Combination of Radiological and Clinical Baseline Data for Outcome Prediction of Patients With an Acute Ischemic Stroke.

Background: Accurate prediction of clinical outcome is of utmost importance for choices regarding the endovascular treatment (EVT) of acute stroke. Recent studies on the prediction modeling for stroke focused mostly on clinical characteristics and radiological scores available at baseline. Radiological images are composed of millions of voxels, and a lot of information can be lost when representing this information by a single value. Therefore, in this study we aimed at developing prediction models that take into account the whole imaging data combined with clinical data available at baseline. Methods: We included 3,279 patients from the MR CLEAN Registry; a prospective, observational, multicenter registry of patients with ischemic stroke treated with EVT. We developed two approaches to combine the imaging data with the clinical data. The first approach was based on radiomics features, extracted from 70 atlas regions combined with the clinical data to train machine learning models. For the second approach, we trained 3D deep learning models using the whole images and the clinical data. Models trained with the clinical data only were compared with models trained with the combination of clinical and image data. Finally, we explored feature importance plots for the best models and identified many known variables and image features/brain regions that were relevant in the model decision process. Results: From 3,279 patients included, 1,241 (37%) patients had a good functional outcome [modified Rankin Scale (mRS) ≤ 2] and 1,954 (60%) patients had good reperfusion [modified Thrombolysis in Cerebral Infarction (eTICI) ≥ 2b]. There was no significant improvement by combining the image data to the clinical data for mRS prediction [mean area under the receiver operating characteristic (ROC) curve (AUC) of 0.81 vs. 0.80] above using the clinical data only, regardless of the approach used. Regarding predicting reperfusion, there was a significant improvement when image and clinical features were combined (mean AUC of 0.54 vs. 0.61), with the highest AUC obtained by the deep learning approach. Conclusions: The combination of radiomics and deep learning image features with clinical data significantly improved the prediction of good reperfusion. The visualization of prediction feature importance showed both known and novel clinical and imaging features with predictive values.

Predicting Poor Outcome Before Endovascular Treatment in Patients With Acute Ischemic Stroke.

Background: Although endovascular treatment (EVT) has greatly improved outcomes in acute ischemic stroke, still one third of patients die or remain severely disabled after stroke. If we could select patients with poor clinical outcome despite EVT, we could prevent futile treatment, avoid treatment complications, and further improve stroke care. We aimed to determine the accuracy of poor functional outcome prediction, defined as 90-day modified Rankin Scale (mRS) score ≥5, despite EVT treatment. Methods: We included 1,526 patients from the MR CLEAN Registry, a prospective, observational, multicenter registry of ischemic stroke patients treated with EVT. We developed machine learning prediction models using all variables available at baseline before treatment. We optimized the models for both maximizing the area under the curve (AUC), reducing the number of false positives. Results: From 1,526 patients included, 480 (31%) of patients showed poor outcome. The highest AUC was 0.81 for random forest. The highest area under the precision recall curve was 0.69 for the support vector machine. The highest achieved specificity was 95% with a sensitivity of 34% for neural networks, indicating that all models contained false positives in their predictions. From 921 mRS 0-4 patients, 27-61 (3-6%) were incorrectly classified as poor outcome. From 480 poor outcome patients in the registry, 99-163 (21-34%) were correctly identified by the models. Conclusions: All prediction models showed a high AUC. The best-performing models correctly identified 34% of the poor outcome patients at a cost of misclassifying 4% of non-poor outcome patients. Further studies are necessary to determine whether these accuracies are reproducible before implementation in clinical practice.

Predicting Outcome of Endovascular Treatment for Acute Ischemic Stroke: Potential Value of Machine Learning Algorithms.

