Evaluation of convolutional neural networks for the detection of inter-breath-hold motion from a stack of cardiac short axis slice images.

PURPOSE: This study aimed to develop and validate a deep learning-based method that detects inter-breath-hold motion from an estimated cardiac long axis image reconstructed from a stack of short axis cardiac cine images. METHODS: Cardiac cine magnetic resonance image data from all short axis slices and 2-/3-/4-chamber long axis slices were considered for the study. Data from 740 subjects were used for model development, and data from 491 subjects were used for testing. The method utilized the slice orientation information to calculate the intersection line of a short axis plane and a long axis plane. An estimated long axis image is shown along with a long axis image as a motion-free reference image, which enables visual assessment of the inter-breath-hold motion from the estimated long axis image. The estimated long axis image was labeled as either a motion-corrupted or a motion-free image. Deep convolutional neural network (CNN) models were developed and validated using the labeled data. RESULTS: The method was fully automatic in obtaining long axis images reformatted from a 3D stack of short axis slices and predicting the presence/absence of inter-breath-hold motion. The deep CNN model with EfficientNet-B0 as a feature extractor was effective at motion detection with an area under the receiver operating characteristic (AUC) curve of 0.87 for the testing data. CONCLUSION: The proposed method can automatically assess inter-breath-hold motion in a stack of cardiac cine short axis slices. The method can help prospectively reacquire problematic short axis slices or retrospectively correct motion.

Fast calculation software for modified Look-Locker inversion recovery (MOLLI) T1 mapping.

BACKGROUND: The purpose of this study was to develop a software tool and evaluate different T1 map calculation methods in terms of computation time in cardiac magnetic resonance imaging. METHODS: The modified Look-Locker inversion recovery (MOLLI) sequence was used to acquire multiple inversion time (TI) images for pre- and post-contrast T1 mapping. The T1 map calculation involved pixel-wise curve fitting based on the T1 relaxation model. A variety of methods were evaluated using data from 30 subjects for computational efficiency: MRmap, python Levenberg-Marquardt (LM), python reduced-dimension (RD) non-linear least square, C++ single- and multi-core LM, and C++ single- and multi-core RD. RESULTS: Median (interquartile range) computation time was 126 s (98-141) for the publicly available software MRmap, 261 s (249-282) for python LM, 77 s (74-80) for python RD, 3.4 s (3.1-3.6) for C++ multi-core LM, and 1.9 s (1.9-2.0) for C++ multi-core RD. The fastest C++ multi-core RD and the publicly available MRmap showed good agreement of myocardial T1 values, resulting in 95% Bland-Altman limits of agreement of (- 0.83 to 0.58 ms) and (- 6.57 to 7.36 ms) with mean differences of - 0.13 ms and 0.39 ms, for the pre- and post-contrast, respectively. CONCLUSION: The C++ multi-core RD was the fastest method on a regular eight-core personal computer for pre- or post-contrast T1 map calculation. The presented software tool (fT1fit) facilitated rapid T1 map and extracellular volume fraction map calculations.

Admission Diffusion-Weighted Imaging Lesion Volume in Patients With Large Vessel Occlusion Stroke and Alberta Stroke Program Early CT Score of ≥6 Points: Serial Computed Tomography-Magnetic Resonance Imaging Collateral Measurements.

Background and Purpose- We hypothesized that the pial collateral status at the time of presentation could predict the infarct size on magnetic resonance imaging in patients with similar degrees of early ischemic changes on computed tomography. We tested the association between serial changes in collateral status and infarct volume defined on diffusion-weighted imaging (DWI) in patients with large vessel occlusion and small core. Methods- Consecutive patients who were candidates for endovascular treatment (Alberta Stroke Program Early CT Score [ASPECTS] of ≥6 points) and who underwent both pretreatment multiphasic computed tomography angiography (mCTA) and multimodal magnetic resonance imaging were enrolled. The baseline early ischemic changes and collateral status were determined using both mCTA and magnetic resonance imaging-based collateral maps. Multivariable linear regression was used to evaluate adjusted estimates of the effect of collateral status on predicting MR DWI lesion volume before endovascular treatment. Results- Of 65 patients (39 men; median age, 76 years; median ASPECTS, 8 points [range, 6-10]), 10 (15.4%), 8 (12.3%), and 47 (72.3%) presented poor, intermediate, and good collaterals on mCTA, respectively. After adjusting for the initial stroke severity, ASPECTS, time to DWI, and mismatch volume, the mCTA collateral grade was the only factor independently associated with the DWI lesion volume (ß=-35.657, SE mean=3.539; P<0.0001). An excellent correlation between the mCTA- and magnetic resonance imaging-based collateral grades was observed (matching grade seen in 92.3%), suggesting a collateral status persistence during the hyperacute stroke phase. Conclusions- The mCTA assessed collateral adequacy is the sole predictor of eventual DWI lesion volume before endovascular treatment. The added value of collateral assessment in early ischemic changes and large vessel occlusion for decision-making regarding more aggressive revascularizations requires further evaluation. Clinical Trial Registration- URL: https://www.clinicaltrials.gov. Unique identifier: NCT03234634 and NCT02668627.

