Semantic Representations Of Objects Research Articles

Using drawings and deep neural networks to characterize the building blocks of human visual similarity.

Early in life and without special training, human beings discern resemblance between abstract visual stimuli, such as drawings, and the real-world objects they represent. We used this capacity for visual abstraction as a tool for evaluating deep neural networks (DNNs) as models of human visual perception. Contrasting five contemporary DNNs, we evaluated how well each explains human similarity judgments among line drawings of recognizable and novel objects. For object sketches, human judgments were dominated by semantic category information; DNN representations contributed little additional information. In contrast, such features explained significant unique variance perceived similarity of abstract drawings. In both cases, a vision transformer trained to blend representations of images and their natural language descriptions showed the greatest ability to explain human perceptual similarity-an observation consistent with contemporary views of semantic representation and processing in the human mind and brain. Together, the results suggest that the building blocks of visual similarity may arise within systems that learn to use visual information, not for specific classification, but in service of generating semantic representations of objects.

Recurrent connectivity supports higher-level visual and semantic object representations in the brain

Visual object recognition has been traditionally conceptualised as a predominantly feedforward process through the ventral visual pathway. While feedforward artificial neural networks (ANNs) can achieve human-level classification on some image-labelling tasks, it’s unclear whether computational models of vision alone can accurately capture the evolving spatiotemporal neural dynamics. Here, we probe these dynamics using a combination of representational similarity and connectivity analyses of fMRI and MEG data recorded during the recognition of familiar, unambiguous objects. Modelling the visual and semantic properties of our stimuli using an artificial neural network as well as a semantic feature model, we find that unique aspects of the neural architecture and connectivity dynamics relate to visual and semantic object properties. Critically, we show that recurrent processing between the anterior and posterior ventral temporal cortex relates to higher-level visual properties prior to semantic object properties, in addition to semantic-related feedback from the frontal lobe to the ventral temporal lobe between 250 and 500 ms after stimulus onset. These results demonstrate the distinct contributions made by semantic object properties in explaining neural activity and connectivity, highlighting it as a core part of object recognition not fully accounted for by current biologically inspired neural networks.

THINGSplus: New norms and metadata for the THINGS database of 1854 object concepts and 26,107 natural object images

To study visual and semantic object representations, the need for well-curated object concepts and images has grown significantly over the past years. To address this, we have previously developed THINGS, a large-scale database of 1854 systematically sampled object concepts with 26,107 high-quality naturalistic images of these concepts. With THINGSplus, we significantly extend THINGS by adding concept- and image-specific norms and metadata for all 1854 concepts and one copyright-free image example per concept. Concept-specific norms were collected for the properties of real-world size, manmadeness, preciousness, liveliness, heaviness, naturalness, ability to move or be moved, graspability, holdability, pleasantness, and arousal. Further, we provide 53 superordinate categories as well as typicality ratings for all their members. Image-specific metadata includes a nameability measure, based on human-generated labels of the objects depicted in the 26,107 images. Finally, we identified one new public domain image per concept. Property (M = 0.97, SD = 0.03) and typicality ratings (M = 0.97, SD = 0.01) demonstrate excellent consistency, with the subsequently collected arousal ratings as the only exception (r = 0.69). Our property (M = 0.85, SD = 0.11) and typicality (r = 0.72, 0.74, 0.88) data correlated strongly with external norms, again with the lowest validity for arousal (M = 0.41, SD = 0.08). To summarize, THINGSplus provides a large-scale, externally validated extension to existing object norms and an important extension to THINGS, allowing detailed selection of stimuli and control variables for a wide range of research interested in visual object processing, language, and semantic memory.

