Spatial regularization of SVM for the detection of diffusion alterations associated with stroke outcome

Artificial intelligence reinforced upconversion nanoparticle-based lateral flow assay via transfer learning

The combination of upconverting nanoparticles (UCNPs) and immunochromatography has become a widely used and promising new detection technique for point-of-care testing (POCT). However, their low luminescence efficiency, non-specific adsorption, and image noise have always limited their progress toward practical applications. Recently, artificial intelligence (AI) has demonstrated powerful representational learning and generalization capabilities in computer vision. We report for the first time a combination of AI and upconversion nanoparticle-based lateral flow assays (UCNP-LFAs) for the quantitative detection of commercial internet of things (IoT) devices. This universal UCNPs quantitative detection strategy combines high accuracy, sensitivity, and applicability in the field detection environment. By using transfer learning to train AI models in a small self-built database, we not only significantly improved the accuracy and robustness of quantitative detection, but also efficiently solved the actual problems of data scarcity and low computing power of POCT equipment. Then, the trained AI model was deployed in IoT devices, whereby the detection process does not require detailed data preprocessing to achieve real-time inference of quantitative results. We validated the quantitative detection of two detectors using eight transfer learning models on a small dataset. The AI quickly provided ultra-high accuracy prediction results (some models could reach 100% accuracy) even when strong noise was added. Simultaneously, the high flexibility of this strategy promises to be a general quantitative detection method for optical biosensors. We believe that this strategy and device have a scientific significance in revolutionizing the existing POCT technology landscape and providing excellent commercial value in the in vitro diagnostics (IVD) industry.

Interpretable deep learning for the prognosis of long-term functional outcome post-stroke using acute diffusion weighted imaging

Advances in deep learning can be applied to acute stroke imaging to build powerful and explainable prediction models that could supersede traditionally used biomarkers. We aimed to evaluate the performance and interpretability of a deep learning model based on convolutional neural networks (CNN) in predicting long-term functional outcome with diffusion-weighted imaging (DWI) acquired at day 1 post-stroke. Ischemic stroke patients (n = 322) were included from the ASTER and INSULINFARCT trials as well as the Pitié-Salpêtrière registry. We trained a CNN to predict long-term functional outcome assessed at 3 months with the modified Rankin Scale (dichotomized as good [mRS ≤ 2] vs. poor [mRS ≥ 3]) and compared its performance to two logistic regression models using lesion volume and ASPECTS. The CNN contained an attention mechanism, which allowed to visualize the areas of the brain that drove prediction. The deep learning model yielded a significantly higher area under the curve (0.83 95%CI [0.78-0.87]) than lesion volume (0.78 [0.73-0.83]) and ASPECTS (0.77 [0.71-0.83]) (p < 0.05). Setting all classifiers to the specificity as the deep learning model (i.e., 0.87 [0.82-0.92]), the CNN yielded a significantly higher sensitivity (0.67 [0.59-0.73]) than lesion volume (0.48 [0.40-0.56]) and ASPECTS (0.50 [0.41-0.58]) (p = 0.002). The attention mechanism revealed that the network learned to naturally attend to the lesion to predict outcome.

Smaller baseline subcortical infarct volume predicts good outcomes in patients with a large core in early acute ischemic stroke after endovascular treatment

Background

Patients with acute ischemic stroke (AIS) and a large core may benefit from endovascular treatment (EVT) in the early time window.

Purpose

To examine the prognostic factors for good outcomes in patients with a large core (70-130 ml) after EVT.

Materials and methods

We retrospectively reviewed 40 patients who met the criteria from October 2019 to April 2021. Based on the modified Rankin Score (mRS) at 90 days, the patients were divided into a good outcome group (mRS 0-2) and a poor outcome group (mRS 3-6). Baseline and procedural characteristics were collected for unilateral and multivariate regression analyses to explore the factors that influence good outcomes. In particular, the infarct territories were quantified as subcortical infarct volume (SIV) and cortical infarct volume (CIV).

Results

Of the 40 patients included, good outcomes were observed in 11 (27.5%) patients. Younger age, smaller SIV and larger mismatch volume were noted in the good outcome group than in the poor outcome group (all P < 0.05). Multivariate logistic regression analysis showed that only a smaller SIV [odds ratio (OR) 0.801; 95% CI 0.644-0.996; P = 0.046] was an independent predictor for good outcomes. The receiver operating characteristic curve indicated a moderate value of SIV for predicting good outcomes, with an area under the receiver operating characteristic curve of 0.735 (95% CI 0.572-0.862; P = 0.007).

Conclusion

Subcortical infarct volume was a potential useful predictor of good outcomes in patients with a large core after EVT in the early time window.

Using natural language processing to automatically classify written self-reported narratives by patients with migraine or cluster headache

Background

Headache medicine is largely based on detailed history taking by physicians analysing patients’ descriptions of headache. Natural language processing (NLP) structures and processes linguistic data into quantifiable units. In this study, we apply these digital techniques on self-reported narratives by patients with headache disorders to research the potential of analysing and automatically classifying human-generated text and information extraction in clinical contexts.

Methods

A prospective cross-sectional clinical trial collected self-reported narratives on headache disorders from participants with either migraine or cluster headache. NLP was applied for the analysis of lexical, semantic and thematic properties of the texts. Machine learning (ML) algorithms were applied to classify the descriptions of headache attacks from individual participants into their correct group (migraine versus cluster headache).

