Structured Sparse Principal Components Analysis With the TV-Elastic Net Penalty

Compressed Feature Vector-based Effective Object Recognition Model in Detection of COVID-19

To better understand the structure of the COVID-19, and to improve the recognition speed, an effective recognition model based on compressed feature vector is proposed. Object recognition plays an important role in computer vison aera. To improve the recognition accuracy, most recent approaches always adopt a set of complicated hand-craft feature vectors and build the complex classifiers. Although such approaches achieve the favourable performance on recognition accuracy, they are inefficient. To raise the recognition speed without decreasing the accuracy loss, this paper proposed an efficient recognition modeltrained witha kind of compressed feature vectors. Firstly, we propose a kind of compressed feature vector based on the theory of compressive sensing. A sparse matrix is adopted to compress feature vector from very high dimensions to very low dimensions, which reduces the computation complexity and saves enough information for model training and predicting. Moreover, to improve the inference efficiency during the classification stage, an efficient recognition model is built by a novel optimization approach, which reduces the support vectors of kernel-support vector machine (kernel SVM). The SVM model is established with whether the subject is infected with the COVID-19 as the dependent variable, and the age, gender, nationality, and other factors as independent variables. The proposed approach iteratively builds a compact set of the support vectors from the original kernel SVM, and then the new generated model achieves approximate recognition accuracy with the original kernel SVM. Additionally, with the reduction of support vectors, the recognition time of new generated is greatly improved. Finally, the COVID-19 patients have specific epidemiological characteristics, and the SVM recognition model has strong fitting ability. From the extensive experimental results conducted on two datasets, the proposed object recognition model achieves favourable performance not only on recognition accuracy but also on recognition speed.

Multivariate semi-blind deconvolution of fMRI time series

Whole brain estimation of the haemodynamic response function (HRF) in functional magnetic resonance imaging (fMRI) is critical to get insight on the global status of the neurovascular coupling of an individual in healthy or pathological condition. Most of existing approaches in the literature works on task-fMRI data and relies on the experimental paradigm as a surrogate of neural activity, hence remaining inoperative on resting-stage fMRI (rs-fMRI) data. To cope with this issue, recent works have performed either a two-step analysis to detect large neural events and then characterize the HRF shape or a joint estimation of both the neural and haemodynamic components in an univariate fashion. In this work, we express the neural activity signals as a combination of piece-wise constant temporal atoms associated with sparse spatial maps and introduce an haemodynamic parcellation of the brain featuring a temporally dilated version of a given HRF model in each parcel with unknown dilation parameters. We formulate the joint estimation of the HRF shapes and spatio-temporal neural representations as a multivariate semi-blind deconvolution problem in a paradigm-free setting and introduce constraints inspired from the dictionary learning literature to ease its identifiability. A fast alternating minimization algorithm, along with its efficient implementation, is proposed and validated on both synthetic and real rs-fMRI data at the subject level. To demonstrate its significance at the population level, we apply this new framework to the UK Biobank data set, first for the discrimination of haemodynamic territories between balanced groups (n=24 individuals in each) patients with an history of stroke and healthy controls and second, for the analysis of normal aging on the neurovascular coupling. Overall, we statistically demonstrate that a pathology like stroke or a condition like normal brain aging induce longer haemodynamic delays in certain brain areas (e.g. Willis polygon, occipital, temporal and frontal cortices) and that this haemodynamic feature may be predictive with an accuracy of 74 % of the individual's age in a supervised classification task performed on n=459 subjects.

Hemodynamic matrix factorization for functional magnetic resonance imaging

The General Linear Model (GLM) used in task-fMRI relates activated brain areas to extrinsic task conditions. The translation of resulting neural activation into a hemodynamic response is commonly approximated with a linear convolution model using a hemodynamic response function (HRF). There are two major limitations in GLM analysis. Firstly, the GLM assumes that neural activation is either on or off and matches the exact stimulus duration in the corresponding task timings. Secondly, brain networks observed in resting-state fMRI experiments present also during task experiments, but the GLM approach models these task-unrelated brain activity as noise. A novel kernel matrix factorization approach, called hemodynamic matrix factorization (HMF), is therefore proposed that addresses both limitations by assuming that task-related and task-unrelated brain activity can be modeled with the same convolution model as in GLM analysis. By contrast to the GLM, the proposed HMF is a blind source separation (BSS) technique, which decomposes fMRI data into modes. Each mode comprises of a neural activation time course and a spatial mapping. Two versions of HMF are proposed in which the neural activation time course of each mode is convolved with either the canonical HRF or predetermined subject-specific HRFs. Firstly, HMF with the canonical HRF is applied to two open-source cohorts. These cohorts comprise of several task experiments including motor, incidental memory, spatial coherence discrimination, verbal discrimination task and a very short localization task, engaging multiple parts of the eloquent cortex. HMF modes were obtained whose neural activation time course followed original task timings and whose corresponding spatial map matched cortical areas known to be involved in the respective task processing. Secondly, the alignment of these neural activation time courses to task timings were further improved by replacing the canonical HRF with subject-specific HRFs during HMF mode computation. In addition to task-related modes, HMF also produced seemingly task-unrelated modes whose spatial maps matched known resting-state networks. The validity of a fMRI task experiment relies on the assumption that the exposure to a stimulus for a given time causes an imminent increase in neural activation of equal duration. The proposed HMF is an attempt to falsify this assumption and allows to identify subject task participation that does not comply with the experiment instructions.