Evaluation of Diffusion Lesion Volume Measurements in Acute Ischemic Stroke Using Encoder-Decoder Convolutional Network.

Background and Purpose- Automatic segmentation of cerebral infarction on diffusion-weighted imaging (DWI) is typically performed based on a fixed apparent diffusion coefficient (ADC) threshold. Fixed ADC threshold methods may not be accurate because ADC values vary over time after stroke onset. Deep learning has the potential to improve the accuracy, provided that a large set of correctly annotated lesion data is used for training. The purpose of this study was to evaluate deep learning-based methods and compare them with commercial software in terms of lesion volume measurements. Methods- U-net, an encoder-decoder convolutional neural network, was adopted to train segmentation models. Two U-net models were developed: a U-net (DWI+ADC) model, trained on DWI and ADC data, and a U-net (DWI) model, trained on DWI data only. A total of 296 subjects were used for training and 134 for external validation. An expert neurologist manually delineated the stroke lesions on DWI images, which were used as the ground-truth reference. Lesion volume measurements from the U-net methods were compared against the expert's manual segmentation and Rapid Processing of Perfusion and Diffusion (RAPID; iSchemaView Inc) analysis. Results- In external validation, U-net (DWI+ADC) showed the highest intraclass correlation coefficient with manual segmentation (intraclass correlation coefficient, 1.0; 95% CI, 0.99-1.00) and sufficiently high correlation with the RAPID results (intraclass correlation coefficient, 0.99; 95% CI, 0.98-0.99). U-net (DWI+ADC) and manual segmentation resulted in the smallest 95% Bland-Altman limits of agreement (-5.31 to 4.93 mL) with a mean difference of -0.19 mL. Conclusions- The presented deep learning-based method is fully automatic and shows a high correlation of diffusion lesion volume measurements with manual segmentation and commercial software. The method has the potential to be used in patient selection for endovascular reperfusion therapy in the late time window of acute stroke.

Intra-individual comparison of gadolinium-enhanced MRI using pseudo-golden-angle radial acquisition with gadoxetic acid-enhanced MRI for diagnosis of HCCs using LI-RADS.

OBJECTIVES: To determine the usefulness of extracellular contrast agent (ECA)-enhanced multiphasic liver magnetic resonance imaging (MRI) using a pseudo-golden-angle radial acquisition scheme by intra-individual comparison with gadoxetic acid-MRI (EOB-MRI) with regard to image quality and the diagnosis of hepatocellular carcinoma (HCC). MATERIALS AND METHODS: This prospective study enrolled 15 patients with 18 HCCs who underwent EOB-MRI using a Cartesian approach and ECA-MRI using the pseudo-golden-angle radial acquisition scheme (free-breathing continuous data acquisition for 64 s following ECA injection, generating six images). Two reviewers evaluated the arterial and portal phases of each MRI for artifacts, organ sharpness, and conspicuity of intrahepatic vessels and the hepatic tumors. A Liver Imaging Reporting and Data System category was also assigned to each lesion. RESULTS: There were no differences in the subjective image quality analysis between the arterial phases of two MRIs (p > 0.05). However, ghosting artifact was seen only in EOB-MRI (N = 3). Six HCCs showed different signal intensities in the arterial phase or portal phase between the two MRIs; five HCCs showed arterial hyperenhancement on ECA-MRI, but not on EOB-MRI. The capsule was observed in 15 HCCs on ECA-MRI and 6 HCCs on EOB-MRI. Five and one HCC were assigned as LR-5 and LR-4 with ECA-MRI and LR-4 and LR-3 with EOB-MRI, respectively. CONCLUSION: Free-breathing ECA-enhanced multiphasic liver MRI using a pseudo-golden-angle radial acquisition was more sensitive in detecting arterial hyperenhancement of HCC than conventional EOB-MRI, and the image quality was acceptable. KEY POINTS: • The pseudo-golden-angle radial acquisition scheme can be applied to perform free-breathing multiphasic dynamic liver MRI. • Adopting the pseudo-golden-angle radial acquisition scheme can improve the detection of arterial enhancement of HCC. • The pseudo-golden-angle radial acquisition scheme enables motion-free liver MRI.

