Common-sense reasoning for human action recognition

This paper presents a novel method that leverages reasoning capabilities in a computer vision system dedicated to human action recognition. The proposed methodology is decomposed into two stages. First, a machine learning based algorithm - known as bag of words - gives a first estimate of action classification from video sequences, by performing an image feature analysis. Those results are afterward passed to a common-sense reasoning system, which analyses, selects and corrects the initial estimation yielded by the machine learning algorithm. This second stage resorts to the knowledge implicit in the rationality that motivates human behaviour. Experiments are performed in realistic conditions, where poor recognition rates by the machine learning techniques are significantly improved by the second stage in which common-sense knowledge and reasoning capabilities have been leveraged. This demonstrates the value of integrating common-sense capabilities into a computer vision pipeline. © 2012 Elsevier B.V. All rights reserved.

Respiratory syncytial virus interaction with human airway epithelium

Although respiratory syncytial virus (RSV) is a major human respiratory pathogen, our knowledge of how it causes disease in humans is limited. Airway epithelial cells are the primary targets of RSV infection in vivo, so the generation and exploitation of RSV infection models based on morphologically and physiologically authentic well-differentiated primary human airway epithelial cells cultured at an air-liquid interface (WD-PAECs) provide timely developments that will help to bridge this gap. Here we review the interaction of RSV with WD-PAEC cultures, the authenticity of the RSV-WD-PAEC models relative to RSV infection of human airway epithelium in vivo, and future directions for their exploitation in our quest to understand RSV pathogenesis in humans.

Parallels between machine and brain decoding

We report some existing work, inspired by analogies between human thought and machine computation, showing that the informational state of a digital computer can be decoded in a similar way to brain decoding. We then discuss some proposed work that would leverage this analogy to shed light on the amount of information that may be missed by the technical limitations of current neuroimaging technologies. © 2012 Springer-Verlag.

Effect of the crown design and interface lute parameters on the stress-state of a machined crown-tooth system: A finite element analysis

The effect of preparation design and the physical properties of the interface lute on the restored machined ceramic crown-tooth complex are poorly understood. The aim of this work was to determine, by means of three-dimensional finite element analysis (3D FEA) the effect of the tooth preparation design and the elastic modulus of the cement on the stress state of the cemented machined ceramic crown-tooth complex. The three-dimensional structure of human premolar teeth, restored with adhesively cemented machined ceramic crowns, was digitized with a micro-CT scanner. An accurate, high resolution, digital replica model of a restored tooth was created. Two preparation designs, with different occlusal morphologies, were modeled with cements of 3 different elastic moduli. Interactive medical image processing software (mimics and professional CAD modeling software) was used to create sophisticated digital models that included the supporting structures; periodontal ligament and alveolar bone. The generated models were imported into an FEA software program (hypermesh version 10.0, Altair Engineering Inc.) with all degrees of freedom constrained at the outer surface of the supporting cortical bone of the crown-tooth complex. Five different elastic moduli values were given to the adhesive cement interface 1.8 GPa, 4 GPa, 8 GPa, 18.3 GPa and 40 GPa; the four lower values are representative of currently used cementing lutes and 40 GPa is set as an extreme high value. The stress distribution under simulated applied loads was determined. The preparation design demonstrated an effect on the stress state of the restored tooth system. The cement elastic modulus affected the stress state in the cement and dentin structures but not in the crown, the pulp, the periodontal ligament or the cancellous and cortical bone. The results of this study suggest that both the choice of the preparation design and the cement elastic modulus can affect the stress state within the restored crown-tooth complex.

Recombinant Subgroup B Human Respiratory Syncytial Virus Expressing Enhanced Green Fluorescent Protein Efficiently Replicates in Primary Human Cells and Is Virulent in Cotton Rats

Human respiratory syncytial virus (HRSV) is the most important viral cause of severe respiratory tract disease in infants. Two subgroups (A and B) have been identified, which cocirculate during, or alternate between, yearly epidemics and cause indistinguishable disease. Existing in vitro and in vivo models of HRSV focus almost exclusively on subgroup A viruses. Here, a recombinant (r) subgroup B virus (rHRSV(B05)) was generated based on a consensus genome sequence obtained directly from an unpassaged clinical specimen from a hospitalized infant. An additional transcription unit containing the gene encoding enhanced green fluorescent protein (EGFP) was introduced between the phosphoprotein and matrix genes (position 5) of the genome to generate rHRSV(B05)EGFP(5). The recombinant viruses replicated efficiently in both HEp-2 cells and in well-differentiated normal human bronchial cells grown at air-liquid interface. Intranasal infection of cotton rats (Sigmodon hispidus) resulted in high numbers of EGFP(+) cells in epithelia of the nasal septum and conchae. When administered in a relatively large inoculum volume, the virus also replicated efficiently in bronchiolar epithelial cells and spread extensively in both the upper and lower respiratory tracts. Virus replication was not observed in ciliated epithelial cells of the trachea. This is the first virulent rHRSV strain with the genetic composition of a currently circulating wild-type virus. In vivo tracking of infected cells by means of EGFP fluorescence in the absence of cytopathic changes increases the sensitivity of virus detection in HRSV pathogenesis studies.

IMPORTANCE

Virology as a discipline has depended on monitoring cytopathic effects following virus culture in vitro. However, wild-type viruses isolated from patients often do not cause significant changes to infected cells, necessitating blind passage. This can lead to genetic and phenotypic changes and the generation of high-titer, laboratory-adapted viruses with diminished virulence in animal models of disease. To address this, we determined the genome sequence of an unpassaged human respiratory syncytial virus from a sample obtained directly from an infected infant, assembled a molecular clone, and recovered a wild-type recombinant virus. Addition of a gene encoding enhanced green fluorescent protein allowed this wild-type virus to be tracked in primary human cells and living animals in the absence of significant cytopathic effects. Imaging of fluorescent cells proved to be a highly valuable tool for monitoring the spread of virus and may help improve assays for evaluating novel intervention strategies.

Machine learning for transient discovery in Pan-STARRS1 difference imaging

Efficient identification and follow-up of astronomical transients is hindered by the need for humans to manually select promising candidates from data streams that contain many false positives. These artefacts arise in the difference images that are produced by most major ground-based time-domain surveys with large format CCD cameras. This dependence on humans to reject bogus detections is unsustainable for next generation all-sky surveys and significant effort is now being invested to solve the problem computationally. In this paper, we explore a simple machine learning approach to real-bogus classification by constructing a training set from the image data of similar to 32 000 real astrophysical transients and bogus detections from the Pan-STARRS1 Medium Deep Survey. We derive our feature representation from the pixel intensity values of a 20 x 20 pixel stamp around the centre of the candidates. This differs from previous work in that it works directly on the pixels rather than catalogued domain knowledge for feature design or selection. Three machine learning algorithms are trained (artificial neural networks, support vector machines and random forests) and their performances are tested on a held-out subset of 25 per cent of the training data. We find the best results from the random forest classifier and demonstrate that by accepting a false positive rate of 1 per cent, the classifier initially suggests a missed detection rate of around 10 per cent. However, we also find that a combination of bright star variability, nuclear transients and uncertainty in human labelling means that our best estimate of the missed detection rate is approximately 6 per cent.