Results

One-hundred and twenty-one patients (81 participants with migraine and 40 participants with cluster headache) provided a self-reported narrative on their headache disorder. Lexical analysis of this text corpus resulted in several specific key words per diagnostic group (cluster headache: Dutch (nl): “oog” | English (en): “eye”, nl: “pijn” | en: “pain” and nl: “terug” | en: “back/to come back”; migraine: nl: “hoofdpijn” | en: “headache”, nl: “stress” | en: “stress” and nl: “misselijkheid” | en: “nausea”). Thematic and sentiment analysis of text revealed largely negative sentiment in texts by both patients with migraine and cluster headache. Logistic regression and support vector machine algorithms with different feature groups performed best for the classification of attack descriptions (with F1-scores for detecting cluster headache varying between 0.82 and 0.86) compared to naïve Bayes classifiers.

Conclusions

Differences in lexical choices between patients with migraine and cluster headache are detected with NLP and are congruent with domain expert knowledge of the disorders. Our research shows that ML algorithms have potential to classify patients’ self-reported narratives of migraine or cluster headache with good performance. NLP shows its capability to discern relevant linguistic aspects in narratives from patients with different headache disorders and demonstrates relevance in clinical information extraction. The potential benefits on the classification performance of larger datasets and neural NLP methods can be investigated in the future.

Trial registration

This study was registered with clinicaltrials.gov with ID NCT05377437.

Supplementary Information

The online version contains supplementary material available at 10.1186/s10194-022-01490-0.

Metabolic detection of malignant brain gliomas through plasma lipidomic analysis and support vector machine-based machine learning

Background

Most malignant brain gliomas (MBGs) are associated with dismal outcomes, mainly due to their late diagnosis. Current diagnostic methods for MBGs are based on imaging and histological examination, which limits their early detection. Here, we aimed to identify reliable plasma lipid biomarkers for non-invasive diagnosis for MBGs.

Methods

Untargeted lipidomic analysis was firstly performed using a discovery cohort (n=107). The data were processed by a support vector machine (SVM)-based discriminating model to retrieve a panel of candidate biomarkers. Then, a targeted quantification method was developed, and the SVM-based diagnostic model was constructed using a training cohort (n=750) and tested using a test cohort (n=225). Finally, the performance of the diagnostic model was further evaluated in an independent validation cohort (n=920) enrolled from multiple medical centers.

Findings

A panel of 11 plasma lipids was identified as candidate biomarkers with an accuracy of 0.999. The diagnostic model developed achieved a high performance in distinguishing MBGs patients from normal controls with an area under the receiver-operating characteristic curve (AUC) of 0.9877 and 0.9869 in the training and test cohorts, respectively. In the validation cohort, the 11 lipid panel still achieved an accuracy of 0.9641 and an AUC of 0.9866.

Interpretation

The present study demonstrates the applicability and robustness of utilizing a machine learning algorithm to analyze lipidomic data for efficient and reliable biomarker screening. The 11 lipid biomarkers show great potential for the non-invasive diagnosis of MBGs with high throughput.

Funding

A full list of funding bodies that contributed to this study can be found in the Acknowledgments section.

A combined model based on feature selection and support vector machine for PM2.5 prediction

With the increasing attention to the environment and air quality, PM2.5 has been paid more and more attention. It is expected to excavate useful information in meteorological data to predict air pollution, however, the air quality is greatly affected by meteorological factors, and how to establish an effective air quality prediction model has always been a problem that people urgently need to solve. This paper proposed a combined model based on feature selection and Support Vector Machine (SVM) for PM2.5 prediction. Firstly, aiming at the influence of meteorological factors on PM2.5, a feature selection method based on linear causality is proposed to find out the causality between features and select the features with strong causality, so as to remove the redundant features in air pollution data and reduce the workload of data analysis. Then, a method based on SVM is proposed to analyze and solve the nonlinear problems in the data, for reducing the prediction error, a method of particle swarm optimization is also used to optimize SVM parameters. Finally, the above methods are combined into a prediction model, which is suitable for the current air pollution control. 12 representative data sets on the UCI (University of California, Irvine) website are used to verify the combined model, and the experimental results show that the model is feasible and effective.

An ANN Model for Predicting the Compressive Strength of Concrete

An artificial neural network (ANN) model for predicting the compressive strength of concrete is established in this study. The Back Propagation (BP) network with one hidden layer is chosen as the structure of the ANN. The database of real concrete mix proportioning listed in earlier research by another author is used for training and testing the ANN. The proper number of neurons in the hidden layer is determined by checking the features of over-fitting while the synaptic weights and the thresholds are finalized by checking the features of over-training. After that, we use experimental data from other papers to verify and validate our ANN model. The final result of the synaptic weights and the thresholds in the ANN are all listed. Therefore, with them, and using the formulae expressed in this article, anyone can predict the compressive strength of concrete according to the mix proportioning on his/her own.