The human language effective connectome

To advance understanding of brain networks involved in language, the effective connectivity between 26 cortical regions implicated in language by a community analysis and 360 cortical regions was measured in 171 humans from the Human Connectome Project, and complemented with functional connectivity and diffusion tractography, all using the HCP multimodal parcellation atlas. A (semantic) network (Group 1) involving inferior cortical regions of the superior temporal sulcus cortex (STS) with the adjacent inferior temporal visual cortex TE1a and temporal pole TG, and the connected parietal PGi region, has effective connectivity with inferior temporal visual cortex (TE) regions; with parietal PFm which also has visual connectivity; with posterior cingulate cortex memory-related regions; with the frontal pole, orbitofrontal cortex, and medial prefrontal cortex; with the dorsolateral prefrontal cortex; and with 44 and 45 for output regions. It is proposed that this system can build in its temporal lobe (STS and TG) and parietal parts (PGi and PGs) semantic representations of objects incorporating especially their visual and reward properties. Another (semantic) network (Group 3) involving superior regions of the superior temporal sulcus cortex and more superior temporal lobe regions including STGa, auditory A5, TPOJ1, the STV and the Peri-Sylvian Language area (PSL) has effective connectivity with auditory areas (A1, A4, A5, Pbelt); with relatively early visual areas involved in motion, e.g., MT and MST, and faces/words (FFC); with somatosensory regions (frontal opercular FOP, insula and parietal PF); with other TPOJ regions; and with the inferior frontal gyrus regions (IFJa and IFSp). It is proposed that this system builds semantic representations specialising in auditory and related facial motion information useful in theory of mind and somatosensory / body image information, with outputs directed not only to regions 44 and 45, but also to premotor 55b and midcingulate premotor cortex. Both semantic networks (Groups 1 and 3) have access to the hippocampal episodic memory system via parahippocampal TF. A third largely frontal network (Group 2) (44, 45, 47l; 55b; the Superior Frontal Language region SFL; and including temporal pole TGv) receives effective connectivity from the two semantic systems, and is implicated in syntax and speech output.

Introducing the structural bases of typicality effects in deep learning

In this paper, we hypothesize that the effects of the degree of typicality in natural semantic categories can be generated based on the structure of artificial categories learned with deep learning models. Motivated by the human approach to representing natural semantic categories and based on the Prototype Theory foundations, we propose a novel Computational Prototype Model (CPM) to represent the internal structure of semantic categories. Unlike other prototype learning approaches, our mathematical framework proposes a first approach to provide deep neural networks with the ability to model abstract semantic concepts such as category central semantic meaning, typicality degree of an object's image, and family resemblance relationship. We proposed several methodologies based on the typicality's concept to evaluate our CPM-model in image semantic processing tasks such as image classification, a global semantic description, and transfer learning. Our experiments on different image datasets, such as ImageNet and Coco, showed that our approach might be an admissible proposition in the effort to endow machines with greater power of abstraction for the semantic representation of objects' categories.

Impaired integration of object feature knowledge in amnestic mild cognitive impairment.

Accessing semantic representations of real-world objects requires integration of multimodal perceptual features that are represented across relevant neocortical areas. Early Alzheimer's disease (AD) neuropathology, including neurofibrillary tangles in the perirhinal cortex as well as disrupted cortico-cortical connectivity, would be expected to disrupt the integration of object features. This integration deficit may underlie AD patients' semantic memory deficits and would be predicted to be more prominent for living objects, which tend to be more defined by sensory features compared with nonliving objects. Two experiments were conducted to assess feature integration in cognitively healthy older adults and patients with amnestic mild cognitive impairment (MCI). In both experiments, pictures of real-world objects were presented in congruent or incongruent colors. Participants were instructed to make a speeded color congruency judgment (Experiment 1) or name the presented surface color (Experiment 2). Across experiments, MCI patients showed a selective integration deficit for living, but not nonliving, objects across both experimental paradigms that was consistent with a deterioration in semantic structural representations rather than a deficit in controlled semantic retrieval. Planned secondary analyses with a subset of patients (Experiment 1) for whom PET imaging was available indicated that the degree of impairment was associated with the magnitude of cortical amyloid burden. These findings suggest that early AD pathology leads to impaired integration of distributed semantic object representations. The development of integration tasks as sensitive markers of early AD pathology may lead to more effective diagnostic tools for early detection and intervention. (PsycInfo Database Record (c) 2020 APA, all rights reserved).

