A review of feature reduction techniques in neuroimaging.

Machine learning techniques are increasingly being used in making relevant predictions and inferences on individual subjects neuroimaging scan data. Previous studies have mostly focused on categorical discrimination of patients and matched healthy controls and more recently, on prediction of individual continuous variables such as clinical scores or age. However, these studies are greatly hampered by the large number of predictor variables (voxels) and low observations (subjects) also known as the curse-of-dimensionality or small-n-large-p problem. As a result, feature reduction techniques such as feature subset selection and dimensionality reduction are used to remove redundant predictor variables and experimental noise, a process which mitigates the curse-of-dimensionality and small-n-large-p effects. Feature reduction is an essential step before training a machine learning model to avoid overfitting and therefore improving model prediction accuracy and generalization ability. In this review, we discuss feature reduction techniques used with machine learning in neuroimaging studies.

EEG/MEG source reconstruction with spatial-temporal two-way regularized regression.

In this work, we propose a spatial-temporal two-way regularized regression method for reconstructing neural source signals from EEG/MEG time course measurements. The proposed method estimates the dipole locations and amplitudes simultaneously through minimizing a single penalized least squares criterion. The novelty of our methodology is the simultaneous consideration of three desirable properties of the reconstructed source signals, that is, spatial focality, spatial smoothness, and temporal smoothness. The desirable properties are achieved by using three separate penalty functions in the penalized regression framework. Specifically, we impose a roughness penalty in the temporal domain for temporal smoothness, and a sparsity-inducing penalty and a graph Laplacian penalty in the spatial domain for spatial focality and smoothness. We develop a computational efficient multilevel block coordinate descent algorithm to implement the method. Using a simulation study with several settings of different spatial complexity and two real MEG examples, we show that the proposed method outperforms existing methods that use only a subset of the three penalty functions.

Treatment of mild traumatic brain injury with an erythropoietin-mimetic peptide.

Mild traumatic brain injury (mTBI) results in an estimated 75-90% of the 1.7 million TBI-related emergency room visits each year. Post-concussion symptoms, which can include impaired memory problems, may persist for prolonged periods of time in a fraction of these cases. The purpose of this study was to determine if an erythropoietin-mimetic peptide, pyroglutamate helix B surface peptide (pHBSP), would improve neurological outcomes following mTBI. Sixty-four rats were randomly assigned to pHBSP or control (inactive peptide) 30 µg/kg IP every 12 h for 3 days, starting at either 1 hour (early treatment) or 24 h (delayed treatment), after mTBI (cortical impact injury 3 m/sec, 2.5 mm deformation). Treatment with pHBSP resulted in significantly improved performance on the Morris water maze task. Rats that received pHBSP required 22.3±1.3 sec to find the platform, compared to 26.3±1.3 sec in control rats (p=0.022). The rats that received pHBSP also traveled a significantly shorter distance to get to the platform, 5.0±0.3 meters, compared to 6.1±0.3 meters in control rats (p=0.019). Motor tasks were only transiently impaired in this mTBI model, and no treatment effect on motor performance was observed with pHBSP. Despite the minimal tissue injury with this mTBI model, there was significant activation of inflammatory cells identified by labeling with CD68, which was reduced in the pHBSP-treated animals. The results suggest that pHBSP may improve cognitive function following mTBI.

Neuroprotection with an erythropoietin mimetic peptide (pHBSP) in a model of mild traumatic brain injury complicated by hemorrhagic shock.

Pyroglutamate helix B surface peptide (pHBSP) is an 11 amino acid peptide, designed to interact with a novel cell surface receptor, composed of the classical erythropoietin (EPO) receptor disulfide linked to the beta common receptor. pHBSP has the cytoprotective effects of EPO without stimulating erythropoiesis. Effects on early cerebral hemodynamics and neurological outcome at 2 weeks post-injury were compared in a rat model of mild cortical impact injury (3m/sec, 2.5 mm deformation) followed by 50 min of hemorrhagic hypotension (MAP 40 mm Hg for 50 min). Rats were randomly assigned to receive 5000 U/kg of EPO, 30 µg/kg of pHBSP, or an inactive substance every 12 h for 3 days, starting at the end of resuscitation from the hemorrhagic hypotension, which was 110 min post-injury. Both treatments reduced contusion volume at 2 weeks post-injury, from 20.8±2.8 mm(3) in the control groups to 7.7±2.0 mm(3) in the EPO-treated group and 5.9±1.5 mm(3) in the pHBSP-treated group (p=0.001). Both agents improved recovery of cerebral blood flow in the injured brain following resuscitation, and resulted in more rapid recovery of performance on beam balancing and beam walking tests. These studies suggest that pHBSP has neuroprotective effects similar to EPO in this model of combined brain injury and hypotension. pHBSP may be more useful in the clinical situation because there is less risk of thrombotic adverse effects.

Functional data analysis in brain imaging studies.

Functional data analysis (FDA) considers the continuity of the curves or functions, and is a topic of increasing interest in the statistics community. FDA is commonly applied to time-series and spatial-series studies. The development of functional brain imaging techniques in recent years made it possible to study the relationship between brain and mind over time. Consequently, an enormous amount of functional data is collected and needs to be analyzed. Functional techniques designed for these data are in strong demand. This paper discusses three statistically challenging problems utilizing FDA techniques in functional brain imaging analysis. These problems are dimension reduction (or feature extraction), spatial classification in functional magnetic resonance imaging studies, and the inverse problem in magneto-encephalography studies. The application of FDA to these issues is relatively new but has been shown to be considerably effective. Future efforts can further explore the potential of FDA in functional brain imaging studies.