Support vector machine based aphasia classification of transcranial magnetic stimulation language mapping in brain tumor patients

Repetitive TMS (rTMS) allows for non-invasive and transient disruption of local neuronal functioning. We used machine learning approaches to assess whether brain tumor patients can be accurately classified into aphasic and non-aphasic groups using their rTMS language mapping results as input features. Given that each tumor affects the subject-specific language networks differently, resulting in heterogenous rTMS functional mappings, we propose the use of machine learning strategies to classify potential patterns of rTMS language mapping results. We retrospectively included 90 patients with left perisylvian world health organization (WHO) grade II-IV gliomas that underwent presurgical navigated rTMS language mapping. Within our cohort, 29 of 90 (32.2%) patients suffered from at least mild aphasia as shown in the Aachen Aphasia Test based Berlin Aphasia Score (BAS). After spatial normalization to MNI 152 of all rTMS spots, we calculated the error rate (ER) in each stimulated cortical area (28 regions of interest, ROI) by automated anatomical labeling parcellation (AAL3) and IIT. We used a support vector machine (SVM) to classify significant areas in relation to aphasia. After feeding the ROIs into the SVM model, it revealed that in addition to age (w = 2.98), the ERs of the left supramarginal gyrus (w = 3.64), left inferior parietal gyrus (w = 2.28) and right pars triangularis (w = 1.34) contributed more than other features to the model. The model's sensitivity was 86.2%, the specificity was 82.0%, the overall accuracy was 85.5% and the AUC was 89.3%. Our results demonstrate an increased vulnerability of right inferior pars triangularis to rTMS in aphasic patients due to left perisylvian gliomas. This finding points towards a functional relevant involvement of the right pars triangularis in response to aphasia. The tumor location feature, specified by calculating overlaps with white and grey matter atlases, did not affect the SVM model. The left supramarginal gyrus as a feature improved our SVM model the most. Additionally, our results could point towards a decreasing potential for neuroplasticity with age.

Localized cytokine responses to total knee arthroplasty and total knee revision complications

Background

The study of localized immune-related factors has proven beneficial for a variety of conditions, and one area of interest in the field of orthopaedics is the impact of implants and localized infections on immune response. Several cytokines have shown increased systemic concentrations (in serum/plasma) in response to implants and infection, but tissue-level cytokines have not been investigated as thoroughly.

Methods

This exploratory study investigated tissue-level cytokines in a cohort of patients (N = 17) in response to total knee arthroplasty and total knee revision to better understand the immune response to implants and localized infection (e.g., prosthetic joint infection). The overall goal of this study was to provide insight into the localized cytokine response of tissues and identify tissue-level markers specific to inflammation caused by implants vs. inflammation caused by infection. Tissues were collected across several anatomical locations and assayed with a panel of 20 human inflammatory cytokines to understand spatial differences in cytokine levels.

Results

In this study, six cytokines were elevated in implanted joints, as compared to native joints: IL-10, IL-12p70, IL-13, IL-17A, IL-4, and TNF-α (p < 0.05). Seven cytokines showed infection-dependent increases in localized tissues: IL-1α, IL-1β, IL-6, IL-8, MCP-1, MIP-1α, and MIP-1β (p < 0.05).

Conclusions

This study demonstrated that differences exist in tissue-level cytokines in response to presence of implant, and some cytokines were specifically elevated for infection; these responses may be informative of overall tissue health. These results highlight the utility of investigating localized cytokine concentrations to offer novel insights for total knee arthroplasty and total knee revision procedures, as well as their complications. Ultimately, this information could provide additional, quantitative measurements of tissue to aid clinical decision making and patient treatment options.

A Survey for Predicting Enzyme Family Classes Using Machine Learning Methods

Enzymes are proteins that act as biological catalysts to speed up cellular biochemical processes. According to their main Enzyme Commission (EC) numbers, enzymes are divided into six categories: EC-1: oxidoreductase; EC-2: transferase; EC-3: hydrolase; EC-4: lyase; EC-5: isomerase and EC-6: synthetase. Different enzymes have different biological functions and acting objects. Therefore, knowing which family an enzyme belongs to can help infer its catalytic mechanism and provide information about the relevant biological function. With the large amount of protein sequences influxing into databanks in the post-genomics age, the annotation of the family for an enzyme is very important. Since the experimental methods are cost ineffective, bioinformatics tool will be a great help for accurately classifying the family of the enzymes. In this review, we summarized the application of machine learning methods in the prediction of enzyme family from different aspects. We hope that this review will provide insights and inspirations for the researches on enzyme family classification.

Brain-Machine Interface Induced Morpho-Functional Remodeling of the Neural Motor System in Severe Chronic Stroke

Brain-machine interfaces (BMI) permit bypass motor system disruption by coupling contingent neuroelectric signals related to motor activity with prosthetic devices that enhance afferent and proprioceptive feedback to the somatosensory cortex. In this study, we investigated neural plasticity in the motor network of severely impaired chronic stroke patients after an EEG-BMI-based treatment reinforcing sensorimotor contingency of ipsilesional motor commands. Our structural connectivity analysis revealed decreased fractional anisotropy in the splenium and body of the corpus callosum, and in the contralesional hemisphere in the posterior limb of the internal capsule, the posterior thalamic radiation, and the superior corona radiata. Functional connectivity analysis showed decreased negative interhemispheric coupling between contralesional and ipsilesional sensorimotor regions, and decreased positive intrahemispheric coupling among contralesional sensorimotor regions. These findings indicate that BMI reinforcing ipsilesional brain activity and enhancing proprioceptive function of the affected hand elicits reorganization of contralesional and ipsilesional somatosensory and motor-assemblies as well as afferent and efferent connection-related motor circuits that support the partial re-establishment of the original neurophysiology of the motor system even in severe chronic stroke.

