMIS) for analyzing myocardial perfusion using magnetic resonance imaging (MRI) images. Eleven patients having a significant stenosis of at least one coronary artery detected by coronarography were examined by thallium tomoscintigraphy and perfusion MRI (first-pass of Gd-DTPA-BMA) at rest and under pharmacologic stress. The MRI images were analyzed by ASSERT to correct for rigid motion in the acquisition plane and to reject those images that were severely deformed or acquired outside the slice plane. The images thus obtained were analyzed by FAMIS. The resulting factor images representing myocardial perfusion were read to identify the ischemic territories corresponding to left anterior descending coronary arteries and right coronary arteries. ASSERT allowed automatic correction for motion and the exclusion of images that could not be registered. The ischemic territories, identified from the factor images of the myocardium, agreed with those identified by coronarography and tomoscintigraphy for 20 of the 22 territories. The results demonstrate the feasibility of analyzing myocardial perfusion using MRI acquisition and treating the resulting images by ASSERT and FAMIS. Extending this method to multislice examinations will enable evaluation of the perfusion of the whole myocardium. J. Magn. Reson. Imaging 2003;18:681–690. © 2003 Wiley-Liss, Inc. [ABSTRACT FROM AUTHOR]

Detection of activation in functional MRI (fMRI) is often complicated by the low contrast-to-noise ratio (CNR) in the data. The primary source of the difficulty is the fact that for activities that are subtle the signal can be hidden inside the inherent noise in the data. Classical univariate methods based on t-test or F-test are susceptible to noise, as they fail to harness systematic correlations in evoked responses within neighboring voxels. Here the power of a multivariate statistical analysis tool known as canonical correlation analysis (CCA) in fMRI studies is demonstrated where the CNR is low. As a further illustration of the power of the method, a comparative study of CCA and ordinary correlation analysis using simulated data under various noise levels is also performed. A novel nonparametric approach is introduced to calculate the P-values from the distribution of the complicated test statistic. The circumstances under which CCA is a better performer as well as when it is not the case are discussed. As an example, this method is applied to detect hippocampal activation from memory-related tasks. Magn Reson Med 50:354–365, 2003. © 2003 Wiley-Liss, Inc. [ABSTRACT FROM AUTHOR]

Diffusion‐derived vessel density (DVDD) is a physiological surrogate of the area of microvessels per unit tissue area. DDVD is calculated according to DDVD(b0b2) = Sb0/ROIarea0 – Sb2/ROIarea2, where Sb0 and Sb2 refer to the liver signal when b is 0 or 2 s/mm2. Pathohistological studies and contrast‐enhanced CT/MRI data showed higher blood volume in hepatocellular carcinoma (HCC) relative to native liver tissue. With intravoxel incoherent motion (IVIM) imaging, most authors paradoxically reported a decreased perfusion fraction of HCC relative to the adjacent liver. This study applied DDVD to assess the perfusion of HCC. MRI was performed with a 3.0‐T magnet. Diffusion‐weighted images with b‐values of 0 and 2 s/mm2 were acquired in 72 HCC patients. Thirty‐two patients had microvascular invasion (MVI(+)) and 40 patients did not have microvascular invasion (MVI(−)). Fifty‐eight patients had Edmondson–Steiner grade I or II HCC, and 14 patients had Edmondson–Steiner grade III or IV HCC. DDVD measurement was conducted on the axial slice that showed the largest HCC size. DDVD(b0b2) T/L = HCC DDVD(b0b2)/liver DDVD(b0b2). DDVD(b0b2) T/L median (95% confidence interval) of all HCCs was 2.942 (2.419–3.522), of MVI(−) HCCs was 2.699 (2.030–3.522), of MVI(+) HCCs was 2.988 (2.423–3.990), of Edmondson–Steiner grade I/II HCCs was 2.873 (2.277–3.465), and of Edmondson–Steiner grade III/IV HCCs was 3.403 (2.008–4.485). DDVD(b0b2) T/L approximately agrees with contrast agent dynamically enhanced CT/MRI literature data, whereas it differs from earlier IVIM study results, where HCC perfusion fraction was paradoxically lower relative to native liver tissue. A weak trend was noted with MIV(+) HCCs had a higher DDVD(b0b2) T/L than that of MVI(−) HCCs, and a weak trend was noted with the poorly differentiated group of HCCs (Edmondson–Steiner grade III and IV) had a higher DDVD(b0b2) T/L than that of the better differentiated group of HCCs (Edmondson–Steiner grade I and II). [ABSTRACT FROM AUTHOR]

