Deep Artificial Neural Networks Reveal a Distributed Cortical Network Encoding Propositional Sentence-Level Meaning

Understanding how and where in the brain sentence-level meaning is constructed from words presents a major scientific challenge. Recent advances have begun to explain brain activation elicited by sentences using vector models of word meaning derived from patterns of word co-occurrence in text corpora. These studies have helped map out semantic representation across a distributed brain network spanning temporal, parietal, and frontal cortex. However, it remains unclear whether activation patterns within regions reflect unified representations of sentence-level meaning, as opposed to superpositions of context-independent component words. This is because models have typically represented sentences as “bags-of-words” that neglect sentence-level structure. To address this issue, we interrogated fMRI activation elicited as 240 sentences were read by 14 participants (9 female, 5 male), using sentences encoded by a recurrent deep artificial neural-network trained on a sentence inference task (InferSent). Recurrent connections and nonlinear filters enable InferSent to transform sequences of word vectors into unified “propositional” sentence representations suitable for evaluating intersentence entailment relations. Using voxelwise encoding modeling, we demonstrate that InferSent predicts elements of fMRI activation that cannot be predicted by bag-of-words models and sentence models using grammatical rules to assemble word vectors. This effect occurs throughout a distributed network, which suggests that propositional sentence-level meaning is represented within and across multiple cortical regions rather than at any single site. In follow-up analyses, we place results in the context of other deep network approaches (ELMo and BERT) and estimate the degree of unpredicted neural signal using an “experiential” semantic model and cross-participant encoding.SIGNIFICANCE STATEMENT A modern-day scientific challenge is to understand how the human brain transforms word sequences into representations of sentence meaning. A recent approach, emerging from advances in functional neuroimaging, big data, and machine learning, is to computationally model meaning, and use models to predict brain activity. Such models have helped map a cortical semantic information-processing network. However, how unified sentence-level information, as opposed to word-level units, is represented throughout this network remains unclear. This is because models have typically represented sentences as unordered “bags-of-words.” Using a deep artificial neural network that recurrently and nonlinearly combines word representations into unified propositional sentence representations, we provide evidence that sentence-level information is encoded throughout a cortical network, rather than in a single region.     

Antibiotic resistance, population structure and spread of Staphylococcus aureus in nursing homes in the Euregion Meuse-Rhine

To determine the spread of Staphylococcus aureus within and between nursing home (NH) residents in the Euregion Meuse-Rhine, a cross-border region of the Netherlands and Germany, we investigated the prevalence of antibiotic resistance, genetic background and population structure of both methicillin-susceptible S. aureus (MSSA) and methicillin-resistant S. aureus (MRSA) isolates. A total of 245 S. aureus isolates were collected from NH residents. Susceptibility testing was performed with microbroth dilution. The genetic background was determined using spa typing, SCCmec typing, pulsed-field gel electrophoresis (PFGE) and multilocus sequence typing (MLST). Differences in the prevalence of resistance between the German and Dutch MSSA isolates were observed for the macrolides (15 % vs. 2 %, p?=?0.003), clindamycin (15 % vs. 0 %, p?=?0.003) and ciprofloxacin (34 % vs. 25 %). The macrolide and ciprofloxacin resistance varied between the NHs, while trimethoprim–sulfamethoxazole resistance was low in all residents. The MRSA prevalence was 3.5 % and <1 % among the German and Dutch NH residents, respectively (p?=?0.005). The German MRSAs, isolated in 7 out of 10 NHs, belonged to ST22-MRSA-IV or ST225-MRSA-II. spa clonal complexes (spa-CCs) 015 and 002 were prevalent among the German MSSA isolates and spa-CCs 024 and 1716 were prevalent among the Dutch MSSA isolates. The antibiotic resistance of MSSA and the MRSA prevalence were significantly higher among the German NH residents. The spread of two MRSA clones was observed within and between the German NHs, but not between the Dutch and German NHs. Differences in the prevalence of resistance and the prevalence of MRSA between NHs on both sides of the border warrant the continuation of surveillance at a local level.     

Elektrophysiologische Diagnostik der Critical Illness Polyneuropathie und Critical Illness Myopathie

Neurological complications of severe critical illness and sepsis are critical illness polyneuropathy (CIP) and critical illness myopathy (CIM). Both present with an ICU acquired muscular weakness and combinations of both are frequent. Weaning problems often may be the first symptoms at the intensive care unit. Electrophysiological assessment plays a major role in early diagnosis as the critically ill patients often cannot be adequately neurologically examined. In this review these disease entities and current diagnostic possibilities of electroneurography, classical electromyography, and direct muscle stimulation are discussed. Early diagnosis is most important as it improves the therapeutic regime in the ICU setting. Moreover, the differentiation between CIM and CIP has prognostic implications as the prognosis of CIP in its longtime course is more harmful.     