Resting-State Functional Network Scale Effects and Statistical Significance-Based Feature Selection in Machine Learning Classification

In recent years, functional brain network topological features have been widely used as classification features. Previous studies have found that network node scale differences caused by different network parcellation definitions significantly affect the structure of the constructed network and its topological properties. However, we still do not know how network scale differences affect the classification accuracy, performance of classification features, and effectiveness of the feature selection strategy using P values in terms of the machine learning method. This study used five scale parcellations, involving 90, 256, 497, 1003, and 1501 nodes. Three local properties of resting-state functional brain networks were selected (degree, betweenness centrality, and nodal efficiency), and the support vector machine method was used to construct classifiers to identify patients with major depressive disorder. We analyzed the impact of the five scales on classification accuracy. In addition, the effectiveness and redundancy of features obtained by the different scale parcellations were compared. Finally, traditional statistical significance (P value) was verified as a feature selection criterion. The results showed that the feature effectiveness of different scales was similar; in other words, parcellation with more regions did not provide more effective discriminative features. Nevertheless, parcellation with more regions did provide a greater quantity of discriminative features, which led to an improvement in the accuracy of the classification. However, due to the close distance between brain regions, the redundancy of parcellation with more regions was also greater. The traditional P value feature selection strategy is feasible with different scales, but our analysis showed that the traditional P < 0.05 threshold was too strict for feature selection. This study provides an important reference for the selection of network scales when applying topological properties of brain networks to machine learning methods.

Multi-Modality Sparse Representation for Alzheimer's Disease Classification

BACKGROUND

Alzheimer's disease (AD) and mild cognitive impairment (MCI) are age-related neurodegenerative diseases characterized by progressive loss of memory and irreversible cognitive functions. The hippocampus, a brain area critical for learning and memory processes, is especially susceptible to damage at early stages of AD.

OBJECTIVE

We aimed to develop prediction model using a multi-modality sparse representation approach.

METHODS

We proposed a sparse representation approach to the hippocampus using structural T1-weighted magnetic resonance imaging (MRI) and 18-fluorodeoxyglucose-positron emission tomography (FDG-PET) to distinguish AD/MCI from healthy control subjects (HCs). We considered structural and function information for the hippocampus and applied a sparse patch-based approach to effectively reduce the dimensions of neuroimaging biomarkers.

RESULTS

In experiments using Alzheimer's Disease Neuroimaging Initiative data, our proposed method demonstrated more reliable than previous classification studies. The effects of different parameters on segmentation accuracy were also evaluated. The mean classification accuracy obtained with our proposed method was 0.94 for AD/HCs, 0.82 for MCI/HCs, and 0.86 for AD/MCI.

CONCLUSION

We extracted multi-modal features from automatically defined hippocampal regions of training subjects and found this method to be discriminative and robust for AD and MCI classification. The extraction of features in T1 and FDG-PET images is expected to improve classification performance due to the relationship between brain structure and function.

Multivariate prediction of functional outcome using lesion topography characterized by acute diffusion tensor imaging

The relationship between stroke topography and functional outcome has largely been studied with binary manual lesion segmentations. However, stroke topography may be better characterized by continuous variables capable of reflecting the severity of ischemia, which may be more pertinent for long-term outcome. Diffusion Tensor Imaging (DTI) constitutes a powerful means of quantifying the degree of acute ischemia and its potential relation to functional outcome. Our aim was to investigate whether using more clinically pertinent imaging parameters with powerful machine learning techniques could improve prediction models and thus provide valuable insight on critical brain areas important for long-term outcome. Eighty-seven thrombolyzed patients underwent a DTI sequence at 24 h post-stroke. Functional outcome was evaluated at 3 months post-stroke with the modified Rankin Score and was dichotomized into good (mRS ≤ 2) and poor (mRS > 2) outcome. We used support vector machines (SVM) to classify patients into good vs. poor outcome and evaluate the accuracy of different models built with fractional anisotropy, mean diffusivity, axial diffusivity, radial diffusivity asymmetry maps, and lesion segmentations in combination with lesion volume, age, recanalization status, and thrombectomy treatment. SVM classifiers built with axial diffusivity maps yielded the best accuracy of all imaging parameters (median [IQR] accuracy = 82.8 [79.3-86.2]%), compared to that of lesion segmentations (76.7 [73.3-82.8]%) when predicting 3-month functional outcome. The analysis revealed a strong contribution of clinical variables, notably - in descending order - lesion volume, thrombectomy treatment, and recanalization status, in addition to the deep white matter at the crossroads of major white matter tracts, represented by brain regions where model weights were highest. Axial diffusivity is a more appropriate imaging marker to characterize stroke topography for predicting long-term outcome than binary lesion segmentations.

Impact of Ischemic Lesion Location on the mRS Score in Patients with Ischemic Stroke: A Voxel-Based Approach

BACKGROUND AND PURPOSE

Previous studies indicated that ischemic lesion volume might be a useful surrogate marker for functional outcome in ischemic stroke but should be considered in the context of lesion location. In contrast to previous studies using the ROI approach, which has several drawbacks, the present study aimed to measure the impact of ischemic lesion location on functional outcome using a more precise voxelwise approach.