A fast and flexible MRI system for the study of dynamic vocal tract shaping.

PURPOSE: The aim of this work was to develop and evaluate an MRI-based system for study of dynamic vocal tract shaping during speech production, which provides high spatial and temporal resolution. METHODS: The proposed system utilizes (a) custom eight-channel upper airway coils that have high sensitivity to upper airway regions of interest, (b) two-dimensional golden angle spiral gradient echo acquisition, (c) on-the-fly view-sharing reconstruction, and (d) off-line temporal finite difference constrained reconstruction. The system also provides simultaneous noise-cancelled and temporally aligned audio. The system is evaluated in 3 healthy volunteers, and 1 tongue cancer patient, with a broad range of speech tasks. RESULTS: We report spatiotemporal resolutions of 2.4 × 2.4 mm2 every 12 ms for single-slice imaging, and 2.4 × 2.4 mm2 every 36 ms for three-slice imaging, which reflects roughly 7-fold acceleration over Nyquist sampling. This system demonstrates improved temporal fidelity in capturing rapid vocal tract shaping for tasks, such as producing consonant clusters in speech, and beat-boxing sounds. Novel acoustic-articulatory analysis was also demonstrated. CONCLUSION: A synergistic combination of custom coils, spiral acquisitions, and constrained reconstruction enables visualization of rapid speech with high spatiotemporal resolution in multiple planes. Magn Reson Med 77:112-125, 2017. © 2016 Wiley Periodicals, Inc.

Evaluation of an independent linear model for acoustic noise on a conventional MRI scanner and implications for acoustic noise reduction.

PURPOSE: To evaluate an independent linear model for gradient acoustic noise on a conventional MRI scanner, and to explore implications for acoustic noise reduction in routine imaging. METHODS: Acoustic noise generated from each physical gradient axis was modeled as the prescribed gradient waveform passed through a linear time-invariant system. Homogeneity and superposition properties were experimentally determined. We also developed a new method to correct relative time shifts between the measured impulse responses for different physical gradient axes. Model accuracy was determined by comparing predicted and measured sound using normalized energy difference. Transfer functions were also measured in subjects with different body habitus and at multiple microphone locations. RESULTS: Both superposition and homogeneity held for each physical gradient axis with errors less than 3%. When all gradients were on simultaneous sound prediction, error was reduced from 32% to 4% after time-shift correction. Transfer functions also showed high sensitivity to body habitus and microphone location. CONCLUSION: The independent linear model predicts MRI acoustic noise with less than 4% error. Acoustic transfer functions are highly sensitive to body habitus and position within the bore, making it challenging to produce a general approach to acoustic noise reduction based on avoiding system resonance peaks.

Real-time 3D magnetic resonance imaging of the pharyngeal airway in sleep apnea.