Purpose: This study aims to evaluate two distinct approaches for fiber radius estimation using diffusion-relaxation MRI data acquired in biomimetic microfiber phantoms that mimic hollow axons. The methods considered are the spherical mean power-law approach and a T2-based pore size estimation technique. Theory and Methods: A general diffusion-relaxation theoretical model for the spherical mean signal from water molecules within a distribution of cylinders with varying radii was introduced, encompassing the evaluated models as particular cases. Additionally, a new numerical approach was presented for estimating effective radii (i.e., MRI-visible mean radii) from the ground truth radii distributions, not reliant on previous theoretical approximations and adaptable to various acquisition sequences. The ground truth radii were obtained from scanning electron microscope images. Results: Both methods show a linear relationship between effective radii estimated from MRI data and ground-truth radii distributions, although some discrepancies were observed. The spherical mean power-law method overestimated fiber radii. Conversely, the T2-based method exhibited higher sensitivity to smaller fiber radii, but faced limitations in accurately estimating the radius in one particular phantom, possibly because of material-specific relaxation changes. Conclusion: The study demonstrates the feasibility of both techniques to predict pore sizes of hollow microfibers. The T2-based technique, unlike the spherical mean power-law method, does not demand ultra-high diffusion gradients, but requires calibration with known radius distributions. This research contributes to the ongoing development and evaluation of neuroimaging techniques for fiber radius estimation, highlights the advantages and limitations of both methods, and provides datasets for reproducible research. [ABSTRACT FROM AUTHOR]

Multivariate techniques better fit the anatomy of complex neuropsychiatric disorders which are characterized not by alterations in a single region, but rather by variations across distributed brain networks. Here, we used principal component analysis (PCA) to identify patterns of covariance across brain regions and relate them to clinical and demographic variables in a large generalizable dataset of individuals with bipolar disorders and controls. We then compared performance of PCA and clustering on identical sample to identify which methodology was better in capturing links between brain and clinical measures. Using data from the ENIGMA‐BD working group, we investigated T1‐weighted structural MRI data from 2436 participants with BD and healthy controls, and applied PCA to cortical thickness and surface area measures. We then studied the association of principal components with clinical and demographic variables using mixed regression models. We compared the PCA model with our prior clustering analyses of the same data and also tested it in a replication sample of 327 participants with BD or schizophrenia and healthy controls. The first principal component, which indexed a greater cortical thickness across all 68 cortical regions, was negatively associated with BD, BMI, antipsychotic medications, and age and was positively associated with Li treatment. PCA demonstrated superior goodness of fit to clustering when predicting diagnosis and BMI. Moreover, applying the PCA model to the replication sample yielded significant differences in cortical thickness between healthy controls and individuals with BD or schizophrenia. Cortical thickness in the same widespread regional network as determined by PCA was negatively associated with different clinical and demographic variables, including diagnosis, age, BMI, and treatment with antipsychotic medications or lithium. PCA outperformed clustering and provided an easy‐to‐use and interpret method to study multivariate associations between brain structure and system‐level variables. Practitioner Points: In this study of 2770 Individuals, we confirmed that cortical thickness in widespread regional networks as determined by principal component analysis (PCA) was negatively associated with relevant clinical and demographic variables, including diagnosis, age, BMI, and treatment with antipsychotic medications or lithium.Significant associations of many different system‐level variables with the same brain network suggest a lack of one‐to‐one mapping of individual clinical and demographic factors to specific patterns of brain changes.PCA outperformed clustering analysis in the same data set when predicting group or BMI, providing a superior method for studying multivariate associations between brain structure and system‐level variables. [ABSTRACT FROM AUTHOR]

