The relation between cognitive dysfunction and diffusion tensor imaging parameters in traumatic brain injury.

PURPOSE OF THE STUDY: To investigate the association among global and regional white matter fractional anisotropy (FA) values following traumatic brain injury (TBI) and cognitive functioning. MATERIALS AND METHODS: This research was conducted in an urban rehabilitation hospital. Participants included adults who were healthy controls (n = 18) or who had a TBI (n = 27). Diffusion tensor imaging using a Siemens VERIO 3T scanner and calculation of global and regional FA values were undertaken. FA values were correlated with neuropsychological test scores and injury severity variables. Logistic regression and receiver operating characteristic (ROC) curve analysis were used to investigate discriminative ability of the FA values. Neuropsychological measures, including the Symbol Digit Modalities Test (SDMT), Trail Making Test, Wechsler Test of Adult Reading, California Verbal Learning Test 2nd Edition, Digit Vigilance Test, and Wisconsin Card Sorting Test, comprised the cognitive measures. RESULTS: Within the TBI group, regional FA values were significantly lower across regions compared with controls; global FA and five brain regions were associated with SDMT scores. The FA value within the body of the corpus callosum (CC) yielded excellent discrimination between groups. CONCLUSIONS: Convergent findings support the discriminability and potential clinical utility of the CC body FA value in the context of TBI.

[Imaging changes in brain microstructural in long-term abstinent from methamphetamine-dependence].

OBJECTIVE: To explore the mechanism for changes in brain microstructure in long-term abstinent from methamphetamine-dependence by using the diffusion tensor imaging (DTI).
 Methods: A total of 26 patients with long-term abstinent methamphetamine-dependence, whose abstinence time more than 14 months, and 26 normal controls all underwent cognitive executive function tests and DTI scans. We used voxel-based analysis to compare the fractional anisotropy (FA) and mean diffusivity (MD) to obtain the abnormal brain regions of DTI parameters between the two groups. Spearman correlation analysis was used to explore the correlation between FA, MD of the brain regions with abnormal parameters and cognitive executive function tests.
 Results: There were no statistical differences in the cognitive executive function tests between the two groups (P>0.05). Compared with the normal control group, the long-term abstinent from methamphetamine-dependence group showed the decreased FA in the right precuneus, right superior frontal gyrus, right calcarine, left inferior temporal gyrus and the increased MD in the right triangular part of inferior frontal gyrus, right precuneus, right posterior cingulate, right middle temporal gyrus, bilateral middle occipital gyrus, left superior parietal lobule, and lobule VIII of cerebellar hemisphere. The MD values of the right middle temporal gyrus in the long-term abstinent group were negatively correlated with the number of completions within 60 seconds (r=-0.504) and within 120 seconds (r=-0.464) .
 Conclusion: The DTI parameters in multiple brain regions from the methamphetamine-dependence patients are still abnormal after a long-term abstinence. DTI can provide imaging evidence for brain microstructural abnormalities in long-term abstinent from methamphetamine-dependence.

[Change of brain structure imaging of long-term withdrawal of methamphetamine-dependent patients].

OBJECTIVE: To explore the characteristics of brain structure in patients with long-term withdrawal of methamphetamine-dependence.
 Methods: A total of 44 patients with withdrawal of methamphetamine-dependent for more than 14 months were recruited, who met the diagnostic criteria for substance dependence in the fifth edition of the American Mental Disorders Diagnostic and Statistical Manual (DSM-V), and 40 healthy subjects were used as the control. In addition to the general scale of drug-relevant survey, the subjects received the 3.0T magnetic resonance high-resolution scan. The voxel-based morphometric measurements for the subject's brain gray volume were conducted.
 Results: There was no significant difference in age, education, smoking and alcohol consumption between the methamphetamine-dependent withdrawal group and the control group (P>0.05). The volumes for the bilateral cerebellum, the left side of temporal gyrus and the right side of the lingual gyrus in the methamphetamine-dependent withdrawal group were increased than those in the control group. The volumes for the bilateral lingual gyrus and bilateral cuneus in the methamphetamine-dependent withdrawal group were decreased than those in the control group. The volumes of left of cuneus and cerebellum were positively correlated with the duration of abstinence.
 Conclusion: After long-term abstinence, although the patients still show abnormal brain structure, their behavior and cognitive function is improved. The cerebral nerve structural is recovered from long-term abstinence.