Influence of clodronate and compressive force on IL-1ß-stimulated human periodontal ligament fibroblasts

Objectives

The aim of this study was to investigate in vitro the effect of clodronate on interleukin-1ß (IL-1ß)–stimulated human periodontal ligament fibroblasts (HPdLFs) with the focus on inflammatory factors of orthodontic tooth movement with and without compressive force.

Materials and methods

HPdLFs were incubated with 5 μM clodronate and 10 ng/mL IL-1ß. After 48 h, cells were exposed to 3 h of compressive force using a centrifuge. The gene expression of cyclooxygenase-2 (COX-2), interleukin-6 (IL-6), matrix metalloproteinase 8 (MMP-8), and the tissue inhibitor of MMP (TIMP-1) was analyzed using RT-PCR. Prostaglandin E2 (PGE-2), IL-6, and TIMP-1 protein syntheses were quantified via ELISA.

Results

Compressive force and IL-1ß induced an overexpression of COX-2 gene expression (61.8-fold; p < 0.05 compared with control), diminished by clodronate (41.1-fold; p < 0.05 compared with control). Clodronate slowed down the compression and IL-1ß induced IL-6 gene expression (161-fold vs. 85.6-fold; p < 0.05 compared with control). TNF-α was only slightly affected without statistical significance. Clodronate reduced IL-1ß-stimulated MMP-8 expression with and without compressive force. TIMP-1 on gene and protein level was downregulated in all groups. Analyzing the MMP-8/TIMP-1 ratio, the highest ratio was detected in IL-1ß-stimulated HPdLFs with compressive force (21.2-fold; p < 0.05 compared with control). Clodronate diminished IL-1ß-induced upregulation of MMP-8/TIMP-1 ratio with (11.5-fold; p < 0.05 compared with control) and without (12.5-fold; p < 0.05 compared with control) compressive force.

Conclusion

Our study demonstrates a slightly anti-inflammatory effect by clodronate under compressive force in vitro. Additionally, the periodontal remodeling presented by the MMP-8/TIMP-1 ratio seems to be diminished by clodronate.

Clinical relevance

Reduction of pro-inflammatory factors and reduction of periodontal remodeling might explain reduced orthodontic tooth movement under clodronate intake.

     

Tumor-Stroma-Interaktionen im Harnblasenkarzinom

Die Urologie - Der histopathologische Aufbau maligner Tumoren schließt zwei Kompartimente ein – das Tumorparenchym mit den transformierten Zellen selbst und jene Strukturen, die deren...     

Left ventricular shape-luminal pressure relationship An open-chest study

Summary Left ventricular dimensions were measured in Cd²⁺ arrested (presumably diastolic), openchest rats. Aortic pressure was maintained at 137 cm H₂O (100 mm Hg) and left-ventricular (luminal) pressures were established and maintained at their chosen values, each by means of reservoir systems. The selected left-ventricular pressures were chosen to be within or to even broaden the range of conceivable diastolic pressures (–3 to 48 cm H₂O). After in situ fixation with 4 % formaldehyde and gelatin embedding, the hearts were serially sectioned in the apex base direction to obtain information at 11 levels (10, 20, ... 90, 100%). Tracings of selected sections were made along the edge of the left ventricular lumen and the pericardial surface. Volumes, surface areas, and mean external and internal radii of the left ventricle were derived. To quantify the circularity of sections a form factor (FF) was introduced (FF = 1 for a circular cross-section and less than one for other shapes). Ventricular lengths, radial dimensions, endocardial and epicardial surface areas, and total and luminal volumes increased with the increasing intraventricular pressures; as expected, the wall simultaneously thinned. Though its appearance was altered by the wall thinning, the curving muscle fascicular pattern was present over the entire pressure range examined. Endocardial surface areas increased more than did the epicardial surface areas. The endocardial FF value increased (more circular) at each section level as the pressure increased. The epicardial FF relationship was apparently constant (0.798 ± 0.014) for all section levels from 10 % through 90 %, regardless of luminal pressure. These results, when taken in conjunction with the results of our previous published studies, prompted the following speculation.The wall of the diastolic ventricle is a fluid-filled chamber with intramyocardial pressures that may be higher than ventricular pressures.