Alzheimer’s Disease (AD) and other neurodegenerative diseases affect over 20 million people worldwide, and this number is projected to significantly increase in the coming decades. Proposed imaging-based markers have shown steadily improving levels of sensitivity/specificity in classifying individual subjects as AD or normal. Several of these efforts have utilized statistical machine learning techniques, using brain images as input, as means of deriving such AD-related markers. A common characteristic of this line of research is a focus on either (1) using a single imaging modality for classification, or (2) incorporating several modalities, but reporting separate results for each. One strategy to improve on the success of these methods is to leverage all available imaging modalities together in a single automated learning framework. The rationale is that some subjects may show signs of pathology in one modality but not in another – by combining all available images a clearer view of the progression of disease pathology will emerge. Our method is based on the Multi-Kernel Learning (MKL) framework, which allows the inclusion of an arbitrary number of views of the data in a maximum margin, kernel learning framework. The principal innovation behind MKL is that it learns an optimal combination of kernel (similarity) matrices while simultaneously training a classifier. In classification experiments MKL outperformed an SVM trained on all available features by 3% – 4%. We are especially interested in whether such markers are capable of identifying early signs of the disease. To address this question, we have examined whether our multi-modal disease marker (MMDM) can predict conversion from Mild Cognitive Impairment (MCI) to AD. Our experiments reveal that this measure shows significant group differences between MCI subjects who progressed to AD, and those who remained stable for 3 years. These differences were most significant in MMDMs based on imaging data. We also discuss the relationship between our MMDM and an individual’s conversion from MCI to AD.

This paper is focused on the Co-segmentation problem [1] -where the objective is to segment a similar object from a pair of images. The background in the two images may be arbitrary; therefore, simultaneous segmentation of both images must be performed with a requirement that the appearance of the two sets of foreground pixels in the respective images are consistent. Existing approaches [1,2] cast this problem as a Markov Random Field (MRF) based segmentation of the image pair with a regularized difference of the two histograms -assuming a Gaussian prior on the foreground appearance [1] or by calculating the sum of squared differences [2]. Both are interesting formulations but lead to difficult optimization problems, due to the presence of the second (histogram difference) term. The model proposed here bypasses measurement of the histogram differences in a direct fashion; we show that this enables obtaining efficient solutions to the underlying optimization model. Our new algorithm is similar to the existing methods in spirit, but differs substantially in that it can be solved to optimality in polynomial time using a maximum flow procedure on an appropriately constructed graph. We discuss our ideas and present promising experimental results.

Structural and functional brain images are playing an important role in helping us understand the changes associated with neurological disorders such as Alzheimer's disease (AD). Recent efforts have now started investigating their utility for diagnosis purposes. This line of research has shown promising results where methods from machine learning (such as Support Vector Machines) have been used to identify AD-related patterns from images, for use in diagnosing new individual subjects. In this paper, we propose a new framework for AD classification which makes use of the Linear Program (LP) boosting with novel additional regularization based on spatial "smoothness" in 3D image coordinate spaces. The algorithm formalizes the expectation that since the examples for training the classifier are images, the voxels eventually selected for specifying the decision boundary must constitute spatially contiguous chunks, i.e., "regions" must be preferred over isolated voxels. This prior belief turns out to be useful for significantly reducing the space of possible classifiers and leads to substantial benefits in generalization. In our method, the requirement of spatial contiguity (of selected discriminating voxels) is incorporated within the optimization framework directly. Other methods have made use of similar biases as a pre-or post-processing step, however, our model incorporates this emphasis on spatial smoothness directly into the learning step. We report on extensive evaluations of our algorithm on MR and FDG-PET images from the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset, and discuss the relationship of the classification output with the clinical and cognitive biomarker data available within ADNI.