Despite the importance of understanding how intelligence is ingrained in the function and structure of the brain in some neurological disorders, the alterations of intelligence‐associated neurological factors in atypical neurodevelopmental disorders, such as attention deficit/hyperactivity disorder (ADHD), are limited. Therefore, we aimed to explore the relationship between the brain functional and morphological characteristics and the intellectual performance of 139 patients with ADHD. Resting‐state functional and T1‐weighted structural magnetic resonance imaging (MRI) data and intellectual‐performance data of the patients were collected. The MRI data were preprocessed to extract four indicators characterizing the participants' brain features: fractional amplitude of low‐frequency fluctuation, regional homogeneity, and gray and white matter volumes. Then, we used a two‐layer feature‐selection method with support vector regression models based on three kernel functions to predict the verbal and performance intelligent quotients of the patients, along with ten fold cross‐validation to evaluate the models' predictive performance. All models showed good performance; the correlation coefficients between the predicted and observed values for each predictive phenotypic variable were >0.41, with statistical significance. The brain features that could best predict the intellectual performance of the patients were concentrated in the superior and inferior frontal gyrus of the prefrontal areas, the angular gyrus and precuneus of the parietal lobe, the inferior and middle temporal gyrus of the temporal lobe, and part of the cerebellar regions. Thus, the voxel‐based brain‐feature indicators could adequately predict the intellectual performance of patients with ADHD, providing a foundation for future neuroimaging studies of this disorder. [ABSTRACT FROM AUTHOR]

Introduction: The prefrontal–cingulate–thalamic areas are associated with ejaculation control. Functional abnormalities of these areas and decreased grey matter volume (GMV) in the subcortical areas have been confirmed in premature ejaculation (PE) patients. However, no study has explored the corresponding GMV changes in the prefrontal–cingulate–thalamic areas, which are considered as the important basis for functional abnormalities. This study aimed to investigated whether PE patients exhibited impaired GMV in the brain, especially the prefrontal–cingulate–thalamic areas, and whether these structural deficits were associated with declined ejaculatory control. Methods: T1‐weighted structural magnetic resonance imaging (MRI) data were acquired from 50 lifelong PE patients and 50 age‐, and education‐matched healthy controls (HCs). The PE diagnostic tool (PEDT) was applied to assess the subjective symptoms of PE. Based on the method of voxel‐based morphometry (VBM), GMV were measured and compared between groups. In addition, the correlations between GMV of brain regions showed differences between groups and PEDT scores were evaluated in the patient group. Results: PE patients showed decreased GMV in the right dorsolateral superior frontal gyrus (clusters = 13, peak T‐values = −4.30) and left thalamus (clusters = 47, T = −4.33), and increased GMV in the left middle cingulate gyrus (clusters = 12, T = 4.02) when compared with HCs. In the patient group, GMV of the left thalamus were negatively associated with PEDT scores (r = −0.35; P = 0.01). Receiver operating characteristic (ROC) analysis showed that GMV of the right dorsolateral superior frontal gyrus (AUC = 0.71, P < 0.01, sensitivity = 60%, specificity = 78%), left thalamus (AUC = 0.72, P < 0.01, sensitivity = 92%, specificity = 46%) and middle cingulate gyrus (AUC = 0.69, P < 0.01, sensitivity = 50%, specificity = 90%), and the combined model (AUC = 0.84, P < 0.01, sensitivity = 78%, specificity = 80%) all had the ability to distinguish PE patients from HCs. Conclusion: Disturbances in GMV were revealed in the prefrontal–cingulate–thalamic areas of PE patients. The findings implied that decreased GMV in the dorsolateral prefrontal cortex and thalamus might be associated with the central pathological neural mechanism underlying the declined ejaculatory control while increased GMV in the middle cingulate gyrus might be the compensatory mechanism underlying PE. [ABSTRACT FROM AUTHOR]