Improved Visual-Semantic Alignment for Zero-Shot Object Detection

Zero-shot object detection is an emerging research topic that aims to recognize and localize previously ‘unseen’ objects. This setting gives rise to several unique challenges, e.g., highly imbalanced positive vs. negative instance ratio, proper alignment between visual and semantic concepts and the ambiguity between background and unseen classes. Here, we propose an end-to-end deep learning framework underpinned by a novel loss function that handles class-imbalance and seeks to properly align the visual and semantic cues for improved zero-shot learning. We call our objective the ‘Polarity loss’ because it explicitly maximizes the gap between positive and negative predictions. Such a margin maximizing formulation is not only important for visual-semantic alignment but it also resolves the ambiguity between background and unseen objects. Further, the semantic representations of objects are noisy, thus complicating the alignment between visual and semantic domains. To this end, we perform metric learning using a ‘Semantic vocabulary’ of related concepts that refines the noisy semantic embeddings and establishes a better synergy between visual and semantic domains. Our approach is inspired by the embodiment theories in cognitive science, that claim human semantic understanding to be grounded in past experiences (seen objects), related linguistic concepts (word vocabulary) and the visual perception (seen/unseen object images). Our extensive results on MS-COCO and Pascal VOC datasets show significant improvements over state of the art.1

Auditory and Semantic Cues Facilitate Decoding of Visual Object Category in MEG.

Sounds (e.g., barking) help us to visually identify objects (e.g., a dog) that are distant or ambiguous. While neuroimaging studies have revealed neuroanatomical sites of audiovisual interactions, little is known about the time course by which sounds facilitate visual object processing. Here we used magnetoencephalography to reveal the time course of the facilitatory influence of natural sounds (e.g., barking) on visual object processing and compared this to the facilitatory influence of spoken words (e.g., "dog"). Participants viewed images of blurred objects preceded by a task-irrelevant natural sound, a spoken word, or uninformative noise. A classifier was trained to discriminate multivariate sensor patterns evoked by animate and inanimate intact objects with no sounds, presented in a separate experiment, and tested on sensor patterns evoked by the blurred objects in the 3 auditory conditions. Results revealed that both sounds and words, relative to uninformative noise, significantly facilitated visual object category decoding between 300-500ms after visual onset. We found no evidence for earlier facilitation by sounds than by words. These findings provide evidence for a semantic route of facilitation by both natural sounds and spoken words, whereby the auditory input first activates semantic object representations, which then modulate the visual processing of objects.

The relationship between response consistency in picture naming and storage impairment in people with semantic variant primary progressive aphasia.

The progressive loss of stored knowledge about word meanings in semantic variant primary progressive aphasia (svPPA) has been attributed to an amodal "storage" deficit of the semantic system. Performance consistency has been proposed to be a key characteristic of storage deficits but has not been examined in close detail and larger participant cohorts. We assessed whether 10 people with svPPA showed consistency in picture naming across 3 closely consecutive sessions. We examined item-by-item consistency of naming accuracy and specific error types, while controlling for the effects of word-related variables such as word frequency, familiarity, and age of acquisition. Participants were very consistent in their accurate and inaccurate responses over and above any effects of the word-related variables. Analyses of error types that compared consistency of semantic errors, correct responses, and other error types (e.g., phonologically related errors, unrelated errors) revealed lower consistency. Our findings support the assumption that semantic features constituting semantic representations of objects are progressively lost in people with svPPA and are therefore consistently unavailable during naming. Variability in the production of error types occurs when distinctive features of an object are lost, resulting in the selection of semantically or visually similar items or in the failure to select an item and a response omission. The assessment of performance consistency sheds light on the underlying impairment of people with semantic deficits (semantic storage vs. access deficit). This can support the choice of an appropriate treatment technique to maintain or reteach semantic information. (PsycINFO Database Record (c) 2019 APA, all rights reserved).

MODELING CITIES FOR 3D_GIS PURPOSES

Abstract. 3D Geographic Information Systems (3D GIS) are systems that are capable of making spatial analyses that consider the tridimentional and semantic representation of objects. These systems make these analyses through its planialtimetric coordinates. The City Geography Markup Language (CityGML) is used for the representation of cities and urban applications. The CityGML is an international standardized data model based on XML used to store and exchange information through 3D representation of cities. This standardized data model has 5 Levels of Detail – LOD, varying from LOD 0 (least detailed) to 4 (most detailed). The main challenges for the implementation of these systems refer to the techniques used for obtaining data and the data format, and also all the software used in the geometric modeling of the urban model. The data related to the buildings were manipulated with the QGIS software in this study. This made it possible to obtain the height of the buildings by the elevation difference between the Digital Surface Model and the Digital Terrain Model. This paper presents and discusses the first results of the geometric modeling made in the campus of the Federal University of Bahia (UFBA), by using airborne laser scanning data, integrating QGIS, Rhinoceros and CityGML.