The connectivity domain: Analyzing resting state fMRI data using feature-based data-driven and model-based methods.

Spontaneous fluctuations of resting state functional MRI (rsfMRI) have been widely used to understand the macro-connectome of the human brain. However, these fluctuations are not synchronized among subjects, which leads to limitations and makes utilization of first-level model-based methods challenging. Considering this limitation of rsfMRI data in the time domain, we propose to transfer the spatiotemporal information of the rsfMRI data to another domain, the connectivity domain, in which each value represents the same effect across subjects. Using a set of seed networks and a connectivity index to calculate the functional connectivity for each seed network, we transform data into the connectivity domain by generating connectivity weights for each subject. Comparison of the two domains using a data-driven method suggests several advantages in analyzing data using data-driven methods in the connectivity domain over the time domain. We also demonstrate the feasibility of applying model-based methods in the connectivity domain, which offers a new pathway for the use of first-level model-based methods on rsfMRI data. The connectivity domain, furthermore, demonstrates a unique opportunity to perform first-level feature-based data-driven and model-based analyses. The connectivity domain can be constructed from any technique that identifies sets of features that are similar across subjects and can greatly help researchers in the study of macro-connectome brain function by enabling us to perform a wide range of model-based and data-driven approaches on rsfMRI data, decreasing susceptibility of analysis techniques to parameters that are not related to brain connectivity information, and evaluating both static and dynamic functional connectivity of the brain from a new perspective.

Compensation through Functional Hyperconnectivity: A Longitudinal Connectome Assessment of Mild Traumatic Brain Injury.

Mild traumatic brain injury (mTBI) is a major public health concern. Functional MRI has reported alterations in several brain networks following mTBI. However, the connectome-scale brain network changes are still unknown. In this study, sixteen mTBI patients were prospectively recruited from an emergency department and followed up at 4-6 weeks after injury. Twenty-four healthy controls were also scanned twice with the same time interval. Three hundred fifty-eight brain landmarks that preserve structural and functional correspondence of brain networks across individuals were used to investigate longitudinal brain connectivity. Network-based statistic (NBS) analysis did not find significant difference in the group-by-time interaction and time effects. However, 258 functional pairs show group differences in which mTBI patients have higher functional connectivity. Meta-analysis showed that "Action" and "Cognition" are the most affected functional domains. Categorization of connectomic signatures using multiview group-wise cluster analysis identified two patterns of functional hyperconnectivity among mTBI patients: (I) between the posterior cingulate cortex and the association areas of the brain and (II) between the occipital and the frontal lobes of the brain. Our results demonstrate that brain concussion renders connectome-scale brain network connectivity changes, and the brain tends to be hyperactivated to compensate the pathophysiological disturbances.

[Advances in clinical application of quantitative susceptibility mapping in central nervous system].

Susceptibility weighted imaging (SWI) is a magnetic resonance technology with a high resolution, three-dimensional gradient echo and fully velocity compensated sequence. It is also sensitive to substance with different magnetisability. Based on the quantitative reconstruction of SWI, quantitative susceptibility mapping (QSM) has been used in monitoring the change of intravenous iron, calcium, microbleeds and oxygen content. This article summarizes principle of QSM and the latest progress of QSM application in the diseases of central nervous system.

Traumatic white matter injury and glial activation: from basic science to clinics.

