Narrow Results

A medical emergency, acute ischemic stroke necessitates immediate treatment as the longer the neural is without blood flow, the greater the risk of irreversible neural damage and disability. Depending on the extent, the lesion may vary in size and location and location of the blood flow blockage, and can have significant effects on a person’s cognitive, motor, and sensory function. The process of identifying and classifying regions of affected neural tissue during a ischemic stroke is referred to as “ischemic stroke lesion segmentation.” Medical imaging techniques like computed tomography (CT) and magnetic resonance imaging (MRI) are typically used to accomplish this specialized software tools. The goal of segmentation is to provide a detailed and accurate map of the extent and location of the stroke damage, which can be used to guide treatment decisions and monitor the patient’s recovery. Although manual segmentation is regarded as the most accurate method, it takes time and is subject to variability between observers, which leads to inconsistent results. There are a few unique ways to deal with dividing ischemic stroke sores, ranging from manual delineation by trained clinician to fully automated algorithms that use machine learning and other advanced techniques. Automated segmentation methods are faster and more objective, but they may require large training sets and can be sensitive to imaging artifacts and other sources of noise. Although fully automatic methods hold promise, semi-automatic methods remain the preferred approach in clinical research. We performed a systematic review of the literature to explore the latest advancements and trends in analyzing ischemic stroke lesions using automated methods developed in the past five years. Our search of IEEE explores, Springer, Science Direct, Taylor & Francis, etc. yielded 1580 papers, from which we selected 50 for detailed analysis. Of these studies, 12 employed supervised segmentation, 12 used unsupervised segmentation and 27 employed deep learning segmentation methods. Only a limited number of studies have validated their fully automatic methods using longitudinal samples, and a mere eight studies included validation using clinical parameters. Furthermore, only 23 of the 50 studies made their methods publicly available. To advance the field, there is a need for fully automatic methods that are validated with longitudinal samples and clinical parameters. Moreover, making methods publicly available is essential for promoting reproducibility and facilitating comparison of results.

Motion tracking is a critical task in many computer vision applications. Existing motion tracking techniques require either a great amount of knowledge on the target object or specific hardware. These requirements discourage the wide spread of commercial applications based on motion tracking. In this paper, we present a novel three degrees of freedom motion tracking system that needs no knowledge on the target object and that only requires a single low-budget camera that can be found installed in most computers and smartphones. Our system estimates, in real time, the three-dimensional position of a nonmodeled unmarked object that may be nonrigid, nonconvex, partially occluded, self-occluded, or motion blurred, given that it is opaque, evenly colored, enough contrasting with the background in each frame, and that it does not rotate. Our system is also able to determine the most relevant object to track in the screen. Our proposal does not impose additional constraints, therefore it allows a market-wide implementation of applications that require the estimation of the three position degrees of freedom of an object.

Extracellular neuronal microelectrode recordings can include action potentials from multiple neurons. To separate spikes from different neurons, they can be sorted according to their shape, a procedure referred to as spike-sorting. Several algorithms have been reported to solve this task. However, when clustering outcomes are unsatisfactory, most of them are difficult to adjust to achieve the desired results. We present an online spike-sorting framework that uses feature normalization and weighting to maximize the distinctiveness between different spike shapes. Furthermore, multiple criteria are applied to either facilitate or prevent cluster fusion, thereby enabling experimenters to fine-tune the sorting process. We compare our method to established unsupervised offline (Wave_Clus (WC)) and online (OSort (OS)) algorithms by examining their performance in sorting various test datasets using two different scoring systems (AMI and the Adamos metric). Furthermore, we evaluate sorting capabilities on intra-operative recordings using established quality metrics. Compared to WC and OS, our algorithm achieved comparable or higher scores on average and produced more convincing sorting results for intra-operative datasets. Thus, the presented framework is suitable for both online and offline analysis and could substantially improve the quality of microelectrode-based data evaluation for research and clinical application.

The proportion, number of bouts, and mean bout duration of different vigilance states (Wake, NREM, REM) are useful indices of dynamics in experimental sleep research. These metrics are estimated by first scoring state, sometimes using an algorithm, based on electrophysiological measurements such as the electroencephalogram (EEG) and electromyogram (EMG), and computing their values from the score sequence. Isolated errors in the scores can lead to large discrepancies in the estimated sleep metrics. But most algorithms score sleep by classifying the state from EEG/EMG features independently in each time epoch without considering the dynamics across epochs, which could provide contextual information. The objective here is to improve estimation of sleep metrics by fitting a probabilistic dynamical model to mouse EEG/EMG data and then predicting the metrics from the model parameters. Hidden Markov models (HMMs) with multivariate Gaussian observations and Markov state transitions were fitted to unlabeled 24-h EEG/EMG feature time series from 20 mice to model transitions between the latent vigilance states; a similar model with unbiased transition probabilities served as a reference. Sleep metrics predicted from the HMM parameters did not deviate significantly from manual estimates except for rapid eye movement sleep (REM) ($p<0.01$; Wilcoxon signed-rank test). Changes in value from Light to Dark conditions correlated well with manually estimated differences (Spearman’s rho 0.43–0.84) except for REM. HMMs also scored vigilance state with over 90% accuracy. HMMs of EEG/EMG features can therefore characterize sleep dynamics from EEG/EMG measurements, a prerequisite for characterizing the effects of perturbation in sleep monitoring and control applications.