Background: Endovascular treatment (EVT) is effective for stroke patients with a large vessel occlusion (LVO) of the anterior circulation. To further improve personalized stroke care, it is essential to accurately predict outcome after EVT. Machine learning might outperform classical prediction methods as it is capable of addressing complex interactions and non-linear relations between variables. Methods: We included patients from the Multicenter Randomized Clinical Trial of Endovascular Treatment for Acute Ischemic Stroke in the Netherlands (MR CLEAN) Registry, an observational cohort of LVO patients treated with EVT. We applied the following machine learning algorithms: Random Forests, Support Vector Machine, Neural Network, and Super Learner and compared their predictive value with classic logistic regression models using various variable selection methodologies. Outcome variables were good reperfusion (post-mTICI ≥ 2b) and functional independence (modified Rankin Scale ≤2) at 3 months using (1) only baseline variables and (2) baseline and treatment variables. Area under the ROC-curves (AUC) and difference of mean AUC between the models were assessed. Results: We included 1,383 EVT patients, with good reperfusion in 531 (38%) and functional independence in 525 (38%) patients. Machine learning and logistic regression models all performed poorly in predicting good reperfusion (range mean AUC: 0.53-0.57), and moderately in predicting 3-months functional independence (range mean AUC: 0.77-0.79) using only baseline variables. All models performed well in predicting 3-months functional independence using both baseline and treatment variables (range mean AUC: 0.88-0.91) with a negligible difference of mean AUC (0.01; 95%CI: 0.00-0.01) between best performing machine learning algorithm (Random Forests) and best performing logistic regression model (based on prior knowledge). Conclusion: In patients with LVO machine learning algorithms did not outperform logistic regression models in predicting reperfusion and 3-months functional independence after endovascular treatment. For all models at time of admission radiological outcome was more difficult to predict than clinical outcome.

Age-related differences in autism: The case of white matter microstructure.

Autism spectrum disorder (ASD) is typified as a brain connectivity disorder in which white matter abnormalities are already present early on in life. However, it is unknown if and to which extent these abnormalities are hard-wired in (older) adults with ASD and how this interacts with age-related white matter changes as observed in typical aging. The aim of this first cross-sectional study in mid- and late-aged adults with ASD was to characterize white matter microstructure and its relationship with age. We utilized diffusion tensor imaging with head motion control in 48 adults with ASD and 48 age-matched controls (30-74 years), who also completed a Flanker task. Intra-individual variability of reaction times (IIVRT) measures based on performance on the Flanker interference task were used to assess IIVRT-white matter microstructure associations. We observed primarily higher mean and radial diffusivity in white matter microstructure in ASD, particularly in long-range fibers, which persisted after taking head motion into account. Importantly, group-by-age interactions revealed higher age-related mean and radial diffusivity in ASD, in projection and association fiber tracts. Subtle dissociations were observed in IIVRT-white matter microstructure relations between groups, with the IIVRT-white matter association pattern in ASD resembling observations in cognitive aging. The observed white matter microstructure differences are lending support to the structural underconnectivity hypothesis in ASD. These reductions seem to have behavioral percussions given the atypical relationship with IIVRT. Taken together, the current results may indicate different age-related patterns of white matter microstructure in adults with ASD. Hum Brain Mapp 38:82-96, 2017. © 2016 Wiley Periodicals, Inc.

1001 Ways to run AutoDock Vina for virtual screening.

Large-scale computing technologies have enabled high-throughput virtual screening involving thousands to millions of drug candidates. It is not trivial, however, for biochemical scientists to evaluate the technical alternatives and their implications for running such large experiments. Besides experience with the molecular docking tool itself, the scientist needs to learn how to run it on high-performance computing (HPC) infrastructures, and understand the impact of the choices made. Here, we review such considerations for a specific tool, AutoDock Vina, and use experimental data to illustrate the following points: (1) an additional level of parallelization increases virtual screening throughput on a multi-core machine; (2) capturing of the random seed is not enough (though necessary) for reproducibility on heterogeneous distributed computing systems; (3) the overall time spent on the screening of a ligand library can be improved by analysis of factors affecting execution time per ligand, including number of active torsions, heavy atoms and exhaustiveness. We also illustrate differences among four common HPC infrastructures: grid, Hadoop, small cluster and multi-core (virtual machine on the cloud). Our analysis shows that these platforms are suitable for screening experiments of different sizes. These considerations can guide scientists when choosing the best computing platform and set-up for their future large virtual screening experiments.