MATERIALS AND METHODS

Datasets of patients with acute ischemic strokes from the Multicenter Randomized Clinical Trial of Endovascular Therapy for Acute Ischemic Stroke in the Netherlands (MR CLEAN) were used. Primary outcome was functional outcome as assessed by the modified Rankin Scale 3 months after stroke. Ischemic lesion volume was determined on CT scans 3-9 days after stroke. Voxel-based lesion-symptom mapping techniques, including covariates that are known to be associated with functional outcome, were used to determine the impact of ischemic lesion location for outcome.

RESULTS

Of the 500 patients in the MR CLEAN trial, 216 were included for analysis. The mean age was 63 years. Lesion-symptom mapping with inclusion of covariates revealed that especially left-hemispheric lesions in the deep periventricular white matter and adjacent internal capsule showed a great influence on functional outcome.

CONCLUSIONS

Our study confirms that infarct location has an important impact on functional outcome of patients with stroke and should be considered in prediction models. After we adjusted for covariates, the left-hemispheric corticosubcortical fiber tracts seemed to be of higher functional importance compared with cortical lesions.

MIDAS: Regionally linear multivariate discriminative statistical mapping

Statistical parametric maps formed via voxel-wise mass-univariate tests, such as the general linear model, are commonly used to test hypotheses about regionally specific effects in neuroimaging cross-sectional studies where each subject is represented by a single image. Despite being informative, these techniques remain limited as they ignore multivariate relationships in the data. Most importantly, the commonly employed local Gaussian smoothing, which is important for accounting for registration errors and making the data follow Gaussian distributions, is usually chosen in an ad hoc fashion. Thus, it is often suboptimal for the task of detecting group differences and correlations with non-imaging variables. Information mapping techniques, such as searchlight, which use pattern classifiers to exploit multivariate information and obtain more powerful statistical maps, have become increasingly popular in recent years. However, existing methods may lead to important interpretation errors in practice (i.e., misidentifying a cluster as informative, or failing to detect truly informative voxels), while often being computationally expensive. To address these issues, we introduce a novel efficient multivariate statistical framework for cross-sectional studies, termed MIDAS, seeking highly sensitive and specific voxel-wise brain maps, while leveraging the power of regional discriminant analysis. In MIDAS, locally linear discriminative learning is applied to estimate the pattern that best discriminates between two groups, or predicts a variable of interest. This pattern is equivalent to local filtering by an optimal kernel whose coefficients are the weights of the linear discriminant. By composing information from all neighborhoods that contain a given voxel, MIDAS produces a statistic that collectively reflects the contribution of the voxel to the regional classifiers as well as the discriminative power of the classifiers. Critically, MIDAS efficiently assesses the statistical significance of the derived statistic by analytically approximating its null distribution without the need for computationally expensive permutation tests. The proposed framework was extensively validated using simulated atrophy in structural magnetic resonance imaging (MRI) and further tested using data from a task-based functional MRI study as well as a structural MRI study of cognitive performance. The performance of the proposed framework was evaluated against standard voxel-wise general linear models and other information mapping methods. The experimental results showed that MIDAS achieves relatively higher sensitivity and specificity in detecting group differences. Together, our results demonstrate the potential of the proposed approach to efficiently map effects of interest in both structural and functional data.

Addressing Confounding in Predictive Models with an Application to Neuroimaging

Understanding structural changes in the brain that are caused by a particular disease is a major goal of neuroimaging research. Multivariate pattern analysis (MVPA) comprises a collection of tools that can be used to understand complex disease efxcfects across the brain. We discuss several important issues that must be considered when analyzing data from neuroimaging studies using MVPA. In particular, we focus on the consequences of confounding by non-imaging variables such as age and sex on the results of MVPA. After reviewing current practice to address confounding in neuroimaging studies, we propose an alternative approach based on inverse probability weighting. Although the proposed method is motivated by neuroimaging applications, it is broadly applicable to many problems in machine learning and predictive modeling. We demonstrate the advantages of our approach on simulated and real data examples.

BrainNetCNN: Convolutional neural networks for brain networks; towards predicting neurodevelopment

We propose BrainNetCNN, a convolutional neural network (CNN) framework to predict clinical neurodevelopmental outcomes from brain networks. In contrast to the spatially local convolutions done in traditional image-based CNNs, our BrainNetCNN is composed of novel edge-to-edge, edge-to-node and node-to-graph convolutional filters that leverage the topological locality of structural brain networks. We apply the BrainNetCNN framework to predict cognitive and motor developmental outcome scores from structural brain networks of infants born preterm. Diffusion tensor images (DTI) of preterm infants, acquired between 27 and 46 weeks gestational age, were used to construct a dataset of structural brain connectivity networks. We first demonstrate the predictive capabilities of BrainNetCNN on synthetic phantom networks with simulated injury patterns and added noise. BrainNetCNN outperforms a fully connected neural-network with the same number of model parameters on both phantoms with focal and diffuse injury patterns. We then apply our method to the task of joint prediction of Bayley-III cognitive and motor scores, assessed at 18 months of age, adjusted for prematurity. We show that our BrainNetCNN framework outperforms a variety of other methods on the same data. Furthermore, BrainNetCNN is able to identify an infant's postmenstrual age to within about 2 weeks. Finally, we explore the high-level features learned by BrainNetCNN by visualizing the importance of each connection in the brain with respect to predicting the outcome scores. These findings are then discussed in the context of the anatomy and function of the developing preterm infant brain.