Background: The impact of left ventricular mechanical dyssynchrony (LVMD) on the long‐term prognosis of ST‐segment elevation myocardial infarction (STEMI) is unclear. Hypothesis: MR uniformity ratio estimates (URE) can detect LVMD and assess STEMI prognosis. Study Type: Retrospective analysis of a prospective multicenter registry (EARLY‐MYO trial, NCT 03768453). Population: Overall, 450 patients (50 females) with first‐time STEMI were analyzed, as well as 40 participants without cardiovascular disease as controls. Field Strength/Sequence: 3.0‐T, balanced steady‐state free precession cine and late gadolinium enhancement imaging. Assessment: MRI data were acquired within 1 week of symptom onset. Major adverse cardiovascular events (MACEs), including cardiovascular death, nonfatal re‐infarction, hospitalization for heart failure, and stroke, were the primary clinical outcomes. LVMD was represented by circumferential URE (CURE) and radial URE (RURE) calculated using strain measurements. The patients were grouped according to clinical outcomes or URE values. Patients' clinical characteristics and MR indicators were compared. Statistical Tests: The Student's t‐test, Mann–Whitney U test, chi‐square test, Fisher's exact test, receiver operating characteristic curve analysis with area under the curve, Kaplan–Meier analysis, Cox regression, logistic regression, intraclass correlation coefficient, c‐index, and integrated discrimination improvement were used. P < 0.05 was considered statistically significant. Results: CURE and RURE were significantly lower in patients with STEMI than in controls. The median follow‐up was 60.5 months. Patients with both lower CURE and RURE values experienced a significantly higher incidence of MACEs by 3.525‐fold. Both CURE and RURE were independent risk factors for MACEs. The addition of UREs improved diagnostic efficacy and risk stratification based on infarct size and left ventricular ejection fraction (LVEF). The indicators associated with LVMD included male sex, serum biomarkers (peak creatine phosphokinase and cardiac troponin I), infarct size, and LVEF. Data Conclusion: CURE and RURE may be useful to evaluate long‐term prognosis after STEMI. Evidence Level: 4 Technical Efficacy: Stage 2 [ABSTRACT FROM AUTHOR]

Background: MRI can provide information on kidney structure, perfusion, and oxygenation. Furthermore, it allows for the assessment of kidney sodium concentrations and handling, allowing multiparametric evaluation of kidney physiology. Multiparametric MRI is promising for establishing prognosis and monitoring treatment responses in kidney diseases, but its intraindividual variation during the day is unresolved. Purpose: To investigate the variation in multiparametric MRI measurements from the morning to the evening. Study Type: Prospective. Population: Ten healthy volunteers, aged 29 ± 5 without history of kidney disease. Field Strength/Sequence: 3 T/T1 mapping, blood‐oxygen level dependent imaging, arterial spin labeling perfusion imaging, diffusion weighted imaging, and sodium imaging. Assessment: A multiparametric MRI protocol, yielding T1, R2*, ADC, renal blood flow and renal sodium levels, was acquired in the morning, noon, and evening. The participants were fasting prior to the first examination. Urine biochemical analyses were performed to complement MRI data. The cortex and medulla were analyzed separately in a semi‐automatic fashion, and gradients of total sodium concentration (TSC) and R2* gradients were calculated from outer cortex to inner medulla. Statistical Test: Analyses of variance and mixed‐effects models to estimate differences from time of day. Coefficients of variation to assess variability within and between participants. A P‐value <0.05 was considered statistically significant. Results: The coefficients of variation varied from 5% to 18% for proton‐based parametric sequences, while it was 38% for TSC over a day. Data Conclusion: Multiparametric MRI is stable over the day. The coefficients of variation over a day were lower for proton multiparametric MRI, but higher for sodium MRI. Evidence Level: 2 Technical Efficacy: Stage 2 [ABSTRACT FROM AUTHOR]