Polyunsaturated fatty acid concentration predicts myelin integrity in early-phase psychosis.

BACKGROUND: White matter (WM) abnormalities have been implicated in schizophrenia, yet the mechanisms underlying these abnormalities are not fully understood. Several lines of evidence suggest that polyunsaturated fatty acids (PUFAs) play a role in myelination, and there is substantial evidence documenting decreased PUFA concentrations in schizophrenia. We therefore hypothesized that lower membrane PUFA concentrations may be related to reduced WM integrity in schizophrenia and related disorders. METHODS: In 30 male patients with a recent-onset psychotic disorder, erythrocyte membrane PUFA concentrations were assessed and diffusion tensor imaging was performed with voxelwise analysis. RESULTS: Lower total PUFA concentration was associated with lower fractional anisotropy (FA) throughout the corpus callosum and bilateral parietal, occipital, temporal and frontal WM (P < .05, corrected). Of the individual PUFAs, lower arachidonic acid concentration, and to a lesser extent, lower nervonic acid, linoleic acid, and docosapentaenoic acid concentration were significantly associated with lower FA. PUFA concentrations were inversely associated with radial diffusivity but showed little association with axial diffusivity. Greater severity of negative symptoms was associated with lower nervonic acid concentration and lower FA values. CONCLUSIONS: Membrane PUFA concentrations appear to be robustly related to brain WM integrity in early phase psychosis. These findings may provide a basis for studies to investigate the effects of PUFA supplementation on WM integrity and associated symptomatology in early psychosis.

Symptom validity testing in memory clinics: Hippocampal-memory associations and relevance for diagnosing mild cognitive impairment.

Patients with mild cognitive impairment (MCI) do not always convert to dementia. In such cases, abnormal neuropsychological test results may not validly reflect cognitive symptoms due to brain disease, and the usual brain-behavior relationships may be absent. This study examined symptom validity in a memory clinic sample and its effect on the associations between hippocampal volume and memory performance. Eleven of 170 consecutive patients (6.5%; 13% of patients younger than 65 years) referred to memory clinics showed noncredible performance on symptom validity tests (SVTs, viz. Word Memory Test and Test of Memory Malingering). They were compared to a demographically matched group (n = 57) selected from the remaining patients. Hippocampal volume, measured by an automated volumetric method (Freesurfer), was correlated with scores on six verbal memory tests. The median correlation was r = .49 in the matched group. However, the relation was absent (median r = -.11) in patients who failed SVTs. Memory clinic samples may include patients who show noncredible performance, which invalidates their MCI diagnosis. This underscores the importance of applying SVTs in evaluating patients with cognitive complaints that may signify a predementia stage, especially when these patients are relatively young.

Persistent and reversible consequences of combat stress on the mesofrontal circuit and cognition.

Prolonged stress can have long-lasting effects on cognition. Animal models suggest that deficits in executive functioning could result from alterations within the mesofrontal circuit. We investigated this hypothesis in soldiers before and after deployment to Afghanistan and a control group using functional and diffusion tensor imaging. Combat stress reduced midbrain activity and integrity, which was associated to compromised sustained attention. Long-term follow-up showed that the functional and structural changes had normalized within 1.5 y. In contrast, combat stress induced a persistent reduction in functional connectivity between the midbrain and prefrontal cortex. These results demonstrate that combat stress has adverse effects on the human mesofrontal circuit and suggests that these alterations are partially reversible.

Legal constraints on genetic data processing in European grids.

European laws on privacy and data security are not explicit about the storage and processing of genetic data. Especially whole-genome data is identifying and contains a lot of personal information. Is processing of such data allowed in computing grids? To find out, we looked at legal precedents in related fields, current literature, and interviews with legal experts. We found that processing of genetic data is only allowed on distributed systems with specific security measures, both technical and organizational. Informed consent, although important, offers no substitute for such requirements.

Provenance for distributed biomedical workflow execution.