Control-group feature normalization for multivariate pattern analysis of structural MRI data using the support vector machine

Normalization of feature vector values is a common practice in machine learning. Generally, each feature value is standardized to the unit hypercube or by normalizing to zero mean and unit variance. Classification decisions based on support vector machines (SVMs) or by other methods are sensitive to the specific normalization used on the features. In the context of multivariate pattern analysis using neuroimaging data, standardization effectively up- and down-weights features based on their individual variability. Since the standard approach uses the entire data set to guide the normalization, it utilizes the total variability of these features. This total variation is inevitably dependent on the amount of marginal separation between groups. Thus, such a normalization may attenuate the separability of the data in high dimensional space. In this work we propose an alternate approach that uses an estimate of the control-group standard deviation to normalize features before training. We study our proposed approach in the context of group classification using structural MRI data. We show that control-based normalization leads to better reproducibility of estimated multivariate disease patterns and improves the classifier performance in many cases.

A Sparse Bayesian Learning Algorithm for Longitudinal Image Data

Longitudinal imaging studies, where serial (multiple) scans are collected on each individual, are becoming increasingly widespread. The field of machine learning has in general neglected the longitudinal design, since many algorithms are built on the assumption that each datapoint is an independent sample. Thus, the application of general purpose machine learning tools to longitudinal image data can be sub-optimal. Here, we present a novel machine learning algorithm designed to handle longitudinal image datasets. Our approach builds on a sparse Bayesian image-based prediction algorithm. Our empirical results demonstrate that the proposed method can offer a significant boost in prediction performance with longitudinal clinical data.

Interpreting support vector machine models for multivariate group wise analysis in neuroimaging

Machine learning based classification algorithms like support vector machines (SVMs) have shown great promise for turning a high dimensional neuroimaging data into clinically useful decision criteria. However, tracing imaging based patterns that contribute significantly to classifier decisions remains an open problem. This is an issue of critical importance in imaging studies seeking to determine which anatomical or physiological imaging features contribute to the classifier's decision, thereby allowing users to critically evaluate the findings of such machine learning methods and to understand disease mechanisms. The majority of published work addresses the question of statistical inference for support vector classification using permutation tests based on SVM weight vectors. Such permutation testing ignores the SVM margin, which is critical in SVM theory. In this work we emphasize the use of a statistic that explicitly accounts for the SVM margin and show that the null distributions associated with this statistic are asymptotically normal. Further, our experiments show that this statistic is a lot less conservative as compared to weight based permutation tests and yet specific enough to tease out multivariate patterns in the data. Thus, we can better understand the multivariate patterns that the SVM uses for neuroimaging based classification.

2014 Update of the Alzheimer's Disease Neuroimaging Initiative: A review of papers published since its inception

The Alzheimer's Disease Neuroimaging Initiative (ADNI) is an ongoing, longitudinal, multicenter study designed to develop clinical, imaging, genetic, and biochemical biomarkers for the early detection and tracking of Alzheimer's disease (AD). The initial study, ADNI-1, enrolled 400 subjects with early mild cognitive impairment (MCI), 200 with early AD, and 200 cognitively normal elderly controls. ADNI-1 was extended by a 2-year Grand Opportunities grant in 2009 and by a competitive renewal, ADNI-2, which enrolled an additional 550 participants and will run until 2015. This article reviews all papers published since the inception of the initiative and summarizes the results to the end of 2013. The major accomplishments of ADNI have been as follows: (1) the development of standardized methods for clinical tests, magnetic resonance imaging (MRI), positron emission tomography (PET), and cerebrospinal fluid (CSF) biomarkers in a multicenter setting; (2) elucidation of the patterns and rates of change of imaging and CSF biomarker measurements in control subjects, MCI patients, and AD patients. CSF biomarkers are largely consistent with disease trajectories predicted by β-amyloid cascade (Hardy, J Alzheimer's Dis 2006;9(Suppl 3):151-3) and tau-mediated neurodegeneration hypotheses for AD, whereas brain atrophy and hypometabolism levels show predicted patterns but exhibit differing rates of change depending on region and disease severity; (3) the assessment of alternative methods of diagnostic categorization. Currently, the best classifiers select and combine optimum features from multiple modalities, including MRI, [(18)F]-fluorodeoxyglucose-PET, amyloid PET, CSF biomarkers, and clinical tests; (4) the development of blood biomarkers for AD as potentially noninvasive and low-cost alternatives to CSF biomarkers for AD diagnosis and the assessment of α-syn as an additional biomarker; (5) the development of methods for the early detection of AD. CSF biomarkers, β-amyloid 42 and tau, as well as amyloid PET may reflect the earliest steps in AD pathology in mildly symptomatic or even nonsymptomatic subjects and are leading candidates for the detection of AD in its preclinical stages; (6) the improvement of clinical trial efficiency through the identification of subjects most likely to undergo imminent future clinical decline and the use of more sensitive outcome measures to reduce sample sizes. Multimodal methods incorporating APOE status and longitudinal MRI proved most highly predictive of future decline. Refinements of clinical tests used as outcome measures such as clinical dementia rating-sum of boxes further reduced sample sizes; (7) the pioneering of genome-wide association studies that leverage quantitative imaging and biomarker phenotypes, including longitudinal data, to confirm recently identified loci, CR1, CLU, and PICALM and to identify novel AD risk loci; (8) worldwide impact through the establishment of ADNI-like programs in Japan, Australia, Argentina, Taiwan, China, Korea, Europe, and Italy; (9) understanding the biology and pathobiology of normal aging, MCI, and AD through integration of ADNI biomarker and clinical data to stimulate research that will resolve controversies about competing hypotheses on the etiopathogenesis of AD, thereby advancing efforts to find disease-modifying drugs for AD; and (10) the establishment of infrastructure to allow sharing of all raw and processed data without embargo to interested scientific investigators throughout the world.

