Jean Honorio

Two-person interaction detection using body-pose features and multiple instance learning

Human activity recognition has potential to impact a wide range of applications from surveillance to human computer interfaces to content based video retrieval. Recently, the rapid development of inexpensive depth sensors (e.g. Microsoft Kinect) provides adequate accuracy for real-time full-body human tracking for activity recognition applications. In this paper, we create a complex human activity dataset depicting two person interactions, including synchronized video, depth and motion capture data. Moreover, we use our dataset to evaluate various features typically used for indexing and retrieval of motion capture data, in the context of real-time detection of interaction activities via Support Vector Machines (SVMs). Experimentally, we find that the geometric relational features based on distance between all pairs of joints outperforms other feature choices. For whole sequence classification, we also explore techniques related to Multiple Instance Learning (MIL) in which the sequence is represented by a bag of body-pose features. We find that the MIL based classifier outperforms SVMs when the sequences extend temporally around the interaction of interest.

Oral methylphenidate normalizes cingulate activity in cocaine addiction during a salient cognitive task

Anterior cingulate cortex (ACC) hypoactivations during cognitive demand are a hallmark deficit in drug addiction. Methylphenidate (MPH) normalizes cortical function, enhancing task salience and improving associated cognitive abilities, in other frontal lobe pathologies; however, in clinical trials, MPH did not improve treatment outcome in cocaine addiction. We hypothesized that oral MPH will attenuate ACC hypoactivations and improve associated performance during a salient cognitive task in individuals with cocaine-use disorders (CUD). In the current functional MRI study, we used a rewarded drug cue-reactivity task previously shown to be associated with hypoactivations in both major ACC subdivisions (implicated in default brain function) in CUD compared with healthy controls. The task was performed by 13 CUD and 14 matched healthy controls on 2 d: after ingesting a single dose of oral MPH (20 mg) or placebo (lactose) in a counterbalanced fashion. Results show that oral MPH increased responses to this salient cognitive task in both major ACC subdivisions (including the caudal-dorsal ACC and rostroventromedial ACC extending to the medial orbitofrontal cortex) in the CUD. These functional MRI results were associated with reduced errors of commission (a common impulsivity measure) and improved task accuracy, especially during the drug (vs. neutral) cue-reactivity condition in all subjects. The clinical application of such MPH-induced brain-behavior enhancements remains to be tested.

Methylphenidate Enhances Executive Function and Optimizes Prefrontal Function in Both Health and Cocaine Addiction

Previous studies have suggested dopamine to be involved in error monitoring/processing, possibly through impact on reinforcement learning. The current study tested whether methylphenidate (MPH), an indirect dopamine agonist, modulates brain and behavioral responses to error, and whether such modulation is more pronounced in cocaine-addicted individuals, in whom dopamine neurotransmission is disrupted. After receiving oral MPH (20 mg) or placebo (counterbalanced), 15 healthy human volunteers and 16 cocaine-addicted individuals completed a task of executive function (the Stroop color word) during functional magnetic resonance imaging (fMRI). During MPH, despite not showing differences on percent accuracy and reaction time, all subjects committed fewer total errors and slowed down more after committing errors, suggestive of more careful responding. In parallel, during MPH all subjects showed reduced dorsal anterior cingulate cortex response to the fMRI contrast error>correct. In the cocaine subjects only, MPH also reduced error>correct activity in the dorsolateral prefrontal cortex (controls instead showed lower error>correct response in this region during placebo). Taken together, MPH modulated dopaminergically innervated prefrontal cortical areas involved in error-related processing, and such modulation was accentuated in the cocaine subjects. These results are consistent with a dopaminergic contribution to error-related processing during a cognitive control task.

Dopaminergic involvement during mental fatigue in health and cocaine addiction

Dopamine modulates executive function, including sustaining cognitive control during mental fatigue. Using event-related functional magnetic resonance imaging (fMRI) during the color-word Stroop task, we aimed to model mental fatigue with repeated task exposures in 33 cocaine abusers and 20 healthy controls. During such mental fatigue (indicated by increased errors, and decreased post-error slowing and dorsal anterior cingulate response to error as a function of time-on-task), healthy individuals showed increased activity in the dopaminergic midbrain to error. Cocaine abusers, characterized by disrupted dopamine neurotransmission, showed an opposite pattern of response. This midbrain fMRI activity with repetition was further correlated with objective indices of endogenous motivation in all subjects: a state measure (task reaction time) and a trait measure (dopamine D2 receptor availability in caudate, as revealed by positron emission tomography data collected in a subset of this sample, which directly points to a contribution of dopamine to these results). In a second sample of 14 cocaine abusers and 15 controls, administration of an indirect dopamine agonist, methylphenidate, reversed these midbrain responses in both groups, possibly indicating normalization of response in cocaine abusers because of restoration of dopamine signaling but degradation of response in healthy controls owing to excessive dopamine signaling. Together, these multimodal imaging findings suggest a novel involvement of the dopaminergic midbrain in sustaining motivation during fatigue. This region might provide a useful target for strengthening self-control and/or endogenous motivation in addiction.

Enhanced midbrain response at 6-month follow-up in cocaine addiction, association with reduced drug-related choice.