Purpose: Subject movement during the MR examination is inevitable and causes not only image artifacts but also deteriorates the homogeneity of the main magnetic field (B0), which is a prerequisite for high quality data. Thus, characterization of changes to B0, for example induced by patient movement, is important for MR applications that are prone to B0 inhomogeneities. Methods: We propose a deep learning based method to predict such changes within the brain from the change of the head position to facilitate retrospective or even real‐time correction. A 3D U‐net was trained on in vivo gradient‐echo brain 7T MRI data. The input consisted of B0 maps and anatomical images at an initial position, and anatomical images at a different head position (obtained by applying a rigid‐body transformation on the initial anatomical image). The output consisted of B0 maps at the new head positions. We further fine‐trained the network weights to each subject by measuring a limited number of head positions of the given subject, and trained the U‐net with these data. Results: Our approach was compared to established dynamic B0 field mapping via interleaved navigators, which suffer from limited spatial resolution and the need for undesirable sequence modifications. Qualitative and quantitative comparison showed similar performance between an interleaved navigator‐equivalent method and proposed method. Conclusion: It is feasible to predict B0 maps from rigid subject movement and, when combined with external tracking hardware, this information could be used to improve the quality of MR acquisitions without the use of navigators. [ABSTRACT FROM AUTHOR]

Purpose: The objective was to develop a fully automated algorithm that generates confidence maps to identify regions valid for analysis of quantitative proton density fat fraction (PDFF) and R2*$$ {R}_2^{\ast } $$ maps of the liver, generated with chemical shift–encoded MRI (CSE‐MRI). Confidence maps are urgently needed for automated quality assurance, particularly with the emergence of automated segmentation and analysis algorithms. Methods: Confidence maps for both PDFF and R2*$$ {R}_2^{\ast } $$ maps are generated based on goodness of fit, measured by normalized RMS error between measured complex signals and the CSE‐MRI signal model. Based on Cramér‐Rao lower bound and Monte‐Carlo simulations, normalized RMS error threshold criteria were developed to identify unreliable regions in quantitative maps. Simulation, phantom, and in vivo clinical studies were included. To analyze the clinical data, a board‐certified radiologist delineated regions of interest (ROIs) in each of the nine liver segments for PDFF and R2*$$ {R}_2^{\ast } $$ analysis in consecutive clinical CSE‐MRI data sets. The percent area of ROIs in areas deemed unreliable by confidence maps was calculated to assess the impact of confidence maps on real‐world clinical PDFF and R2*$$ {R}_2^{\ast } $$ measurements. Results: Simulations and phantom studies demonstrated that the proposed algorithm successfully excluded regions with unreliable PDFF and R2*$$ {R}_2^{\ast } $$ measurements. ROI analysis by the radiologist revealed that 2.6% and 15% of the ROIs were placed in unreliable areas of PDFF and R2*$$ {R}_2^{\ast } $$ maps, as identified by confidence maps. Conclusion: A proposed confidence map algorithm that identifies reliable areas of PDFF and R2*$$ {R}_2^{\ast } $$ measurements from CSE‐MRI acquisitions was successfully developed. It demonstrated technical and clinical feasibility. [ABSTRACT FROM AUTHOR]

