Faculty Bibliography

Accelerating Magnetic Resonance Imaging (MRI) by taking fewer measurements has the potential to reduce medical costs, minimize stress to patients and make MRI possible in applications where it is currently prohibitively slow or expensive. We introduce the fastMRI dataset, a large-scale collection of both raw MR measurements and clinical MR images, that can be used for training and evaluation of machine-learning approaches to MR image reconstruction. By introducing standardized evaluation criteria and a freely-accessible dataset, our goal is to help the community make rapid advances in the state of the art for MR image reconstruction. We also provide a self-contained introduction to MRI for machine learning researchers with no medical imaging background

We develop and evaluate a neural network-based method for Gibbs artifact and noise removal. A convolutional neural network (CNN) was designed for artifact removal in diffusion-weighted imaging data. Two implementations were considered: one for magnitude images and one for complex images. Both models were based on the same encoder-decoder structure and were trained by simulating MRI acquisitions on synthetic non-MRI images. Both machine learning methods were able to mitigate artifacts in diffusion-weighted images and diffusion parameter maps. The CNN for complex images was also able to reduce artifacts in partial Fourier acquisitions. The proposed CNNs extend the ability of artifact correction in diffusion MRI. The machine learning method described here can be applied on each imaging slice independently, allowing it to be used flexibly in clinical applications

BACKGROUND AND PURPOSE/OBJECTIVE:Na) MR imaging. MATERIALS AND METHODS/METHODS:Na) MR imaging using a 3T scanner. Total sodium concentration was measured in the genu, body, and splenium of the corpus callosum with 5-mm ROIs; total sodium concentration of the genu-to-splenium ratio was calculated and compared between patients and controls. RESULTS:= .001). CONCLUSIONS:Complex differences are seen in callosal total sodium concentration in symptomatic patients with mild traumatic brain injury, supporting the notion of ionic dysfunction in the pathogenesis of mild traumatic brain injury. The total sodium concentration appears to be altered beyond the immediate postinjury phase, and further work is needed to understand the relationship to persistent symptoms and outcome.

Dynamic Causal Modeling (DCM) is an advanced biophysical model which explicitly describes the entire process from experimental stimuli to functional magnetic resonance imaging (fMRI) signals via neural activity and cerebral hemodynamics. To conduct a DCM study, one needs to represent the experimental stimuli as a compact vector-valued function of time, which is hard in complex tasks such as book reading and natural movie watching. Deep learning provides the state-of-the-art signal representation solution, encoding complex signals into compact dense vectors while preserving the essence of the original signals. There is growing interest in using Recurrent Neural Networks (RNNs), a major family of deep learning techniques, in fMRI modeling. However, the generic RNNs used in existing studies work as black boxes, making the interpretation of results in a neuroscience context difficult and obscure. In this paper, we propose a new biophysically interpretable RNN built on DCM, DCM-RNN. We generalize the vanilla RNN and show that DCM can be cast faithfully as a special form of the generalized RNN. DCM-RNN uses back propagation for parameter estimation. We believe DCM-RNN is a promising tool for neuroscience. It can fit seamlessly into classical DCM studies. We demonstrate face validity of DCM-RNN in two principal applications of DCM: causal brain architecture hypotheses testing and effective connectivity estimation. We also demonstrate construct validity of DCM-RNN in an attention-visual experiment. Moreover, DCM-RNN enables end-to-end training of DCM and representation learning deep neural networks, extending DCM studies to complex tasks.

Mild traumatic brain injury is a growing public health problem with an estimated incidence of over 1.7 million people annually in US. Diagnosis is based on clinical history and symptoms, and accurate, concrete measures of injury are lacking. This work aims to directly use diffusion MR images obtained within one month of trauma to detect injury, by incorporating deep learning techniques. To overcome the challenge due to limited training data, we describe each brain region using the bag of word representation, which specifies the distribution of representative patch patterns. We apply a convolutional auto-encoder to learn the patch-level features, from overlapping image patches extracted from the MR images, to learn features from diffusion MR images of brain using an unsupervised approach. Our experimental results show that the bag of word representation using patch level features learnt by the auto encoder provides similar performance as that using the raw patch patterns, both significantly outperform earlier work relying on the mean values of MR metrics in selected brain regions.

BACKGROUND AND PURPOSE/OBJECTIVE:Cerebral microhemorrhages are a known marker of mild traumatic brain injury. Blast-related mild traumatic brain injury relates to a propagating pressure wave, and there is evidence that the mechanism of injury in blast-related mild traumatic brain injury may be different from that in blunt head trauma. Two recent reports in mixed cohorts of blunt and blast-related traumatic brain injury in military personnel suggest that the prevalence of cerebral microhemorrhages is lower than in civilian head injury. In this study, we aimed to characterize the prevalence of cerebral microhemorrhages in military service members specifically with chronic blast-related mild traumatic brain injury. MATERIALS AND METHODS/METHODS:Participants were prospectively recruited and underwent 3T MR imaging. Susceptibility-weighted images were assessed by 2 neuroradiologists independently for the presence of cerebral microhemorrhages. RESULTS:Our cohort included 146 veterans (132 men) who experienced remote blast-related mild traumatic brain injury (mean, 9.4 years; median, 9 years after injury). Twenty-one (14.4%) reported loss of consciousness for <30 minutes. Seventy-seven subjects (52.7%) had 1 episode of blast-related mild traumatic brain injury; 41 (28.1%) had 2 episodes; and 28 (19.2%) had >2 episodes. No cerebral microhemorrhages were identified in any subject, as opposed to the frequency of SWI-detectable cerebral microhemorrhages following blunt-related mild traumatic brain injury in the civilian population, which has been reported to be as high as 28% in the acute and subacute stages. CONCLUSIONS:Our results may reflect differences in pathophysiology and the mechanism of injury between blast- and blunt-related mild traumatic brain injury. Additionally, the chronicity of injury may play a role in the detection of cerebral microhemorrhages.

We seek to elucidate the underlying pathophysiology of injury sustained after mild traumatic brain injury (MTBI) using multi-shell diffusion MRI, deriving compartment-specific WM tract integrity (WMTI) metrics. WMTI allows a more biophysical interpretation of WM changes by describing microstructural characteristics in both intra- and extra-axonal environments. Thirty-two patients with MTBI within 30 days of injury and twenty-one age- and sex-matched controls were imaged on a 3T MR scanner. Multi-shell diffusion acquisition was performed with 5 b-values (250 - 2500 s/mm²) along 6 - 60 diffusion encoding directions. Tract-based spatial statistics (TBSS) was used with family-wise error (FWE) correction for multiple comparisons. TBSS results demonstrate focally lower intra-axonal diffusivity (D_axon) in MTBI patients in the splenium of the corpus callosum (sCC) (p < 0.05, FWE-corrected). The Area Under the Curve (AUC)-value for was 0.76 with low sensitivity of 46.9%, but 100% specificity. These results indicate that D_axon may be a useful imaging biomarker highly specific for MTBI-related WM injury. The observed decrease in D_axon suggests restriction of the diffusion along the axons occurring shortly after injury.