We study the problem of online subspace learning in the context of sequential observations involving structured perturbations. In online subspace learning, the observations are an unknown mixture of two components presented to the model sequentially -the main effect which pertains to the subspace and a residual/error term. If no additional requirement is imposed on the residual, it often corresponds to noise terms in the signal which were unaccounted for by the main effect. To remedy this, one may impose 'structural' contiguity, which has the intended effect of leveraging the secondary terms as a covariate that helps the estimation of the subspace itself, instead of merely serving as a noise residual. We show that the corresponding online estimation procedure can be written as an approximate optimization process on a Grassmannian. We propose an efficient numerical solution, GOSUS, Grassmannian Online Subspace Updates with Structured-sparsity, for this problem. GOSUS is expressive enough in modeling both homogeneous perturbations of the subspace and structural contiguities of outliers, and after certain manipulations, solvable via an alternating direction method of multipliers (ADMM). We evaluate the empirical performance of this algorithm on two problems of interest: online background subtraction and online multiple face tracking, and demonstrate that it achieves competitive performance with the state-of-the-art in near real time.

We study the cosegmentation problem where the objective is to segment the same object (i.e., region) from a pair of images. The segmentation for each image can be cast using a partitioning/segmentation function with an additional constraint that seeks to make the histograms of the segmented regions (based on intensity and texture features) similar. Using Markov Random Field (MRF) energy terms for the simultaneous segmentation of the images together with histogram consistency requirements using the squared L 2 (rather than L 1) distance, after linearization and adjustments, yields an optimization model with some interesting combinatorial properties. We discuss these properties which are closely related to certain relaxation strategies recently introduced in computer vision. Finally, we show experimental results of the proposed approach.

With the proliferation of wearable cameras, the number of videos of users documenting their personal lives using such devices is rapidly increasing. Since such videos may span hours, there is an important need for mechanisms that represent the information content in a compact form (i.e., shorter videos which are more easily browsable/sharable). Motivated by these applications, this paper focuses on the problem of egocentric video summarization. Such videos are usually continuous with significant camera shake and other quality issues. Because of these reasons, there is growing consensus that direct application of standard video summarization tools to such data yields unsatisfactory performance. In this paper, we demonstrate that using gaze tracking information (such as fixation and saccade) significantly helps the summarization task. It allows meaningful comparison of different image frames and enables deriving personalized summaries (gaze provides a sense of the camera wearer's intent). We formulate a summarization model which captures common-sense properties of a good summary, and show that it can be solved as a submodular function maximization with partition matroid constraints, opening the door to a rich body of work from combinatorial optimization. We evaluate our approach on a new gaze-enabled egocentric video dataset (over 15 hours), which will be a valuable standalone resource.

Our primary interest is in generalizing the problem of Cosegmentation to a large group of images, that is, concurrent segmentation of common foreground region(s) from multiple images. We further wish for our algorithm to offer scale invariance (foregrounds may have arbitrary sizes in different images) and the running time to increase (no more than) near linearly in the number of images in the set. What makes this setting particularly challenging is that even if we ignore the scale invariance desiderata, the Cosegmentation problem, as formalized in many recent papers (except [1]), is already hard to solve optimally in the two image case. A straightforward extension of such models to multiple images leads to loose relaxations; and unless we impose a distributional assumption on the appearance model, existing mechanisms for image-pair-wise measurement of foreground appearance variations lead to significantly large problem sizes (even for moderate number of images). This paper presents a surprisingly easy to implement algorithm which performs well, and satisfies all requirements listed above (scale invariance, low computational requirements, and viability for the multiple image setting). We present qualitative and technical analysis of the properties of this framework.

The brain at rest consists of spatially distributed but functionally connected regions, called intrinsic connectivity networks (ICNs). Resting state functional magnetic resonance imaging (rs-fMRI) has emerged as a way to characterize brain networks without confounds associated with task fMRI such as task difficulty and performance. Here we applied a Support Vector Machine (SVM) linear classifier as well as a support vector machine regressor to rs-fMRI data in order to compare age-related differences in four of the major functional brain networks: the default, cingulo-opercular, fronto-parietal, and sensorimotor. A linear SVM classifier discriminated between young and old subjects with 84% accuracy (p-value < 1 × 10−7). A linear SVR age predictor performed reasonably well in continuous age prediction (R2 = 0.419, p-value < 1 × 10−8). These findings reveal that differences in intrinsic connectivity as measured with rs-fMRI exist between subjects, and that SVM methods are capable of detecting and utilizing these differences for classification and prediction.