PURPOSE: To investigate the feasibility of real-time 3D magnetic resonance imaging (MRI) with simultaneous recording of physiological signals for identifying sites of airway obstruction during natural sleep in pediatric patients with sleep-disordered breathing. METHODS: Experiments were performed using a three-dimensional Fourier transformation (3DFT) gradient echo sequence with prospective undersampling based on golden-angle radial spokes, and L1-norm regularized iterative self-consistent parallel imaging (L1-SPIRiT) reconstruction. This technique was demonstrated in three healthy adult volunteers and five pediatric patients with sleep-disordered breathing. External airway occlusion was used to induce partial collapse of the upper airway on inspiration and test the effectiveness of the proposed imaging method. Apneic events were identified using information available from synchronized recording of mask pressure and respiratory effort. RESULTS: Acceptable image quality was obtained in seven of eight subjects. Temporary airway collapse induced via inspiratory loading was successfully imaged in all three volunteers, with average airway volume reductions of 63.3%, 52.5%, and 33.7%. Central apneic events and associated airway narrowing/closure were identified in two pediatric patients. During central apneic events, airway obstruction was observed in the retropalatal region in one pediatric patient. CONCLUSION: Real-time 3D MRI of the pharyngeal airway with synchronized recording of physiological signals is feasible and may provide valuable information about the sites and nature of airway narrowing/collapse during natural sleep.

Real-time magnetic resonance imaging and electromagnetic articulography database for speech production research (TC).

USC-TIMIT is an extensive database of multimodal speech production data, developed to complement existing resources available to the speech research community and with the intention of being continuously refined and augmented. The database currently includes real-time magnetic resonance imaging data from five male and five female speakers of American English. Electromagnetic articulography data have also been presently collected from four of these speakers. The two modalities were recorded in two independent sessions while the subjects produced the same 460 sentence corpus used previously in the MOCHA-TIMIT database. In both cases the audio signal was recorded and synchronized with the articulatory data. The database and companion software are freely available to the research community.

An Improved Path-Finding Method for the Tracking of Centerlines of Tortuous Internal Carotid Arteries in MR Angiography.

Centerline tracking is useful in performing segmental analysis of vessel tortuosity in angiography data. However, a highly tortuous) artery can produce multiple centerlines due to over-segmentation of the artery, resulting in inaccurate path-finding results when using the shortest path-finding algorithm. In this study, the internal carotid arteries (ICAs) from three-dimensional (3D) time-of-flight magnetic resonance angiography (TOF MRA) data were used to demonstrate the effectiveness of a new path-finding method. The method is based on a series of depth-first searches (DFSs) with randomly different orders of neighborhood searches and produces an appropriate path connecting the two endpoints in the ICAs. It was compared with three existing methods which were (a) DFS with a sequential order of neighborhood search, (b) Dijkstra algorithm, and (c) A* algorithm. The path-finding accuracy was evaluated by counting the number of successful paths. The method resulted in an accuracy of 95.8%, outperforming the three existing methods. In conclusion, the proposed method has been shown to be more suitable as a path-finding procedure than the existing methods, particularly in cases where there is more than one centerline resulting from over-segmentation of a highly tortuous artery.

Deep learning for classification of late gadolinium enhancement lesions based on the 16-segment left ventricular model.

PURPOSE: This study aimed to develop and validate a deep learning-based method that allows for segmental analysis of myocardial late gadolinium enhancement (LGE) lesions. METHODS: Cardiac LGE data from 170 patients with coronary artery disease and non-ischemic heart disease were used for training, validation, and testing. Short-axis images were transformed to polar space after identification of the left ventricular (LV) center point and anterior right ventricular (RV) insertion point. Images were obtained after dividing the polar transformed images into segments based on the 16-segment LV model. Five different deep convolutional neural network (CNN) models were developed and validated using the labeled data, where the image after the division corresponded to a segment, and the lesion labeling was based on the 16-segment LV model. Unseen testing data were used to evaluate the performance of the lesion classification. RESULTS: Without manual lesion segmentation and annotation, the proposed method showed an area under the curve (AUC) of 0.875, and a precision, recall, and F1-score of 0.723, 0.783, and 0.752, respectively for the lesion class when the pretrained ResNet50 model was tested for all slice images. The two pretrained models of ResNet50 and EfficientNet-B0 outperformed the three non-pretrained CNN models in terms of AUCs (0.873-0.875 vs. 0.834-0.841). CONCLUSION: The proposed method is based on learning a deep CNN model from polar transformed images to predict LGE lesions with good accuracy and does not require time-consuming annotation procedures such as lesion segmentation.

Effects of Path-Finding Algorithms on the Labeling of the Centerlines of Circle of Willis Arteries.