Human infancy is marked by fastest postnatal brain structural changes. It also coincides with the onset of many neurodevelopmental disorders. Atlas‐based automated structure labeling has been widely used for analyzing various neuroimaging data. However, the relatively large and nonlinear neuroanatomical differences between infant and adult brains can lead to significant offsets of the labeled structures in infant brains when adult brain atlas is used. Age‐specific 1‐ and 2‐year‐old brain atlases covering all major gray and white matter (GM and WM) structures with diffusion tensor imaging (DTI) and structural MRI are critical for precision medicine for infant population yet have not been established. In this study, high‐quality DTI and structural MRI data were obtained from 50 healthy children to build up three‐dimensional age‐specific 1‐ and 2‐year‐old brain templates and atlases. Age‐specific templates include a single‐subject template as well as two population‐averaged templates from linear and nonlinear transformation, respectively. Each age‐specific atlas consists of 124 comprehensively labeled major GM and WM structures, including 52 cerebral cortical, 10 deep GM, 40 WM, and 22 brainstem and cerebellar structures. When combined with appropriate registration methods, the established atlases can be used for highly accurate automatic labeling of any given infant brain MRI. We demonstrated that one can automatically and effectively delineate deep WM microstructural development from 3 to 38 months by using these age‐specific atlases. These established 1‐ and 2‐year‐old infant brain DTI atlases can advance our understanding of typical brain development and serve as clinical anatomical references for brain disorders during infancy. [ABSTRACT FROM AUTHOR]

Objective: The aim of the study was to evaluate the benefits of morphometric magnetic resonance imaging (MRI) postprocessing in patients presenting with a first seizure and negative MRI results and to investigate these findings in the context of the clinical and electroencephalographic data, seizure recurrence rates, and epilepsy diagnosis in these patients. Methods: We retrospectively reviewed 97 MRI scans of patients with first unprovoked epileptic seizure and no evidence of epileptogenic lesion on clinical routine MRI. Morphometric Analysis Program (MAP; v2018), automated postprocessing software, was used to identify subtle, potentially epileptogenic lesions in the three‐dimensional T1‐weighted MRI data. The resulting probability maps were examined together with the conventional MRI images by a reviewer who remained blinded to the patients' clinical and electroencephalographical data. Clinical data were prospectively collected between February 2018 and May 2023. Results: Among the apparently MRI‐negative patients, a total of 18 of 97 (18.6%) showed cortical changes suggestive of focal cortical dysplasia. Within the population with positive MAP findings (MAP+), seizure recurrence rates were 61.1% and 66.7% at 1 and 2 years after the first unprovoked seizure, respectively. Conversely, patients with negative MAP findings (MAP−) had lower seizure recurrence rates of 27.8% and 34.2% at 1 and 2 years after the first unprovoked seizure, respectively. Patients with MAP+ findings were significantly more likely to be diagnosed with epilepsy than those patients with MAP− findings (χ2 [1, n = 97] = 14.820, p <.001, odds ratio = 21.371, 95% CI = 2.710–168.531) during a mean follow‐up time of 22.51 months (SD = 16.7 months, range = 1–61 months). Significance: MRI postprocessing can be a valuable tool for detecting subtle epileptogenic lesions in patients with a first seizure and negative MRI results. Patients with first seizure and MAP+ findings had high seizure recurrence rates, meeting the criteria for beginning epilepsy. [ABSTRACT FROM AUTHOR]

Objective: This study aims to explore the frequency and influencing factors of asymptomatic spinal lesions (ASLs) and their impact on subsequent relapses in patients with AQP4‐IgG‐positive NMOSD (AQP4‐NMOSD) in a real‐world setting. Methods: We retrospectively reviewed clinical information and spinal MRI data from AQP4‐NMOSD patients who had at least one spinal cord MRI during their follow‐ups. Kaplan–Meier curves and Cox proportional hazards models were employed to ascertain potential predictors of remission ASLs and to investigate factors associated with subsequent relapses. Results: In this study, we included 129 patients with AQP4‐NMOSD and reviewed 173 spinal MRIs during attacks and 89 spinal MRIs during remission. Among these, 6 ASLs (3.5%) were identified during acute attacks, while 8 ASLs (9%) were found during remission. Remission ASLs were linked to the use of immunosuppressive agents, particularly conventional ones, whereas no patients using rituximab developed ASLs (p = 0.005). Kaplan–Meier curve analysis indicated that patients with ASLs had a significantly higher relapse risk (HR = 4.658, 95% CI: 1.519–14.285, p = 0.007) compared to those without. Additionally, the use of mycophenolate mofetil (HR = 0.027, 95% CI: 0.003–0.260, p = 0.002) and rituximab (HR = 0.035, 95% CI: 0.006–0.203, p < 0.001) significantly reduced the relapse risk. However, after accounting for other factors, the presence of ASLs did not exhibit a significant impact on subsequent relapses (HR = 2.297, 95% CI: 0.652–8.085, p = 0.195). Interpretation: ASLs may be observed in patients with AQP4‐NMOSD. The presence of ASLs may signify an underlying inflammatory activity due to insufficient immunotherapy. The administration of immunosuppressive agents plays a key role in the presence of remission ASLs and the likelihood of subsequent relapses. [ABSTRACT FROM AUTHOR]