Learning a fast global model that describes the observed phenomenon well is a crucial goal in the inherently distributed Vehicular Networks. This global model is further used for decision-making, which is especially important for some safety-related applications (i.e., the altering of accident and warning of traffic jam). Most existing works have ignored the network overhead caused by synchronizing with neighbors, which inevitably delays the time for agents to stabilize. In this paper, we focus on developing an asynchronous distributed clustering algorithm to learn the global model, where cluster models, rather than raw data points, are shared and updated. Empirical experiments on a message delay simulator show the efficiency of our methods, with a reduced convergence time, declined network overhead and improved accuracy (relative to the standard solution). This algorithm is further improved by introducing a tolerant delay. Compared to the algorithm without delay, the performance is improved significantly in terms of convergence time (by as much as 47%) and network overhead (by around 53%) if the underlying network is geometric or regular.

SUSTAIN (Supervised and Unsupervised STratified Adaptive Incremental Network) is a network model of human category learning. SUSTAIN initially assumes a simple category structure. If simple solutions prove inadequate and SUSTAIN is confronted with a surprising event (e.g. it is told that a bat is a mammal instead of a bird), SUSTAIN recruits an additional cluster to represent the surprising event. Newly recruited clusters are available to explain future events and can themselves evolve into prototypes/attractors/rules. SUSTAIN has expanded the scope of findings that models of human category learning can address. This paper extends SUSTAIN to account for both supervised and unsupervised learning data through a common mechanism. The modified model, uSUSTAIN (unified SUSTAIN), is successfully applied to human learning data that compares unsupervised and supervised learning performances.¹⁸

Dimension reduction methods including feature selection and feature extraction have played an important role in data mining and pattern recognition. In this study, we propose a novel unsupervised feature selection approach based on sparse representation theory, namely Sparsity Score (SS). Due to the sparse representation procedure, SS not only owns the global property of Variance Score (VS) and the local property of Laplacian Score (LS), but also possesses the discriminating nature. Experimental results, based on three well-known face datasets (Yale, ORL and CMU PIE), reveal that SS performs well in the evaluation of the feature significance, and it significantly outperforms VS and LS.

3D human pose estimation describes estimating 3D articulation structure of a person from an image or a video. The technology has massive potential because it can enable tracking people and analyzing motion in real time. Recently, much research has been conducted to optimize human pose estimation, but few works have focused on reviewing 3D human pose estimation. In this paper, we offer a comprehensive survey of the state-of-the-art methods for 3D human pose estimation, referred to as pose estimation solutions, implementations on images or videos that contain different numbers of people and advanced 3D human pose estimation techniques. Furthermore, different kinds of algorithms are further subdivided into sub-categories and compared in light of different methodologies. To the best of our knowledge, this is the first such comprehensive survey of the recent progress of 3D human pose estimation and will hopefully facilitate the completion, refinement and applications of 3D human pose estimation.

We propose an extension to alignment-free approaches that can produce reasonably accurate phylogenetic groupings starting from unaligned genomes, for example, as fast as 1 min on a standard desktop computer for 25 bacterial genomes. A 6-fold speed-up and 11-fold reduction in memory requirements compared to previous alignment-free methods is achieved by reducing the comparison space to a representative sample of k-mers of optimal length and with specific tag motifs. This approach was applied to the test case of fitting the enterohemorrhagic O104:H4 E.coli strain from the 2011 outbreak in Germany into the phylogenetic network of previously known E.coli-related strains and extend the method to allow assigning any new strain to the correct phylogenetic group even directly from unassembled short sequence reads from next generation sequencing data. Hence, this approach is also useful to quickly identify the most suitable reference genome for subsequent assembly steps.

In recent years, simple Japanese has been attracting attention as information transmission for foreigners. Automatic text simplification aims to reduce the complexity of vocabulary and expressions in a sentence while retaining its original meaning. This paper aims at compressing vocabulary, focusing on lexical simplification. Since the construction or expansion of a simplification corpus is very costly, we construct a simplification model by unsupervised learning that does not require a parallel corpus for simplification. We construct a simplification model that does not require a parallel corpus using Unsupervised Statistical Machine Translation. Based on a predetermined vocabulary, a pseudo-corpus for simplification is constructed from a web corpus and we learn the simplification model by the pseudo-corpus. We only need a vocabulary and a plain text corpus to train the simplification model. Moreover, we propose to clean the phrase table by WordNet, which improves the performance in BLEU and SARI metrics. By suppressing distant paraphrasing with WordNet, it became easier to select the correct paraphrase candidate.

Clustering is primarily used to uncover the true underlying structure of a given data set. Most algorithms for clustering often depend on initial guesses of the cluster centers and assumptions made as to the number of subgroups presents in the data. In this paper, we propose a method for fuzzy clustering without initial guesses on cluster number in the data set. Our method assumes that clusters will have the normal distribution. Our method can automatically estimate the cluster number and form the clusters according to the number. In it, Genetic Algorithms (GAs) with two chromosomic coding techniques are evaluated. Graph structured coding can derive high fitness value. Linear structured can save the number of generation.

Electronic Health Records (EHRs) contain a wealth of patient data useful to biomedical researchers. At present, both the extraction of data and methods for analyses are frequently designed to work with a single snapshot of a patient’s record. Health care providers often perform and record actions in small batches over time. By extracting these care events, a sequence can be formed providing a trajectory for a patient’s interactions with the health care system. These care events also offer a basic heuristic for the level of attention a patient receives from health care providers. We show that is possible to learn meaningful embeddings from these care events using two deep learning techniques, unsupervised autoencoders and long short-term memory networks. We compare these methods to traditional machine learning methods which require a point in time snapshot to be extracted from an EHR.