Scientific research has become very data and compute intensive because of the progress in data acquisition and measurement devices, which is particularly true in Life Sciences. To cope with this deluge of data, scientists use distributed computing and storage infrastructures. The use of such infrastructures introduces by itself new challenges to the scientists in terms of proper and efficient use. Scientific workflow management systems play an important role in facilitating the use of the infrastructure by hiding some of its complexity. Although most scientific workflow management systems are provenance-aware, not all of them come with provenance functionality out of the box. In this paper we describe the improvement and integration of a provenance system into an e-infrastructure for biomedical research based on the MOTEUR workflow management system. The main contributions of the paper are: presenting an OPM implementation using relational database backend for the provenance store, providing an e-infrastructure with a comprehensive provenance system, defining a generic approach to provenance implementation, potentially suitable for other workflow systems and application domains and demonstrating the value of this system based on use cases presenting the provenance data through a user-friendly web interface.

Integrated support for neuroscience research: from study design to publication.

Computational neuroscience is a new field of research in which neurodegenerative diseases are studied with the aid of new imaging techniques and computation facilities. Researchers with different expertise collaborate in these studies. A study requires scalable computational and storage capacity and information management facilities to succeed. Many virtual laboratories are proposed and developed to facilitate these studies, however most of them cover only the parts related to the computational data processing. In this paper we describe and analyse the phases of the computational neuroscience studies including the actors, the tasks they perform, and the characteristics of each phase. Based on these we identify the required properties and functionalities of a virtual laboratory that supports the actors and their tasks throughout the complete study.

Performance of VIDISCA-454 in feces-suspensions and serum.

Virus discovery combining sequence unbiased amplification with next generation sequencing is now state-of-the-art. We have previously determined that the performance of the unbiased amplification technique which is operational at our institute, VIDISCA-454, is efficient when respiratory samples are used as input. The performance of the assay is, however, not known for other clinical materials like blood or stool samples. Here, we investigated the sensitivity of VIDISCA-454 with feces-suspensions and serum samples that are positive and that have been quantified for norovirus and human immunodeficiency virus type 1, respectively. The performance of VIDISCA-454 in serum samples was equal to its performance in respiratory material, with an estimated lower threshold of 1,000 viral genome copies. The estimated threshold in feces-suspension is around 200,000 viral genome copies. The decreased sensitivity in feces suspension is mainly due to sequences that share no recognizable identity with known sequences. Most likely these sequences originate from bacteria and phages which are not completely sequenced.

Initial steps towards a production platform for DNA sequence analysis on the grid.

BACKGROUND: Bioinformatics is confronted with a new data explosion due to the availability of high throughput DNA sequencers. Data storage and analysis becomes a problem on local servers, and therefore it is needed to switch to other IT infrastructures. Grid and workflow technology can help to handle the data more efficiently, as well as facilitate collaborations. However, interfaces to grids are often unfriendly to novice users. RESULTS: In this study we reused a platform that was developed in the VL-e project for the analysis of medical images. Data transfer, workflow execution and job monitoring are operated from one graphical interface. We developed workflows for two sequence alignment tools (BLAST and BLAT) as a proof of concept. The analysis time was significantly reduced. All workflows and executables are available for the members of the Dutch Life Science Grid and the VL-e Medical virtual organizations All components are open source and can be transported to other grid infrastructures. CONCLUSIONS: The availability of in-house expertise and tools facilitates the usage of grid resources by new users. Our first results indicate that this is a practical, powerful and scalable solution to address the capacity and collaboration issues raised by the deployment of next generation sequencers. We currently adopt this methodology on a daily basis for DNA sequencing and other applications. More information and source code is available via http://www.bioinformaticslaboratory.nl/

Estimation of diffusion properties in crossing fiber bundles.