Similarity-driven multi-view embeddings from high-dimensional biomedical data

Diverse, high-dimensional modalities collected in large cohorts present new opportunities for the formulation and testing of integrative scientific hypotheses. Similarity-driven multi-view linear reconstruction (SiMLR) is an algorithm that exploits inter-modality relationships to transform large scientific datasets into smaller, more well-powered and interpretable low-dimensional spaces. SiMLR contributes an objective function for identifying joint signal, regularization based on sparse matrices representing prior within-modality relationships and an implementation that permits application to joint reduction of large data matrices. We demonstrate that SiMLR outperforms closely related methods on supervised learning problems in simulation data, a multi-omics cancer survival prediction dataset and multiple modality neuroimaging datasets. Taken together, this collection of results shows that SiMLR may be applied to joint signal estimation from disparate modalities and may yield practically useful results in a variety of application domains.

Will machine learning applied to neuroimaging in bipolar disorder help the clinician? A critical review and methodological suggestions

OBJECTIVES

The existence of anatomofunctional brain abnormalities in bipolar disorder (BD) is now well established by magnetic resonance imaging (MRI) studies. To create diagnostic and prognostic tools, as well as identifying biologically valid subtypes of BD, research has recently turned towards the use of machine learning (ML) techniques. We assessed both supervised ML and unsupervised ML studies in BD to evaluate their robustness, reproducibility and the potential need for improvement.

METHOD

We systematically searched for studies using ML algorithms based on MRI data of patients with BD until February 2019.

RESULT

We identified 47 studies, 45 using supervised ML techniques and 2 including unsupervised ML analyses. Among supervised studies, 43 focused on diagnostic classification. The reported accuracies for classification of BD ranged between (a) 57% and 100%, for BD vs healthy controls; (b) 49.5% and 93.1% for BD vs patients with major depressive disorder; and (c) 50% and 96.2% for BD vs patients with schizophrenia. Reported accuracies for discriminating subjects genetically at risk for BD (either from control or from patients with BD) ranged between 64.3% and 88.93%.

CONCLUSIONS

Although there are strong methodological limitations in previous studies and an important need for replication in large multicentric samples, the conclusions of our review bring hope of future computer-aided diagnosis of BD and pave the way for other applications, such as treatment response prediction. To reinforce the reliability of future results we provide methodological suggestions for good practice in conducting and reporting MRI-based ML studies in BD.

Toward a unified framework for interpreting machine-learning models in neuroimaging

Machine learning is a powerful tool for creating computational models relating brain function to behavior, and its use is becoming widespread in neuroscience. However, these models are complex and often hard to interpret, making it difficult to evaluate their neuroscientific validity and contribution to understanding the brain. For neuroimaging-based machine-learning models to be interpretable, they should (i) be comprehensible to humans, (ii) provide useful information about what mental or behavioral constructs are represented in particular brain pathways or regions, and (iii) demonstrate that they are based on relevant neurobiological signal, not artifacts or confounds. In this protocol, we introduce a unified framework that consists of model-, feature- and biology-level assessments to provide complementary results that support the understanding of how and why a model works. Although the framework can be applied to different types of models and data, this protocol provides practical tools and examples of selected analysis methods for a functional MRI dataset and multivariate pattern-based predictive models. A user of the protocol should be familiar with basic programming in MATLAB or Python. This protocol will help build more interpretable neuroimaging-based machine-learning models, contributing to the cumulative understanding of brain mechanisms and brain health. Although the analyses provided here constitute a limited set of tests and take a few hours to days to complete, depending on the size of data and available computational resources, we envision the process of annotating and interpreting models as an open-ended process, involving collaborative efforts across multiple studies and laboratories.