An improved understanding and characterization of glial activation and its relationship with white matter injury will likely serve as a novel treatment target to curb post injury inflammation and promote axonal remyelination after brain trauma. Traumatic brain injury (TBI) is a significant public healthcare burden and a leading cause of death and disability in the United States. Particularly, traumatic white matter (WM) injury or traumatic axonal injury has been reported as being associated with patients' poor outcomes. However, there is very limited data reporting the importance of glial activation after TBI and its interaction with WM injury. This article presents a systematic review of traumatic WM injury and the associated glial activation, from basic science to clinical diagnosis and prognosis, from advanced neuroimaging perspective. It concludes that there is a disconnection between WM injury research and the essential role of glia which serve to restore a healthy environment for axonal regeneration following WM injury. Particularly, there is a significant lack of non-invasive means to characterize the complex pathophysiology of WM injury and glial activation in both animal models and in humans. An improved understanding and characterization of the relationship between glia and WM injury will likely serve as a novel treatment target to curb post injury inflammation and promote axonal remyelination.

Anatomic Review in 3D Augmented Reality Alters Craniotomy Planning Among Residents.

OBJECTIVE: Objectively examine the effect of 3D-Augmented Reality anatomic review on craniotomy planning among neurosurgical residents as it pertains to craniotomy size, skull positioning, and knowledge of significant anatomic relationships. METHODS: Postgraduate year 1-7 neurosurgery residents were instructed to review standard 2D radiographs, pin a skull, and tailor a craniotomy for 6 different lesions and case vignettes. Participants then reviewed the lesion in a 3D-augmented reality (AR) environment, followed by repeating the craniotomy station for a variety of lesion types and locations (superficial, subcortical, deep, skull base). Quiz with case-specific anatomic and surgical questions followed by an exit survey for qualitative impressions. RESULTS: Eleven of thirteen eligible residents participated. Skull position significantly changed in 5 out of 6 cases after 3D-AR view (P < 0.05, 20° angular adjustment). No significant change in incision length or craniotomy size. Subgroup analysis of junior versus senior residents revealed that craniotomy size was significantly altered in 2 out of 6 cases. Qualitative testimonials (Likert scale 5 = strongly agree) reported a change in craniotomy approach after 3D-review (3.5), improved appreciation of anatomy (4.2), increased confidence in surgical approach (4.33 junior residents, 3.5 senior residents), smaller incision (3.5 junior residents, 1.75 senior residents), better appreciation of white matter tracts (4.6). CONCLUSIONS: The augmented reality platform offers a medium to examine surgical planning skills. Residents uniformly appreciated 3D-AR as a valuable tool for improving appreciation of critical anatomic structures and their relationship to lesional pathology. 3D-AR review significantly altered skull positioning for various lesions and craniotomy approaches, particularly among junior residents.

Extracting functional connectivity brain networks at the resting state from pulsed arterial spin labeling data.

Introduction: Functional connectivity in the brain is often studied with blood oxygenation level dependent (BOLD) resting state functional magnetic resonance imaging (rsfMRI), but the BOLD signal is several steps removed from neuronal activity. Arterial spin labeling (ASL), particularly pulsed ASL (PASL), has also the capacity to measure the blood-flow changes in response to activity. In this paper, we investigated the feasibility of extracting major brain networks from PASL data, in contrast with rsfMRI analsyis. Materials and methods: In this retrospective study, we analyzed a cohort dataset that consists of 21 mild traumatic brain injury (mTBI) patients and 29 healthy controls, which was collected in a previous study. By extracting 10 major brain networks from the data of both PASL and rsfMRI, we contrasted their similarities and differences in the 10 networks extracted from both modalities. Results: Our data demonstrated that PASL could be used to extract all 10 major brain networks. Eight out of 10 networks demonstrated over 60 % similarity to rsfMRI data. Meanwhile, there are similar but not identical changes in networks detected between mTBI patients and healthy controls with both modalities. Notably, the PASL-extracted default mode network (DMN), other than the rsfMRI-extracted DMN, includes some regions known to be associated with the DMN in other studies. It demonstrated that PASL data can be analyzed to identify resting state networks with reasonable reliability, even without rsfMRI data. Conclusion: Our analysis provides an opportunity to extract functional connectivity information in heritage datasets in which ASL but not BOLD was collected.