A feature-based neurocomputational model of semantic memory.

According with a featural organization of semantic memory, this work is aimed at investigating, through an attractor network, the role of different kinds of features in the representation of concepts, both in normal and neurodegenerative conditions. We implemented new synaptic learning rules in order to take into account the role of partially shared features and of distinctive features with different saliency. The model includes semantic and lexical layers, coding, respectively for object features and word-forms. Connections among nodes are strongly asymmetrical. To account for the feature saliency, asymmetrical synapses were created using Hebbian rules of potentiation and depotentiation, setting different pre-synaptic and post-synaptic thresholds. A variable post-synaptic threshold, which automatically changed to reflect the feature frequency in different concepts (i.e., how many concepts share a feature), was used to account for partially shared features. The trained network solved naming tasks and word recognition tasks very well, exploiting the different role of salient versus marginal features in concept identification. In the case of damage, superordinate concepts were preserved better than the subordinate ones. Interestingly, the degradation of salient features, but not of marginal ones, prevented object identification. The model suggests that Hebbian rules, with adjustable post-synaptic thresholds, can provide a reliable semantic representation of objects exploiting the statistics of input features.

Sensory and semantic activations evoked by action attributes of manipulable objects: Evidence from ERPs

“Two route” theories of object-related action processing posit different temporal activation profiles of grasp-to-move actions (rapidly evoked based on object structure) versus skilled use actions (more slowly activated based on semantic knowledge). We capitalized on the exquisite temporal resolution and multidimensionality of Event-Related Potentials (ERPs) to directly test this hypothesis. Participants viewed manipulable objects (e.g., calculator) preceded by objects sharing either “grasp”, “use”, or no action attributes (e.g., bar of soap, keyboard, earring, respectively), as well as by action-unrelated but taxonomically-related objects (e.g., abacus); participants judged whether the two objects were related. The results showed more positive responses to “grasp-to-move” primed objects than “skilled use” primed objects or unprimed objects starting in the P1 (0–150 ms) time window and continuing onto the subsequent N1 and P2 components (150–300 ms), suggesting that only “grasp-to-move”, but not “skilled use”, actions may facilitate visual attention to object attributes. Furthermore, reliably reduced N400s (300–500 ms), an index of semantic processing, were observed to taxonomically primed and “skilled use” primed objects relative to unprimed objects, suggesting that “skilled use” action attributes are a component of distributed, multimodal semantic representations of objects. Together, our findings provide evidence supporting two-route theories by demonstrating that “grasp-to-move” and “skilled use” actions impact different aspects of object processing and highlight the relationship of “skilled use” information to other aspects of semantic memory.

Understanding mid-level representations in visual processing.

It is clear that early visual processing provides an image-based representation of the visual scene: Neurons in Striate cortex (V1) encode nothing about the meaning of a scene, but they do provide a great deal of information about the image features within it. The mechanisms of these "low-level" visual processes are relatively well understood. We can construct plausible models for how neurons, up to and including those in V1, build their representations from preceding inputs down to the level of photoreceptors. It is also clear that at some point we have a semantic, "high-level" representation of the visual scene because we can describe verbally the objects that we are viewing and their meaning to us. A huge number of studies are examining these "high-level" visual processes each year. Less well studied are the processes of "mid-level" vision, which presumably provide the bridge between these "low-level" representations of edges, colors, and lights and the "high-level" semantic representations of objects, faces, and scenes. This article and the special issue of papers in which it is published consider the nature of "mid-level" visual processing and some of the reasons why we might not have made as much progress in this domain as we would like.

Large-scale weakly supervised object localization via latent category learning.

Localizing objects in cluttered backgrounds is challenging under large-scale weakly supervised conditions. Due to the cluttered image condition, objects usually have large ambiguity with backgrounds. Besides, there is also a lack of effective algorithm for large-scale weakly supervised localization in cluttered backgrounds. However, backgrounds contain useful latent information, e.g., the sky in the aeroplane class. If this latent information can be learned, object-background ambiguity can be largely reduced and background can be suppressed effectively. In this paper, we propose the latent category learning (LCL) in large-scale cluttered conditions. LCL is an unsupervised learning method which requires only image-level class labels. First, we use the latent semantic analysis with semantic object representation to learn the latent categories, which represent objects, object parts or backgrounds. Second, to determine which category contains the target object, we propose a category selection strategy by evaluating each category's discrimination. Finally, we propose the online LCL for use in large-scale conditions. Evaluation on the challenging PASCAL Visual Object Class (VOC) 2007 and the large-scale imagenet large-scale visual recognition challenge 2013 detection data sets shows that the method can improve the annotation precision by 10% over previous methods. More importantly, we achieve the detection precision which outperforms previous results by a large margin and can be competitive to the supervised deformable part model 5.0 baseline on both data sets.