Continuation of Nesterov's Smoothing for Regression With Structured Sparsity in High-Dimensional Neuroimaging

Predictive models can be used on high-dimensional brain images to decode cognitive states or diagnosis/prognosis of a clinical condition/evolution. Spatial regularization through structured sparsity offers new perspectives in this context and reduces the risk of overfitting the model while providing interpretable neuroimaging signatures by forcing the solution to adhere to domain-specific constraints. Total variation (TV) is a promising candidate for structured penalization: it enforces spatial smoothness of the solution while segmenting predictive regions from the background. We consider the problem of minimizing the sum of a smooth convex loss, a non-smooth convex penalty (whose proximal operator is known) and a wide range of possible complex, non-smooth convex structured penalties such as TV or overlapping group Lasso. Existing solvers are either limited in the functions they can minimize or in their practical capacity to scale to high-dimensional imaging data. Nesterov's smoothing technique can be used to minimize a large number of non-smooth convex structured penalties. However, reasonable precision requires a small smoothing parameter, which slows down the convergence speed to unacceptable levels. To benefit from the versatility of Nesterov's smoothing technique, we propose a first order continuation algorithm, CONESTA, which automatically generates a sequence of decreasing smoothing parameters. The generated sequence maintains the optimal convergence speed toward any globally desired precision. Our main contributions are: gap to probe the current distance to the global optimum in order to adapt the smoothing parameter and the To propose an expression of the duality convergence speed. This expression is applicable to many penalties and can be used with other solvers than CONESTA. We also propose an expression for the particular smoothing parameter that minimizes the number of iterations required to reach a given precision. Furthermore, we provide a convergence proof and its rate, which is an improvement over classical proximal gradient smoothing methods. We demonstrate on both simulated and high-dimensional structural neuroimaging data that CONESTA significantly outperforms many state-of-the-art solvers in regard to convergence speed and precision.

Different Types of White Matter Hyperintensities in CADASIL

Objective: In CADASIL (Cerebral Autosomal Dominant Arteriopathy with Subcortical Infarcts and Leukoencephalopathy), white matter hyperintensities (WMH) are considered to result from hypoperfusion. We hypothesized that in fact the burden of WMH results from the combination of several regional populations of WMH with different mechanisms and clinical consequences.

Methods: To identify regional WMH populations, we used a 4-step approach. First, we used an unsupervised principal component algorithm to determine, without a priori knowledge, the main sources of variation of the global spatial pattern of WMH. Thereafter, to determine whether these sources are likely to include relevant information regarding regional populations of WMH, we tested their relationships with: (1) MRI markers of the disease; (2) the clinical severity assessed by the Mattis Dementia Rating scale (MDRS) (cognitive outcome) and the modified Rankin's score (disability outcome). Finally, through careful interpretation of all the results, we tried to identify different regional populations of WMH.

Results: The unsupervised principal component algorithm identified 3 main sources of variation of the global spatial pattern of WMH, which showed significant and sometime inverse relationships with MRI markers and clinical scores. The models predicting clinical severity based on these sources outperformed those evaluating WMH by their volume (MDRS, coefficient of determination of 39.0 vs. 35.3%, p = 0.01; modified Rankin's score, 43.7 vs. 38.1%, p = 0.001). By carefully interpreting the visual aspect of these sources as well as their relationships with MRI markers and clinical severity, we found strong arguments supporting the existence of different regional populations of WMH. For instance, in multivariate analyses, larger extents of WMH in anterior temporal poles and superior frontal gyri were associated with better outcomes, while larger extents of WMH in pyramidal tracts were associated with worse outcomes, which could not be explained if WMH in these different areas shared the same mechanisms.

Conclusion: The results of the present study support the hypothesis that the whole extent of WMH results from a combination of different regional populations of WMH, some of which are associated, for yet undetermined reasons, with milder forms of the disease.

Prediction of activation patterns preceding hallucinations in patients with schizophrenia using machine learning with structured sparsity

Despite significant progress in the field, the detection of fMRI signal changes during hallucinatory events remains difficult and time-consuming. This article first proposes a machine-learning algorithm to automatically identify resting-state fMRI periods that precede hallucinations versus periods that do not. When applied to whole-brain fMRI data, state-of-the-art classification methods, such as support vector machines (SVM), yield dense solutions that are difficult to interpret. We proposed to extend the existing sparse classification methods by taking the spatial structure of brain images into account with structured sparsity using the total variation penalty. Based on this approach, we obtained reliable classifying performances associated with interpretable predictive patterns, composed of two clearly identifiable clusters in speech-related brain regions. The variation in transition-to-hallucination functional patterns not only from one patient to another but also from one occurrence to the next (e.g., also depending on the sensory modalities involved) appeared to be the major difficulty when developing effective classifiers. Consequently, second, this article aimed to characterize the variability within the prehallucination patterns using an extension of principal component analysis with spatial constraints. The principal components (PCs) and the associated basis patterns shed light on the intrinsic structures of the variability present in the dataset. Such results are promising in the scope of innovative fMRI-guided therapy for drug-resistant hallucinations, such as fMRI-based neurofeedback.