There is an ongoing debate on how to model diffusivity in fiber crossings. We propose an optimization framework for the selection of a dual tensor model and the set of diffusion weighting parameters b, such that both the diffusion shape and orientation parameters can be precisely as well as accurately estimated. For that, we have adopted the Cramér-Rao lower bound (CRLB) on the variance of the model parameters, and performed Monte Carlo simulations. We have found that the axial diffusion lambda(parallel) needs to be constrained, while an isotropic fraction can be modeled by a single parameter f(iso). Under these circumstances, the Fractional Anisotropy (FA) of both tensors can theoretically be independently estimated with a precision of 9% (at SNR = 25). Levenberg-Marquardt optimization of the Maximum Likelihood function with a Rician noise model approached this precision while the bias was insignificant. A two-element b-vector b = [1.0 3.5] x 10(3) mm(-2)s was found to be sufficient for estimating parameters of heterogeneous tissue with low error. This has allowed us to estimate consistent FA-profiles along crossing tracts. This work defines fundamental limits for comparative studies to correctly analyze crossing white matter structures.

Hitting the ground running: Healthgrid deployment and adoption.

We consider the issues of healthgrid development, deployment and adoption in health care and research environments. While healthgrid technology could be deployed to support advanced medical research, we are not seeing its wide adoption. Understanding why this technology is not being exploited is one purpose of this paper. We do so in light of the seminal Healthgrid White Paper and the SHARE roadmap. We also address barriers to adoption and successes by presenting experiences in North America and Europe. By critically appraising where we are, we hope that we can hit the ground running in the near future.

A virtual laboratory for medical image analysis.

This paper presents the design, implementation, and usage of a virtual laboratory for medical image analysis. It is fully based on the Dutch grid, which is part of the Enabling Grids for E-sciencE (EGEE) production infrastructure and driven by the gLite middleware. The adopted service-oriented architecture enables decoupling the user-friendly clients running on the user's workstation from the complexity of the grid applications and infrastructure. Data are stored on grid resources and can be browsed/viewed interactively by the user with the Virtual Resource Browser (VBrowser). Data analysis pipelines are described as Scufl workflows and enacted on the grid infrastructure transparently using the MOTEUR workflow management system. VBrowser plug-ins allow for easy experiment monitoring and error detection. Because of the strict compliance to the grid authentication model, all operations are performed on behalf of the user, ensuring basic security and facilitating collaboration across organizations. The system has been operational and in daily use for eight months (December 2008), with six users, leading to the submission of 9000 jobs/month in average and the production of several terabytes of data.

Accelerating regional atrophy rates in the progression from normal aging to Alzheimer's disease.

We investigated progression of atrophy in vivo, in Alzheimer's disease (AD), and mild cognitive impairment (MCI). We included 64 patients with AD, 44 with MCI and 34 controls with serial MRI examinations (interval 1.8 ± 0.7 years). A nonlinear registration algorithm (fluid) was used to calculate atrophy rates in six regions: frontal, medial temporal, temporal (extramedial), parietal, occipital lobes and insular cortex. In MCI, the highest atrophy rate was observed in the medial temporal lobe, comparable with AD. AD patients showed even higher atrophy rates in the extramedial temporal lobe. Additionally, atrophy rates in frontal, parietal and occipital lobes were increased. Cox proportional hazard models showed that all regional atrophy rates predicted conversion to AD. Hazard ratios varied between 2.6 (95% confidence interval (CI) = 1.1-6.2) for occipital atrophy and 15.8 (95% CI = 3.5-71.8) for medial temporal lobe atrophy. In conclusion, atrophy spreads through the brain with development of AD. MCI is marked by temporal lobe atrophy. In AD, atrophy rate in the extramedial temporal lobe was even higher. Moreover, atrophy rates also accelerated in parietal, frontal, insular and occipital lobes. Finally, in nondemented elderly, medial temporal lobe atrophy was most predictive of progression to AD, demonstrating the involvement of this region in the development of AD.

Neurotoxic effects of ecstasy on the thalamus.