MRI evaluation of axonal reorganization after bone marrow stromal cell treatment of traumatic brain injury.

We treated traumatic brain injury (TBI) with human bone marrow stromal cells (hMSCs) and evaluated the effect of treatment on white matter reorganization using MRI. We subjected male Wistar rats (n = 17) to controlled cortical impact and either withheld treatment (controls; n = 9) or inserted collagen scaffolds containing hMSCs (n = 8). Six weeks later, the rats were sacrificed and MRI revealed selective migration of grafted neural progenitor cells towards the white matter reorganized boundary of the TBI-induced lesion. Histology confirmed that the white matter had been reorganized, associated with increased fractional anisotropy (FA; p < 0.01) in the recovery regions relative to the injured core region in both treated and control groups. Treatment with hMSCs increased FA in the recovery regions, lowered T(2) in the core region, decreased lesion volume and improved functional recovery relative to untreated controls. Immunoreactive staining showed axonal projections emanating from neurons and extruding from the corpus callosum into the ipsilateral cortex at the boundary of the lesion. Fiber tracking (FT) maps derived from diffusion tensor imaging confirmed the immunohistological data and provided information on axonal rewiring. The apparent kurtosis coefficient (AKC) detected additional axonal remodeling regions with crossing axons, confirmed by immunohistological staining, compared with FA. Our data demonstrate that AKC, FA, FT and T(2) can be used to evaluate treatment-induced white matter recovery, which may facilitate restorative therapy in patients with TBI.

The role of advanced MR imaging findings as biomarkers of traumatic brain injury.

Treatment of traumatic brain injury (TBI) requires proper classification of the pathophysiology. Clinical classifiers and conventional neuroimaging are limited in TBI detection, outcome prediction, and treatment guidance. Advanced magnetic resonance imaging (MRI) techniques such as susceptibility weighted imaging, diffusion tensor imaging, and magnetic resonance spectroscopic imaging are sensitive to microhemorrhages, white matter injury, and abnormal metabolic activities, respectively, in brain injury. In this article, we reviewed these 3 advanced MRI methods and their applications in TBI and report some new findings from our research. These MRI techniques have already demonstrated their potential to improve TBI detection and outcome prediction. As such, they have demonstrated the capacity of serving as a set of biomarkers to reveal the heterogeneous and complex nature of brain injury in a regional and temporal manner. Further longitudinal studies using advanced MRI in a synergistic approach are expected to provide insight in understanding TBI and imaging implications for treatment.

Optimizing accuracy of freehand cannulation of the ipsilateral ventricle for intracranial pressure monitoring in patients with brain trauma.