Quantitative analysis of intracranial vessel segments typically requires the identification of the vessels' centerlines, and a path-finding algorithm can be used to automatically detect vessel segments' centerlines. This study compared the performance of path-finding algorithms for vessel labeling. Three-dimensional (3D) time-of-flight magnetic resonance angiography (MRA) images from the publicly available dataset were considered for this study. After manual annotations of the endpoints of each vessel segment, three path-finding methods were compared: (Method 1) depth-first search algorithm, (Method 2) Dijkstra's algorithm, and (Method 3) A* algorithm. The rate of correctly found paths was quantified and compared among the three methods in each segment of the circle of Willis arteries. In the analysis of 840 vessel segments, Method 2 showed the highest accuracy (97.1%) of correctly found paths, while Method 1 and 3 showed an accuracy of 83.5% and 96.1%, respectively. The AComm artery was highly inaccurately identified in Method 1, with an accuracy of 43.2%. Incorrect paths by Method 2 were noted in the R-ICA, L-ICA, and R-PCA-P1 segments. The Dijkstra and A* algorithms showed similar accuracy in path-finding, and they were comparable in the speed of path-finding in the circle of Willis arterial segments.

Automated identification of myocardial perfusion defects in dynamic cardiac computed tomography using deep learning.

PURPOSE: The purpose of this study was to develop and evaluate deep convolutional neural network (CNN) models for quantifying myocardial blood flow (MBF) as well as for identifying myocardial perfusion defects in dynamic cardiac computed tomography (CT) images. METHODS: Adenosine stress cardiac CT perfusion data acquired from 156 patients having or being suspected with coronary artery disease were considered for model development and validation. U-net-based deep CNN models were developed to segment the aorta and myocardium and to localize anatomical landmarks. Color-coded MBF maps were obtained in short-axis slices from the apex to the base level and were used to train a deep CNN classifier. Three binary classification models were built for the detection of perfusion defect in the left anterior descending artery (LAD), the right coronary artery (RCA), and the left circumflex artery (LCX) territories. RESULTS: Mean Dice scores were 0.94 (±0.07) and 0.86 (±0.06) for the aorta and myocardial deep learning-based segmentations, respectively. With the localization U-net, mean distance errors were 3.5 (±3.5) mm and 3.8 (±2.4) mm for the basal and apical center points, respectively. The classification models identified perfusion defects with the accuracy of mean area under the receiver operating curve (AUROC) values of 0.959 (±0.023) for LAD, 0.949 (±0.016) for RCA, and 0.957 (±0.021) for LCX. CONCLUSION: The presented method has the potential to fully automate the quantification of MBF and subsequently identify the main coronary artery territories with myocardial perfusion defects in dynamic cardiac CT perfusion.

Assessing Machine Learning Models for Predicting Age with Intracranial Vessel Tortuosity and Thickness Information.

This study aimed to develop and validate machine learning (ML) models that predict age using intracranial vessels' tortuosity and diameter features derived from magnetic resonance angiography (MRA) data. A total of 171 subjects' three-dimensional (3D) time-of-flight MRA image data were considered for analysis. After annotations of two endpoints in each arterial segment, tortuosity features such as the sum of the angle metrics, triangular index, relative length, and product of the angle distance, as well as the vessels' diameter features, were extracted and used to train and validate the ML models for age prediction. Features extracted from the right and left internal carotid arteries (ICA) and basilar arteries were considered as the inputs to train and validate six ML regression models with a four-fold cross validation. The random forest regression model resulted in the lowest root mean square error of 14.9 years and the highest average coefficient of determination of 0.186. The linear regression model showed the lowest average mean absolute percentage error (MAPE) and the highest average Pearson correlation coefficient (0.532). The mean diameter of the right ICA vessel segment was the most important feature contributing to prediction of age in two out of the four regression models considered. An ML of tortuosity descriptors and diameter features extracted from MRA data showed a modest correlation between real age and ML-predicted age. Further studies are warranted for the assessment of the model's age predictions in patients with intracranial vessel diseases.

Automatic calculation of myocardial perfusion reserve using deep learning with uncertainty quantification.