BACKGROUND: Neurotoxic effects of ecstasy have been reported, although it remains unclear whether effects can be attributed to ecstasy, other recreational drugs or a combination of these. AIMS: To assess specific/independent neurotoxic effects of heavy ecstasy use and contributions of amphetamine, cocaine and cannabis as part of The Netherlands XTC Toxicity (NeXT) study. METHOD: Effects of ecstasy and other substances were assessed with (1)H-magnetic resonance spectroscopy, diffusion tensor imaging, perfusion weighted imaging and [(123)I]2beta-carbomethoxy-3beta-(4-iodophenyl)-tropane ([(123)I]beta-CIT) single photon emission computed tomography (serotonin transporters) in a sample (n=71) with broad variation in drug use, using multiple regression analyses. RESULTS: Ecstasy showed specific effects in the thalamus with decreased [(123)I]beta-CIT binding, suggesting serotonergic axonal damage; decreased fractional anisotropy, suggesting axonal loss; and increased cerebral blood volume probably caused by serotonin depletion. Ecstasy had no effect on brain metabolites and apparent diffusion coefficients. CONCLUSIONS: Converging evidence was found for a specific toxic effect of ecstasy on serotonergic axons in the thalamus.

Sustained effects of ecstasy on the human brain: a prospective neuroimaging study in novel users.

Previous studies have suggested toxic effects of recreational ecstasy use on the serotonin system of the brain. However, it cannot be excluded that observed differences between users and non-users are the cause rather than the consequence of ecstasy use. As part of the Netherlands XTC Toxicity (NeXT) study, we prospectively assessed sustained effects of ecstasy use on the brain in novel ecstasy users using repeated measurements with a combination of different neuroimaging parameters of neurotoxicity. At baseline, 188 ecstasy-naive volunteers with high probability of first ecstasy use were examined. After a mean period of 17 months follow-up, neuroimaging was repeated in 59 incident ecstasy users and 56 matched persistent ecstasy-naives and their outcomes were compared. Neuroimaging included [(123)I]beta-carbomethoxy-3beta-(4-iodophenyl)tropane (CIT) SPECT to measure serotonin transporter densities as indicators of serotonergic function; (1)H-MR spectroscopy ((1)H-MRS) to measure brain metabolites as indicators of neuronal damage; diffusion tensor imaging (DTI) to measure the apparent diffusion coefficient and fractional anisotropy (FA) of the diffusional motion of water molecules in the brain as indicators of axonal integrity; and perfusion weighted imaging (PWI) to measure regional relative cerebral blood volume (rrCBV) which indicates brain perfusion. With this approach, both structural ((1)H-MRS and DTI) and functional ([(123)I]beta-CIT SPECT and PWI) aspects of neurotoxicity were combined. Compared to persistent ecstasy-naives, novel low-dose ecstasy users (mean 6.0, median 2.0 tablets) showed decreased rrCBV in the globus pallidus and putamen; decreased FA in thalamus and frontoparietal white matter; increased FA in globus pallidus; and increased apparent diffusion coefficient in the thalamus. No changes in serotonin transporter densities and brain metabolites were observed. These findings suggest sustained effects of ecstasy on brain microvasculature, white matter maturation and possibly axonal damage due to low dosages of ecstasy. Although we do not know yet whether these effects are reversible or not, we cannot exclude that ecstasy even in low doses is neurotoxic to the brain.

From "low hanging" to "user ready": initial steps into a HealthGrid.

Grids offer powerful infrastructures and promising concepts for the development and deployment of advanced applications in medical research and healthcare. The construction of HealthGrids in practice, however, is challenging due to reasons of scientific, technical, and cultural nature, among them the large gap between communities that develop and use the technology. Whereas grid developments focus mostly on functionality, usability issues are also very important to enable the potential of grids to be fully exploited by those who could mostly benefit from it, the end-users. In this paper we make a retrospective of our efforts to develop the Virtual Lab for functional Magnetic Resonance Imaging (fMRI). This project aims at providing for the end-users a grid-based system to facilitate research and clinical usage of fMRI data for study of brain activation. We present the evolution of this project in three phases coined "low hanging fruit", "trying out" and "end-user ready", and the lessons learnt in each one. The evolution of the software architecture, which had a large impact on the user front-end, is discussed in more detail. The current architecture facilitates the construction of front-ends that enable users to access the grid infrastructure from a single user-friendly GUI. All (local and grid) resources are accessed directly by the users from a virtual desktop implemented by the Virtual Resource Browser (VBrowser).