Background: There is limited data in the literature regarding the role of nonarthrographic MRI for detecting biceps pulley (BP) lesions. Purpose: To assess the accuracy of nonarthrographic MRI for detecting BP lesions, and to evaluate the diagnostic value of various MRI signs (superior glenohumeral ligament discontinuity/nonvisibility, long head of biceps (LHB) displacement sign or subluxation/dislocation, LHB tendinopathy, and supraspinatus and subscapularis tendon lesions) in detecting such lesions. Study Type: Retrospective. Population: 84 patients (32 in BP‐lesion group and 52 in BP‐intact group‐as confirmed by arthroscopy). Field Strength/Sequence: 1.5‐T, T1‐weighted turbo spin echo (TSE), T2‐weighted TSE, and proton density‐weighted TSE spectral attenuated inversion recovery (SPAIR) sequences. Assessment: Three radiologists independently reviewed all MRI data for the presence of BP lesions and various MRI signs. The MRI signs and final MRI diagnoses were tested for accuracy regarding detecting BP lesions using arthroscopy results as the reference standard. Furthermore, the inter‐reader agreement (IRA) between radiologists was determined. Statistical Tests: Student's t‐tests, Chi‐squared, and Fisher's exact tests, and 4‐fold table test were used. The IRA was calculated using Kappa statistics. A P‐value <0.05 was considered statistically significant. Results: The sensitivity, specificity, and accuracy of nonarthrographic MRI for detecting BP lesions were 65.6%–78.1%, 90.4%–92.3%, and 81%–86.9%, respectively. The highest accuracy was noticed for the LHB displacement sign (84.5%–86.9%), and the highest sensitivity was registered for the LHB tendinopathy sign (87.5%). Furthermore, the highest specificity was observed for the LHB displacement sign and LHB subluxation/dislocation sign (98.1%–100%). The IRA regarding final MRI diagnosis and MRI signs of BP lesions was good to very good (κ = 0.76–0.98). Data Conclusion: Nonarthrographic shoulder MRI may show good diagnostic accuracy for detecting BP lesions. The LHB displacement sign could serve as the most accurate and specific sign for diagnosis of BP lesions. Level of Evidence: 3 Technical Efficacy: Stage 2 [ABSTRACT FROM AUTHOR]