Background: Myocardial perfusion reserve index (MPRI) in magnetic resonance imaging (MRI) is an important indicator of ischemia, and its measurement typically involves manual procedures. The purposes of this study were to develop a fully automatic method for estimating the MPRI and to evaluate its performance. Methods: The method consisted of segmenting the myocardium in dynamic contrast-enhanced (DCE) myocardial perfusion MRI data using Monte Carlo dropout U-Net, dividing the myocardium into segments based on landmark localization with machine learning, and estimating the MPRI after the calculation of the left ventricular and myocardial contrast upslopes. The proposed method was compared with a reference method, which involved manual adjustments of the myocardial contours and upslope ranges. Results: In test subjects, MPRIs measured by the proposed technique correlated with those by the manual reference in segmental assessment [intraclass correlation coefficient (ICC) =0.75, 95% CI: 0.70-0.79, P<0.001]. The automatic and reference MPRI values showed a mean difference of -0.02 and 95% limits of agreement of (-0.86, 0.82). Conclusions: The proposed automatic method is based on deep learning segmentation and machine learning landmark detection for MPRI measurements in DCE perfusion MRI. It holds the potential to efficiently and quantitatively assess myocardial ischemia without any user's interaction.

A Deep Learning-Based Automatic Collateral Assessment in Patients with Acute Ischemic Stroke.

This study aimed to develop a supervised deep learning (DL) model for grading collateral status from dynamic susceptibility contrast magnetic resonance perfusion (DSC-MRP) images from patients with large vessel occlusion (LVO) acute ischemic stroke (AIS) and compare its performance against experts' manual grading. Among consecutive LVO-AIS at three medical center sites, DSC-MRP data were processed to generate collateral flow maps consisting of arterial, capillary, and venous phases. With the use of expert readings as a reference, a DL model was developed to analyze collateral status with output classified into good and poor grades. The resulting model was externally validated in a later-collected population from one medical center site. The model was trained on 255 patients and externally validated on 72 patients. In the all-site internal validation population, DL grading of good collateral probability yielded a c statistic of 0.91; in the external validation population, the c statistic was 0.85. In the external validation population, there was moderate agreement between the experts' grades and DL grades (kappa = 0.53, 95% CI = 0.32-0.73, p < 0.0001). Day 7 infarct growth volume was higher in DL-graded poor collateral group than good collateral group patients (median volume [26 mL vs. 6 mL], p = 0.01) in patients with successful reperfusion (modified treatment in cerebral infarction (mTICI) = 2b-3). In all patients with a 90-day modified Rankin Scale (mRS) score, there was a shift to more favorable outcomes in the good collateral group, with a common odds ratio of 2.99 (95% CI = 1.89-4.76, p < 0.0001). The DL-based collateral grading was in good agreement with expert manual grading in both development and validation populations. After exclusion of patients with large infarct volume, early reperfusion is more likely to benefit patients with the poor collateral flow, and the DL method has the potential to aid the assessment of collateral status.

Automated in-depth cerebral arterial labelling using cerebrovascular vasculature reframing and deep neural networks.

Identifying the cerebral arterial branches is essential for undertaking a computational approach to cerebrovascular imaging. However, the complexity and inter-individual differences involved in this process have not been thoroughly studied. We used machine learning to examine the anatomical profile of the cerebral arterial tree. The method is less sensitive to inter-subject and cohort-wise anatomical variations and exhibits robust performance with an unprecedented in-depth vessel range. We applied machine learning algorithms to disease-free healthy control subjects (n = 42), patients with stroke with intracranial atherosclerosis (ICAS) (n = 46), and patients with stroke mixed with the existing controls (n = 69). We trained and tested 70% and 30% of each study cohort, respectively, incorporating spatial coordinates and geometric vessel feature vectors. Cerebral arterial images were analyzed based on the 'segmentation-stacking' method using magnetic resonance angiography. We precisely classified the cerebral arteries across the exhaustive scope of vessel components using advanced geometric characterization, redefinition of vessel unit conception, and post-processing algorithms. We verified that the neural network ensemble, with multiple joint models as the combined predictor, classified all vessel component types independent of inter-subject variations in cerebral arterial anatomy. The validity of the categorization performance of the model was tested, considering the control, ICAS, and control-blended stroke cohorts, using the area under the receiver operating characteristic (ROC) curve and precision-recall curve. The classification accuracy rarely fell outside each image's 90-99% scope, independent of cohort-dependent cerebrovascular structural variations. The classification ensemble was calibrated with high overall area rates under the ROC curve of 0.99-1.00 [0.97-1.00] in the test set across various study cohorts. Identifying an all-inclusive range of vessel components across controls, ICAS, and stroke patients, the accuracy rates of the prediction were: internal carotid arteries, 91-100%; middle cerebral arteries, 82-98%; anterior cerebral arteries, 88-100%; posterior cerebral arteries, 87-100%; and collections of superior, anterior inferior, and posterior inferior cerebellar arteries, 90-99% in the chunk-level classification. Using a voting algorithm on the queued classified vessel factors and anatomically post-processing the automatically classified results intensified quantitative prediction performance. We employed stochastic clustering and deep neural network ensembles. Ma-chine intelligence-assisted prediction of vessel structure allowed us to personalize quantitative predictions of various types of cerebral arterial structures, contributing to precise and efficient decisions regarding the cerebrovascular disease.