The role of the sound of objects in object identification: evidence from picture naming.

In the present work we were concerned with the role of sound representations in object recognition. In order to address this issue we made use of a picture naming task in which target pictures might be accompanied by a white-noise burst. White-noise was thought to interfere with the representation of the sound possibly associated with the depicted object. We reasoned that if such a representation is critical for the recognition of objects strongly associated with certain sounds, white-noise interference should affect the naming of pictures representing objects with typical sounds leaving the naming of object without typical sounds unaffected. The results were congruent with the predictions and consistent with a view of the semantic representations of objects as collection of related representations, modal in nature, and mandatorily accessed.

Active Monitoring and Control of Light Fixtures during Building Construction and Operation: Cyber-Physical Systems Approach

Digitally addressable lighting systems offer tremendous opportunities for performance monitoring and control of individual light fixtures. However, the locations of individual light fixtures within a building are not easily differentiable; as such, facility managers cannot distinguish and control each fixture. Virtual models contain semantic representations of objects that enable users to visually identify, distinguish, and interactively monitor building components, thus providing opportunities for controlling components. To monitor and control light fixtures from these models, each physical fixture needs to be tightly integrated with its virtual representation in order to enable bidirectional coordination. Bidirectional coordination between virtual models and physical light fixtures offers tremendous opportunities for enhancing progress monitoring during construction and performance monitoring and control during the operations and maintenance phase of a building’s life cycle. This paper presents an approach to facilitating this through cyber-physical systems integration between virtual models and physical light fixtures. The importance of actively monitoring and controlling light fixtures is highlighted. A cyber-physical systems architecture that illustrates the key features, enabling technologies and their roles, is presented. The paper also presents a prototype system that has been developed and implemented to demonstrate the functionality of the proposed approach.

Learning semantic representations of objects and their parts

Recently, large scale image annotation datasets have been collected with millions of images and thousands of possible annotations. Latent variable models, or embedding methods, that simultaneously learn semantic representations of object labels and image representations can provide tractable solutions on such tasks. In this work, we are interested in jointly learning representations both for the objects in an image, and the parts of those objects, because such deeper semantic representations could bring a leap forward in image retrieval or browsing. Despite the size of these datasets, the amount of annotated data for objects and parts can be costly and may not be available. In this paper, we propose to bypass this cost with a method able to learn to jointly label objects and parts without requiring exhaustively labeled data. We design a model architecture that can be trained under a proxy supervision obtained by combining standard image annotation (from ImageNet) with semantic part-based within-label relations (from WordNet). The model itself is designed to model both object image to object label similarities, and object label to object part label similarities in a single joint system. Experiments conducted on our combined data and a precisely annotated evaluation set demonstrate the usefulness of our approach.

Speaking of shape: The effects of language-specific encoding on semantic representations

AbstractThe question of whether different linguistic patterns differentially influence semantic and conceptual representations is of central interest in cognitive science. In this paper, we investigate whether the regular encoding of shape within a nominal classification system leads to an increased salience of shape in speakers' semantic representations by comparing English, (Amazonian) Spanish, and Bora, a shape-based classifier language spoken in the Amazonian regions of Columbia and Peru. Crucially, in displaying obligatory use, pervasiveness in grammar, high discourse frequency, and phonological variability of forms corresponding to particular shape features, the Bora classifier system differs in important ways from those in previous studies investigating effects of nominal classification, thereby allowing better control of factors that may have influenced previous findings. In addition, the inclusion of Spanish monolinguals living in the Bora village allowed control for the possibility that differences found between English and Bora speakers may be attributed to their very different living environments. We found that shape is more salient in the semantic representation of objects for speakers of Bora, which systematically encodes shape, than for speakers of English and Spanish, which do not. Our results are consistent with assumptions that semantic representations are shaped and modulated by our specific linguistic experiences.