BACKGROUND: Intracranial pressure (ICP) monitoring in traumatic brain injury (TBI) usually requires the placement of a catheter into the ipsilateral ventricle. This surgical procedure is commonly performed via a freehand method using surface anatomical landmarks as guides. The current accuracy of the catheter placement remains relatively low and even lower among TBI patients. This study was undertaken to optimize the freehand ventricular cannulation to increase the accuracy for TBI. The authors hypothesized that an optimal surgical plan of cannulation should give an operator the greatest degrees of freedom, which could be measured as the range of operation angle, range of catheter placement depth, and size of the target area. METHODS: An imaging simulation was first performed using the computed tomography (CT) images of 47 adult patients with normal brain anatomy. On the reconstructed 3D head model, four different coronal planes of ventricular cannulation were identified: a 4-cm anterior, a 2-cm anterior, a standard (central), and a 2-cm posterior plane. The degrees of freedom during the cannulation procedure were determined, including the relevant angles, lengths of cannulation, cross-sectional area, and bounding rectangle of the lateral ventricle. Next, a retrospective assessment was performed on the CT scans of another 111 patients with TBI who underwent freehand ventricular cannulation for ICP monitoring. Postoperative measurements were also performed based on CT images to calculate the accuracy and safety of catheter placement between coronal planes in practice. RESULTS: Our simulation results showed that the 2-cm anterior plane had more extensive degrees of freedom for ventricular cannulation, in terms of length of catheter trajectory (7% longer, P<0.001), cross-sectional area of the lateral ventricle (14% larger, P=0.046), and length of the lateral ventricle (17% wider, P<0.001) than that of the standard plane, while both the 4-cm anterior and 2-cm posterior planes did not offer advantages over the standard plane in these ways. The mean length range of catheter trajectory in the 2-cm anterior plane was 41 to 58 mm. Retrospective assessment of TBI patients with ICP monitor placement also confirmed our simulation data. It showed that the accuracy of ipsilateral ventricle cannulation in the 2-cm anterior plane was 70.6%, which was a significant increase from 42.9% in the standard plane (P=0.007). CONCLUSIONS: Our imaging simulation and retrospective study demonstrate that different coronal planes could provide different degrees of freedom for cannulation, the 2-cm anterior plane has the greatest degrees of freedom in terms of larger target area and greater length range of the trajectory. The optimized surgical plan in this manner could improve cannulation accuracy and benefit a significant number of TBI patients.

Characterization of brain microstructural abnormalities in cirrhotic patients without overt hepatic encephalopathy using diffusion kurtosis imaging.

Cirrhosis is a major public health concern. However, little is known about the neurobiological mechanisms underlying brain microstructure alterations in cirrhotic patients. The purpose of this prospective study was to investigate brain microstructural alterations in cirrhosis with or without minimal hepatic encephalopathy (MHE) and their relationship with patients' neurocognitive performance and disease duration using voxel-based analysis of diffusion kurtosis imaging (DKI). DKI data were acquired from 30 cirrhotic patients with MHE, 31 patients without MHE (NMHE) and 59 healthy controls. All DKI-derived parametric maps were compared across the three groups to investigate their group differences. Correlation analyses were further performed to assess relationships between altered imaging parameters and clinical data. Voxel-based analysis of DKI data results showed that MHE/NMHE patients had increased radial diffusivity, axial diffusivity (AD) and mean diffusivity in addition to decreased axial kurtosis (AK) and fractional anisotropy of kurtosis in several regions. Compared to controls, these regions were primarily the cingulum, temporal and frontal cortices. The DKI metrics (i.e., AK and AD) were correlated with clinical variables in the two patient groups. In conclusion, DKI is useful for detecting brain microstructural abnormalities in MHE and NMHE patients. Abnormal DKI parameters suggest alterations in brain microstructural complexity in cirrhotic patients, which may contribute to the neurobiological basis of neurocognitive impairment. These results may provide additional information on the pathophysiology of cirrhosis.

A high spatial resolution 1H magnetic resonance spectroscopic imaging technique for breast cancer with a short echo time.

The high sensitivity but low specificity of breast MRI has prompted exploration of breast (1)H MRS for breast cancer detection. However, several obstacles still prevent the routine application of in vivo breast (1)H MRS, including poor spatial resolution, long acquisition time associated with conventional multi-voxel MRS imaging (MRSI) techniques, and the difficulty of "extra" lipid suppression in a magnetic field with relatively poor achievable homogeneity compared to the brain. Using a combination of a recently developed echo-filter (EF) suppression technique and an elliptical sampling scheme, we demonstrate the feasibility of overcoming these difficulties. It is robust (the suppression technique is insensitive to magnetic field inhomogeneity), fast (acquisition time of about 12 min) and offers high spatial resolution (up to 0.6 cm(3) per voxel at 1.5 T with a TE of only 60 ms). This approach should be even better at 3 T with higher resolution and/or shorter TE.

Texture analysis of magnetic resonance T1 mapping with dilated cardiomyopathy: A machine learning approach.