Improved imaging of lingual articulation using real-time multislice MRI.

PURPOSE: To develop a real-time imaging technique that allows for simultaneous visualization of vocal tract shaping in multiple scan planes, and provides dynamic visualization of complex articulatory features. MATERIALS AND METHODS: Simultaneous imaging of multiple slices was implemented using a custom real-time imaging platform. Midsagittal, coronal, and axial scan planes of the human upper airway were prescribed and imaged in real-time using a fast spiral gradient-echo pulse sequence. Two native speakers of English produced voiceless and voiced fricatives /f/-/v/, /θ/-/ð/, /s/-/z/, /∫/- in symmetrical maximally contrastive vocalic contexts /a_a/, /i_i/, and /u_u/. Vocal tract videos were synchronized with noise-cancelled audio recordings, facilitating the selection of frames associated with production of English fricatives. RESULTS: Coronal slices intersecting the postalveolar region of the vocal tract revealed tongue grooving to be most pronounced during fricative production in back vowel contexts, and more pronounced for sibilants /s/-/z/ than for /∫/-. The axial slice best revealed differences in dorsal and pharyngeal articulation; voiced fricatives were observed to be produced with a larger cross-sectional area in the pharyngeal airway. Partial saturation of spins provided accurate location of imaging planes with respect to each other. CONCLUSION: Real-time MRI of multiple intersecting slices can provide valuable spatial and temporal information about vocal tract shaping, including details not observable from a single slice.

Estimation of Cardiac Short Axis Slice Levels with a Cascaded Deep Convolutional and Recurrent Neural Network Model.

Automatic identification of short axis slice levels in cardiac magnetic resonance imaging (MRI) is important in efficient and precise diagnosis of cardiac disease based on the geometry of the left ventricle. We developed a combined model of convolutional neural network (CNN) and recurrent neural network (RNN) that takes a series of short axis slices as input and predicts a series of slice levels as output. Each slice image was labeled as one of the following five classes: out-of-apical, apical, mid, basal, and out-of-basal levels. A variety of multi-class classification models were evaluated. When compared with the CNN-alone models, the cascaded CNN-RNN models resulted in higher mean F1-score and accuracy. In our implementation and testing of four different baseline networks with different combinations of RNN modules, MobileNet as the feature extractor cascaded with a two-layer long short-term memory (LSTM) network produced the highest scores in four of the seven evaluation metrics, i.e., five F1-scores, area under the curve (AUC), and accuracy. Our study indicates that the cascaded CNN-RNN models are superior to the CNN-alone models for the classification of short axis slice levels in cardiac cine MR images.

Flexible retrospective selection of temporal resolution in real-time speech MRI using a golden-ratio spiral view order.

In speech production research using real-time magnetic resonance imaging (MRI), the analysis of articulatory dynamics is performed retrospectively. A flexible selection of temporal resolution is highly desirable because of natural variations in speech rate and variations in the speed of different articulators. The purpose of the study is to demonstrate a first application of golden-ratio spiral temporal view order to real-time speech MRI and investigate its performance by comparison with conventional bit-reversed temporal view order. Golden-ratio view order proved to be more effective at capturing the dynamics of rapid tongue tip motion. A method for automated blockwise selection of temporal resolution is presented that enables the synthesis of a single video from multiple temporal resolution videos and potentially facilitates subsequent vocal tract shape analysis.