Hyperglycaemia, insulin therapy and critical penumbral regions for prognosis in acute stroke: further insights from the INSULINFARCT trial

BACKGROUND

Recently, the concept of 'clinically relevant penumbra' was defined as an area saved by arterial recanalization and correlated with stroke outcome. This clinically relevant penumbra was located in the subcortical structures, especially the periventricular white matter. Our aims were to confirm this hypothesis, to investigate the impact of admission hyperglycemia and of insulin treatment on the severity of ischemic damages in this area and to study the respective contributions of infarct volume and ischemic damage severity of the clinically relevant penumbra on 3-month outcome.

METHODS

We included 99 patients from the INSULINFARCT trial. Voxel-Based Analysis was carried on the Apparent Diffusion Coefficient (ADC) maps obtained at day one to localize the regions, which were more damaged in patients i) with poor clinical outcomes at three months and ii) without arterial recanalization. We determined the intersection of the detected areas, which represents the clinically relevant penumbra and investigated whether hyperglycemic status and insulin regimen affected the severity of ischemic damages in this area. We performed logistic regression to examine the contribution of infarct volume or early ADC decrease in this strategic area on 3-month outcome.

FINDINGS

Lower ADC values were found in the corona radiata in patients with poor prognosis (p< 0.0001) and in those without arterial recanalization (p< 0.0001). The tracking analysis showed that lesions in this area interrupted many important pathways. ADC values in this area were lower in hyperglycemic than in normoglycemic patients (average decrease of 41.6 ± 20.8 x10(-6) mm2/s) and unaffected by the insulin regimen (p: 0.10). ADC values in the clinically relevant penumbra, but not infarct volumes, were significant predictors of 3-month outcome.

CONCLUSION

These results confirm that the deep hemispheric white matter is part of the clinically relevant penumbra and show that hyperglycaemia exacerbates the apparition of irreversible ischemic damage within 24 hours in this area. However, early intensive insulin therapy fails to protect this area from infarction.

TRIAL REGISTRATION

ClinicalTrials.gov NCT00472381.

The ischemic penumbra: the location rather than the volume of recovery determines outcome

PURPOSE OF REVIEW

The clinical efficiency of thrombolysis is explained by the rescue of ischemic penumbra areas resulting from early arterial recanalization. The perfusion-diffusion weighted imaging mismatch is a commonly used MRI surrogate of the ischemic penumbra. However, the randomized trials testing the mismatch hypothesis have been negative. We will review the 'mismatch concept' and the recent studies that aim to localize the clinically eloquent areas of penumbra in middle cerebral artery (MCA) infarcts.

RECENT FINDINGS

New methods of image analysis have shown that poor outcomes after MCA stroke are related to infarction of an extremely well localized area of the periventricular white matter and adjacent internal capsule, where projections and association tracts are crossing and converging. This area almost colocalizes with the area salvaged by early arterial recanalization and is located extremely close to the initial ischemic core.

SUMMARY

The location of the area that correlates with disability in MCA stroke patients and that is salvaged by early arterial recanalization is in the same specific region of the deep white matter, close to the initial ischemic core. These findings may have important implications for designing new recanalization trials and support the importance of basic research on white-matter neuroprotection.

Identifying endophenotypes of autism: a multivariate approach

The existence of an endophenotype of autism spectrum condition (ASC) has been recently suggested by several commentators. It can be estimated by finding differences between controls and people with ASC that are also present when comparing controls and the unaffected siblings of ASC individuals. In this work, we used a multivariate methodology applied on magnetic resonance images to look for such differences. The proposed procedure consists of combining a searchlight approach and a support vector machine classifier to identify the differences between three groups of participants in pairwise comparisons: controls, people with ASC and their unaffected siblings. Then we compared those differences selecting spatially collocated as candidate endophenotypes of ASC.

Analytic estimation of statistical significance maps for support vector machine based multi-variate image analysis and classification

Multivariate pattern analysis (MVPA) methods such as support vector machines (SVMs) have been increasingly applied to fMRI and sMRI analyses, enabling the detection of distinctive imaging patterns. However, identifying brain regions that significantly contribute to the classification/group separation requires computationally expensive permutation testing. In this paper we show that the results of SVM-permutation testing can be analytically approximated. This approximation leads to more than a thousandfold speedup of the permutation testing procedure, thereby rendering it feasible to perform such tests on standard computers. The speedup achieved makes SVM based group difference analysis competitive with standard univariate group difference analysis methods.