Drug addiction is characterized by dysregulated dopamine neurotransmission. Although dopamine functioning appears to partially recover with abstinence, the specific regions that recover and potential impact on drug seeking remain to be determined. Here we used functional magnetic resonance imaging (fMRI) to study an ecologically valid sample of 15 treatment‐seeking cocaine addicted individuals at baseline and 6‐month follow‐up. At both study sessions, we collected fMRI scans during performance of a drug Stroop task, clinical self‐report measures of addiction severity and behavioral measures of cocaine seeking (simulated cocaine choice); actual drug use in between the two study sessions was also monitored. At 6‐month follow‐up (compared with baseline), we predicted functional enhancement of dopaminergically innervated brain regions, relevant to the behavioral responsiveness toward salient stimuli. Consistent with predictions, whole‐brain analyses revealed responses in the midbrain (encompassing the ventral tegmental area/substantia nigra complex) and thalamus (encompassing the mediodorsal nucleus) that were higher (and more positively correlated) at follow‐up than baseline. Increased midbrain activity from baseline to follow‐up correlated with reduced simulated cocaine choice, indicating that heightened midbrain activations in this context may be marking lower approach motivation for cocaine. Normalization of midbrain function at follow‐up was also suggested by exploratory comparisons with active cocaine users and healthy controls (who were assessed only at baseline). Enhanced self‐control at follow‐up was suggested by a trend for the commonly hypoactive dorsal anterior cingulate cortex to increase response during a drug‐related context. Together, these results suggest that fMRI could be useful in sensitively tracking follow‐up outcomes in drug addiction.

Can a Single Brain Region Predict a Disorder

We perform prediction of diverse disorders (cocaine use, schizophrenia and Alzheimers disease) in unseen subjects from brain functional magnetic resonance imaging. First, we show that for multisubject prediction of simple cognitive states (e.g., motor versus calculation and reading), voxels-as-features methods produce clusters that are similar for different leave-one-subject-out folds; while for group classification (e.g., cocaine addicted versus control subjects), voxels are scattered and less stable. Therefore, we chose to use a single region per experimental condition and a majority vote classifier. Interestingly, our method outperforms state-of-the-art techniques. Our method can integrate multiple experimental conditions and successfully predict disorders in unseen subjects (leave-one-subject-out generalization accuracy: 89.3% and 90.9% for cocaine use, 96.4% for schizophrenia and 81.5% for Alzheimers disease). Our experimental results not only span diverse disorders, but also different experimental designs (block design and event related tasks), facilities, magnetic fields (1.5T, 3T, 4T) and speed of acquisition (interscan interval from 1600 to 3500 ms). We further argue that our method produces a meaningful low-dimensional representation that retains discriminability.

FMRI analysis of cocaine addiction using k-support sparsity

In this paper, we explore various sparse regularization techniques for analyzing fMRI data, such as LASSO, elastic net and the recently introduced k-support norm. Employing sparsity regularization allow us to handle the curse of dimensionality, a problem commonly found in fMRI analysis. We test these methods on real data of both healthy subjects as well as cocaine addicted ones and we show that although LASSO has good prediction, it lacks interpretability since the resulting model is too sparse, and results are highly sensitive to the regularization parameter. We find that we can improve prediction performance over the LASSO using elastic net or the k-support norm, which is a convex relaxation to sparsity with an ℓ2 penalty that is tighter than the elastic net. Elastic net and k-support norm overcome the problem of overly sparse solutions, resulting in both good prediction and interpretable solutions, while the k-support norm gave better prediction performance. Our experimental results support the general applicability of the k-support norm in fMRI analysis, both for prediction performance and interpretability.

Predictive sparse modeling of fMRI data for improved classification, regression, and visualization using the k-support norm

We explore various sparse regularization techniques for analyzing fMRI data, such as the ℓ1 norm (often called LASSO in the context of a squared loss function), elastic net, and the recently introduced k-support norm. Employing sparsity regularization allows us to handle the curse of dimensionality, a problem commonly found in fMRI analysis. In this work we consider sparse regularization in both the regression and classification settings. We perform experiments on fMRI scans from cocaine-addicted as well as healthy control subjects. We show that in many cases, use of the k-support norm leads to better predictive performance, solution stability, and interpretability as compared to other standard approaches. We additionally analyze the advantages of using the absolute loss function versus the standard squared loss which leads to significantly better predictive performance for the regularization methods tested in almost all cases. Our results support the use of the k-support norm for fMRI analysis and on the clinical side, the generalizability of the I-RISA model of cocaine addiction.

fMRI Analysis with Sparse Weisfeiler-Lehman Graph Statistics

fMRI analysis has most often been approached with linear methods. However, this disregards information encoded in the relationships between voxels. We propose to exploit the inherent spatial structure of the brain to improve the prediction performance of fMRI analysis. We do so in an exploratory fashion by representing the fMRI data by graphs. We use the Weisfeiler-Lehman algorithm to efficiently compute subtree features of the graphs. These features encode non-linear interactions between voxels, which contain additional discriminative information that cannot be captured by a linear classifier. In order to make use of the efficiency of the Weisfeiler-Lehman algorithm, we introduce a novel pyramid quantization strategy to approximate continuously labeled graphs with a sequence of discretely labeled graphs. To control the capacity of the resulting prediction function, we utilize the elastic net sparsity regularizer. We validate our method on a cocaine addiction dataset showing a significant improvement over elastic net and kernel ridge regression baselines and a reduction in classification error of over 14%. Source code is also available at https://gitorious.org/wlpyramid .

Simple fully automated group classification on brain fMRI

We propose a simple, well grounded classification technique which is suited for group classification on brain fMRI datasets that have high dimensionality, small number of subjects, high noise level, high subject variability, imperfect registration and capture subtle cognitive effects. We propose threshold-split region as a new feature selection method and majority vote as the classification technique. Our method does not require a predefined set of regions of interest. We use average across sessions, only one feature per experimental condition, feature independence assumption, and simple classifiers. The seeming counter-intuitive approach of using a simple design is supported by signal processing and statistical theory. Experimental results in two block design datasets that capture brain function under distinct monetary rewards for cocaine addicted and control subjects, show that our method exhibits increased generalization accuracy compared to commonly used feature selection and classification techniques.