A lifespan perspective on semantic processing of concrete concepts: does a sensory/motor model have the potential to bridge the gap?

Research regarding semantic knowledge of objects is often conducted independently in children and adults. Review of these bodies of evidence suggests that the two literatures are often complementary. It seems critical to determine what we can learn from a developmental perspective, toward the common goal of understanding semantic organization. Here we focus on the proposal that semantic knowledge about concrete concepts may be built on the foundation of sensory/motor processes. In particular, we focus on a moderate formulation of this viewpoint, the sensory/motor model of semantic representations of objects (e.g., Gainotti 2007; Martin 2007), which has been examined utilizing behavioral, neuroimaging, and neuropsychological evidence. Taken together, behavioral and neuroimaging studies with infants, older children, and adults have suggested that patterns laid down in early childhood remain salient throughout the lifespan and may also predict patterns of deficit that emerge following brain injury.

Perceiving the tree in the woods: separating brain responses to coincidentally presented scenes and objects

Event Abstract Back to Event Perceiving the tree in the woods: separating brain responses to coincidentally presented scenes and objects Ulla Martens1* and Thomas Gruber1 1 University of Osnabrück, Germany Due to methodological problems in separating the brain responses to multiple stimuli, there is a profound lack of electrophysiological studies investigating object recognition in the context of natural scenes. We overcome this limitation by tagging coincidentally presented natural scenes and objects to different presentation frequencies. In due course, the scenes produced a different steady-state visual evoked potential (SSVEP) than the objects. We analyzed the effect of semantic (in)-coherency between natural scenes and a salient object on SSVEP amplitudes and topography. Additionally, the neuronal generators of the coherency effect were analyzed by means of Variable Resolution Electromagnetic Tomography (VARETA). Results showed: 1) An attentional focus on the scene in coherent and on the object in incoherent conditions. 2) Right inferior and middle frontal neuronal sources of the incoherency effect indicate a top-down attentional shift from the scene to the object. 3) Lateralization effects in the anterior temporal lobe indicate right hemispheric scene processing and left hemispheric object processing. Taken together, we state that the SSVEP is an ideal tool to investigate multiple stimuli displays with an excellent signal-to-noise ratio to investigate neuronal generators of higher cognitive functions. The attentional advantage of incoherent objects in a scene is likely the result of a mismatch between scene-based expectations (i.e. activated concepts) and activated semantic object representations. Keywords: Perception, SSVEP Conference: XI International Conference on Cognitive Neuroscience (ICON XI), Palma, Mallorca, Spain, 25 Sep - 29 Sep, 2011. Presentation Type: Poster Presentation Topic: Poster Sessions: Neurophysiology of Sensation and Perception Citation: Martens U and Gruber T (2011). Perceiving the tree in the woods: separating brain responses to coincidentally presented scenes and objects. Conference Abstract: XI International Conference on Cognitive Neuroscience (ICON XI). doi: 10.3389/conf.fnhum.2011.207.00335 Copyright: The abstracts in this collection have not been subject to any Frontiers peer review or checks, and are not endorsed by Frontiers. They are made available through the Frontiers publishing platform as a service to conference organizers and presenters. The copyright in the individual abstracts is owned by the author of each abstract or his/her employer unless otherwise stated. Each abstract, as well as the collection of abstracts, are published under a Creative Commons CC-BY 4.0 (attribution) licence (https://creativecommons.org/licenses/by/4.0/) and may thus be reproduced, translated, adapted and be the subject of derivative works provided the authors and Frontiers are attributed. For Frontiers’ terms and conditions please see https://www.frontiersin.org/legal/terms-and-conditions. Received: 23 Nov 2011; Published Online: 28 Nov 2011. * Correspondence: Dr. Ulla Martens, University of Osnabrück, Osnabrück, Germany, umartens@uni-osnabrueck.de Login Required This action requires you to be registered with Frontiers and logged in. To register or login click here. Abstract Info Abstract The Authors in Frontiers Ulla Martens Thomas Gruber Google Ulla Martens Thomas Gruber Google Scholar Ulla Martens Thomas Gruber PubMed Ulla Martens Thomas Gruber Related Article in Frontiers Google Scholar PubMed Abstract Close Back to top Javascript is disabled. Please enable Javascript in your browser settings in order to see all the content on this page.