Background: Dynamic contrast‐enhanced (DCE) MRI commonly outperforms diffusion‐weighted (DW) MRI in breast cancer discrimination. However, the side effects of contrast agents limit the use of DCE‐MRI, particularly in patients with chronic kidney disease. Purpose: To develop a novel deep learning model to fully exploit the potential of overall b‐value DW‐MRI without the need for a contrast agent in predicting breast cancer molecular subtypes and to evaluate its performance in comparison with DCE‐MRI. Study Type: Prospective. Subjects: 486 female breast cancer patients (training/validation/test: 64%/16%/20%). Field Strength/Sequence: 3.0 T/DW‐MRI (13 b‐values) and DCE‐MRI (one precontrast and five postcontrast phases). Assessment: The breast cancers were divided into four subtypes: luminal A, luminal B, HER2+, and triple negative. A channel‐dimensional feature‐reconstructed (CDFR) deep neural network (DNN) was proposed to predict these subtypes using pathological diagnosis as the reference standard. Additionally, a non‐CDFR DNN (NCDFR‐DNN) was built for comparative purposes. A mixture ensemble DNN (ME‐DNN) integrating two CDFR‐DNNs was constructed to identify subtypes on multiparametric MRI (MP‐MRI) combing DW‐MRI and DCE‐MRI. Statistical Tests: Model performance was evaluated using accuracy, sensitivity, specificity, and area under the receiver operating characteristic curve (AUC). Model comparisons were performed using the one‐way analysis of variance with least significant difference post hoc test and the DeLong test. P < 0.05 was considered significant. Results: The CDFR‐DNN (accuracies, 0.79 ~ 0.80; AUCs, 0.93 ~ 0.94) demonstrated significantly improved predictive performance than the NCDFR‐DNN (accuracies, 0.76 ~ 0.78; AUCs, 0.92 ~ 0.93) on DW‐MRI. Utilizing the CDFR‐DNN, DW‐MRI attained the predictive performance equal (P = 0.065 ~ 1.000) to DCE‐MRI (accuracies, 0.79 ~ 0.80; AUCs, 0.93 ~ 0.95). The predictive performance of the ME‐DNN on MP‐MRI (accuracies, 0.85 ~ 0.87; AUCs, 0.96 ~ 0.97) was superior to those of both the CDFR‐DNN and NCDFR‐DNN on either DW‐MRI or DCE‐MRI. Data Conclusion: The CDFR‐DNN enabled overall b‐value DW‐MRI to achieve the predictive performance comparable to DCE‐MRI. MP‐MRI outperformed DW‐MRI and DCE‐MRI in subtype prediction. Level of Evidence: 2 Technical Efficacy Stage: 1 [ABSTRACT FROM AUTHOR]

Background: Deep stromal invasion (DSI) is one of the predominant risk factors that determined the types of radical hysterectomy (RH). Thus, the accurate assessment of DSI in cervical adenocarcinoma (AC)/adenosquamous carcinoma (ASC) can facilitate optimal therapy decision. Purpose: To develop a nomogram to identify DSI in cervical AC/ASC. Study Type: Retrospective. Population: Six hundred and fifty patients (mean age of 48.2 years) were collected from center 1 (primary cohort, 536), centers 2 and 3 (external validation cohorts 1 and 2, 62 and 52). Field Strength/Sequence: 5‐T, T2‐weighted imaging (T2WI, SE/FSE), diffusion‐weighted imaging (DWI, EPI), and contrast‐enhanced T1‐weighted imaging (CE‐T1WI, VIBE/LAVA). Assessment: The DSI was defined as the outer 1/3 stromal invasion on pathology. The region of interest (ROI) contained the tumor and 3 mm peritumoral area. The ROIs of T2WI, DWI, and CE‐T1WI were separately imported into Resnet18 to calculate the DL scores (TDS, DDS, and CDS). The clinical characteristics were retrieved from medical records or MRI data assessment. The clinical model and nomogram were constructed by integrating clinical independent risk factors only and further combining DL scores based on primary cohort and were validated in two external validation cohorts. Statistical Tests: Student's t‐test, Mann–Whitney U test, or Chi‐squared test were used to compare differences in continuous or categorical variables between DSI‐positive and DSI‐negative groups. DeLong test was used to compare AU‐ROC values of DL scores, clinical model, and nomogram. Results: The nomogram integrating menopause, disruption of cervical stromal ring (DCSRMR), DDS, and TDS achieved AU‐ROCs of 0.933, 0.807, and 0.817 in evaluating DSI in primary and external validation cohorts. The nomogram had superior diagnostic ability to clinical model and DL scores in primary cohort (all P < 0.0125 [0.05/4]) and CDS (P = 0.009) in external validation cohort 2. Data Conclusion: The nomogram achieved good performance for evaluating DSI in cervical AC/ASC. Level of Evidence: 3 Technical Efficacy: Stage 2 [ABSTRACT FROM AUTHOR]