The diagnosis of dilated cardiomyopathy (DCM) remains a challenge in clinical radiology. This study aimed to investigate whether texture analysis (TA) parameters on magnetic resonance T1 mapping can be helpful for the diagnosis of DCM.A total of 50 DCM cases were retrospectively screened and 24 healthy controls were prospectively recruited between March 2015 and July 2017. T1 maps were acquired using the Modified Look-Locker Inversion Recovery (MOLLI) sequence at a 3.0 T MR scanner. The endocardium and epicardium were drawn on the short-axis slices of the T1 maps by an experienced radiologist. Twelve histogram parameters and 5 gray-level co-occurrence matrix (GLCM) features were extracted during the TA. Differences in texture features between DCM patients and healthy controls were evaluated by t test. Support vector machine (SVM) was used to calculate the diagnostic accuracy of those texture parameters.Most histogram features were higher in the DCM group when compared to healthy controls, and 9 of these had significant differences between the DCM group and healthy controls. In terms of GLCM features, energy, correlation, and homogeneity were higher in the DCM group, when compared with healthy controls. In addition, entropy and contrast were lower in the DCM group. Moreover, entropy, contrast, and homogeneity had significant differences between these 2 groups. The diagnostic accuracy when using the SVM classifier with all these histogram and GLCM features was 0.85 ±â€Š0.07.A computer-based TA and machine learning approach of T1 mapping can provide an objective tool for the diagnosis of DCM.

Global white matter analysis of diffusion tensor images is predictive of injury severity in traumatic brain injury.

Conventional clinical neuroimaging is insensitive to axonal injury in traumatic brain injury (TBI). Immunocytochemical staining reveals changes to axonal morphology within hours, suggesting potential for diffusion-weighted magnetic resonance (MR) in early diagnosis and management of TBI. Diffusion tensor imaging (DTI) characterizes the three-dimensional (3D) distribution of water diffusion, which is highly anisotropic in white matter fibers owing to axonal length. Recently, DTI has been used to investigate traumatic axonal injury (TAI), emphasizing regional analysis in more severe TBI. In the current study, we hypothesized that a global white matter (WM) analysis of DTI data would be sensitive to TAI across a spectrum of TBI severity and injury to scan interval. To investigate this, we compared WM-only histograms of a scalar, fractional anisotropy (FA), between 20 heterogeneous TBI patients recruited from Detroit Medical Center, including six mild TBI (GCS 13-15), and 14 healthy age-matched controls. FA histogram parameters were correlated with admission GCS and posttraumatic amnesia (PTA). In all cases, including mild TBI, patients' FA histograms were globally decreased compared with control histograms. The shape of the TBI histograms also differed from controls, being more peaked and skewed. The mean FA, kurtosis and skewness were highly correlated suggesting a common mechanism. FA histogram properties also correlated with injury severity indexed by GCS and PTA, with mean FA being the best predictor and duration of PTA (r = 0.64) being superior to GCS (r = 0.47). Therefore, in this heterogeneous sample, the FA mean accounted for 40% of the variance in PTA. Increased diffusion in the short axis dimension, likely reflecting dysmyelination and swelling of axons, accounted for most of the FA decrease. FA is globally deceased in WM, including mild TBI, possibly reflecting widespread involvement. FA changes appear to be correlated with injury severity suggesting a role in early diagnosis and prognosis of TBI.

In vivo measurement of tissue damage, oxygen saturation changes and blood flow changes after experimental traumatic brain injury in rats using susceptibility weighted imaging.