The Alzheimer's Disease Neuroimaging Initiative: a review of papers published since its inception

The Alzheimer's Disease Neuroimaging Initiative (ADNI) is an ongoing, longitudinal, multicenter study designed to develop clinical, imaging, genetic, and biochemical biomarkers for the early detection and tracking of Alzheimer's disease (AD). The study aimed to enroll 400 subjects with early mild cognitive impairment (MCI), 200 subjects with early AD, and 200 normal control subjects; $67 million funding was provided by both the public and private sectors, including the National Institute on Aging, 13 pharmaceutical companies, and 2 foundations that provided support through the Foundation for the National Institutes of Health. This article reviews all papers published since the inception of the initiative and summarizes the results as of February 2011. The major accomplishments of ADNI have been as follows: (1) the development of standardized methods for clinical tests, magnetic resonance imaging (MRI), positron emission tomography (PET), and cerebrospinal fluid (CSF) biomarkers in a multicenter setting; (2) elucidation of the patterns and rates of change of imaging and CSF biomarker measurements in control subjects, MCI patients, and AD patients. CSF biomarkers are consistent with disease trajectories predicted by β-amyloid cascade (Hardy, J Alzheimers Dis 2006;9(Suppl 3):151-3) and tau-mediated neurodegeneration hypotheses for AD, whereas brain atrophy and hypometabolism levels show predicted patterns but exhibit differing rates of change depending on region and disease severity; (3) the assessment of alternative methods of diagnostic categorization. Currently, the best classifiers combine optimum features from multiple modalities, including MRI, [(18)F]-fluorodeoxyglucose-PET, CSF biomarkers, and clinical tests; (4) the development of methods for the early detection of AD. CSF biomarkers, β-amyloid 42 and tau, as well as amyloid PET may reflect the earliest steps in AD pathology in mildly symptomatic or even nonsymptomatic subjects, and are leading candidates for the detection of AD in its preclinical stages; (5) the improvement of clinical trial efficiency through the identification of subjects most likely to undergo imminent future clinical decline and the use of more sensitive outcome measures to reduce sample sizes. Baseline cognitive and/or MRI measures generally predicted future decline better than other modalities, whereas MRI measures of change were shown to be the most efficient outcome measures; (6) the confirmation of the AD risk loci CLU, CR1, and PICALM and the identification of novel candidate risk loci; (7) worldwide impact through the establishment of ADNI-like programs in Europe, Asia, and Australia; (8) understanding the biology and pathobiology of normal aging, MCI, and AD through integration of ADNI biomarker data with clinical data from ADNI to stimulate research that will resolve controversies about competing hypotheses on the etiopathogenesis of AD, thereby advancing efforts to find disease-modifying drugs for AD; and (9) the establishment of infrastructure to allow sharing of all raw and processed data without embargo to interested scientific investigators throughout the world. The ADNI study was extended by a 2-year Grand Opportunities grant in 2009 and a renewal of ADNI (ADNI-2) in October 2010 through to 2016, with enrollment of an additional 550 participants.

Searchlight analysis: promise, pitfalls, and potential

Multivariate pattern analysis (MVPA) is an increasingly popular approach for characterizing the information present in neural activity as measured by fMRI. For neuroimaging researchers, the searchlight technique serves as the most intuitively appealing means of implementing MVPA with fMRI data. However, searchlight approaches carry with them a number of special concerns and limitations that can lead to serious interpretation errors in practice, such as misidentifying a cluster as informative, or failing to detect truly informative voxels. Here we describe how such distorted results can occur, using both schematic illustrations and examples from actual fMRI datasets. We recommend that confirmatory and sensitivity tests, such as the ones prescribed here, should be considered a necessary stage of searchlight analysis interpretation, and that their adoption will allow the full potential of searchlight analysis to be realized.

Spatial and Anatomical Regularization of SVM: A General Framework for Neuroimaging Data

This paper presents a framework to introduce spatial and anatomical priors in SVM for brain image analysis based on regularization operators. A notion of proximity based on prior anatomical knowledge between the image points is defined by a graph (e.g., brain connectivity graph) or a metric (e.g., Fisher metric on statistical manifolds). A regularization operator is then defined from the graph Laplacian, in the discrete case, or from the Laplace-Beltrami operator, in the continuous case. The regularization operator is then introduced into the SVM, which exponentially penalizes high-frequency components with respect to the graph or to the metric and thus constrains the classification function to be smooth with respect to the prior. It yields a new SVM optimization problem whose kernel is a heat kernel on graphs or on manifolds. We then present different types of priors and provide efficient computations of the Gram matrix. The proposed framework is finally applied to the classification of brain Magnetic Resonance (MR) images (based on Gray Matter (GM) concentration maps and cortical thickness measures) from 137 patients with Alzheimer's Disease (AD) and 162 elderly controls. The results demonstrate that the proposed classifier generates less-noisy and consequently more interpretable feature maps with high classification performances.

The relevance voxel machine (RVoxM): a self-tuning Bayesian model for informative image-based prediction

This paper presents the relevance voxel machine (RVoxM), a dedicated Bayesian model for making predictions based on medical imaging data. In contrast to the generic machine learning algorithms that have often been used for this purpose, the method is designed to utilize a small number of spatially clustered sets of voxels that are particularly suited for clinical interpretation. RVoxM automatically tunes all its free parameters during the training phase, and offers the additional advantage of producing probabilistic prediction outcomes. We demonstrate RVoxM as a regression model by predicting age from volumetric gray matter segmentations, and as a classification model by distinguishing patients with Alzheimer's disease from healthy controls using surface-based cortical thickness data. Our results indicate that RVoxM yields biologically meaningful models, while providing state-of-the-art predictive accuracy.