Traumatic brain injury (TBI) is a prevalent disease, and many TBI patients experience disturbed cerebral blood flow (CBF) after injury. Moreover, TBI is difficult to quantify with conventional imaging modalities. In this paper, we utilized susceptibility weighted imaging (SWI) as a means to monitor functional blood oxygenation changes and to quantify CBF changes in animals after trauma. In this study using six rats, brain trauma was induced by a weight drop model and the brain was scanned over four time points: pre trauma, and 4 h, 24 h and 48 h post trauma. Five rats survived and one died after trauma. A blood phase analysis using filtered SWI phase images suggested that three rats recovered after 48 h and two rats deteriorated. SWI also suggested that CBF decreased by up to 26%. The CBF change is in agreement with the results of arterial spin labeling methods conducted in this study and with previously published results. Furthermore, SWI revealed an enlargement of the major venous vasculature in deep brain structures, in accordance with the location of diffuse axonal injury. Compared with the traditional, invasive, clinical monitoring of cerebral vascular damage and reduction in blood flow, this method offers a novel, safe and noninvasive approach to quantify changes in oxygen saturation and CBF and to visualize structural changes in blood vasculature after TBI.

MR imaging findings in mild traumatic brain injury with persistent neurological impairment.

Traumatic brain injury (TBI) is a widespread cause of neurologic disability, with >70% of cases being mild in severity. Magnetic resonance imaging provides objective biomarkers in the diagnosis of brain injury by detecting brain lesions resulting from trauma. This paper reports on the detection rates of presumed trauma-related pathology using fluid-attenuated inversion recovery (FLAIR) and susceptibility-weighted imaging (SWI) in TBI patients with chronic, persistent symptoms. METHODS: 180 subjects with persistent neurobehavioral symptoms following head trauma referred by personal injury attorneys and 94 asymptomatic, age-matched volunteers were included in the study. 83% of TBI subjects were classified as mild. RESULTS: TBI subjects had a significantly greater number of lesions detected by FLAIR than controls (42% vs. 22%) and more lesions detected by SWI than controls (28% vs. 3%). To reduce the confounding effects of aging, we examined mild TBI subjects <45years of age, which reduced the rate of lesions detected by FLAIR (26% vs. 2%) and SWI (15% vs. 0%). This younger group, which contained few age-related lesions, also demonstrated that subcortical lesions on FLAIR are more specific for TBI than deeper lesions. CONCLUSIONS: While the presence of litigation in mild TBI cases with incomplete recovery has been associated with greater expression of symptomatology and, by extension, poorer outcomes, this study shows that mild TBI patients in litigation with chronic, persistent symptoms may have associated brain injury underlying their symptoms detectable by MRI biomarkers.

Constructing fine-granularity functional brain network atlases via deep convolutional autoencoder.

State-of-the-art functional brain network reconstruction methods such as independent component analysis (ICA) or sparse coding of whole-brain fMRI data can effectively infer many thousands of volumetric brain network maps from a large number of human brains. However, due to the variability of individual brain networks and the large scale of such networks needed for statistically meaningful group-level analysis, it is still a challenging and open problem to derive group-wise common networks as network atlases. Inspired by the superior spatial pattern description ability of the deep convolutional neural networks (CNNs), a novel deep 3D convolutional autoencoder (CAE) network is designed here to extract spatial brain network features effectively, based on which an Apache Spark enabled computational framework is developed for fast clustering of larger number of network maps into fine-granularity atlases. To evaluate this framework, 10 resting state networks (RSNs) were manually labeled from the sparsely decomposed networks of Human Connectome Project (HCP) fMRI data and 5275 network training samples were obtained, in total. Then the deep CAE models are trained by these functional networks' spatial maps, and the learned features are used to refine the original 10 RSNs into 17 network atlases that possess fine-granularity functional network patterns. Interestingly, it turned out that some manually mislabeled outliers in training networks can be corrected by the deep CAE derived features. More importantly, fine granularities of networks can be identified and they reveal unique network patterns specific to different brain task states. By further applying this method to a dataset of mild traumatic brain injury study, it shows that the technique can effectively identify abnormal small networks in brain injury patients in comparison with controls. In general, our work presents a promising deep learning and big data analysis solution for modeling functional connectomes, with fine granularities, based on fMRI data.