Unsupervised feature selection using clustering ensembles and population based incremental learning algorithm

This paper describes a novel feature selection algorithm for unsupervised clustering, that combines the clustering ensembles method and the population based incremental learning algorithm. The main idea of the proposed unsupervised feature selection algorithm is to search for a subset of all features such that the clustering algorithm trained on this feature subset can achieve the most similar clustering solution to the one obtained by an ensemble learning algorithm. In particular, a clustering solution is firstly achieved by a clustering ensembles method, then the population based incremental learning algorithm is adopted to find the feature subset that best fits the obtained clustering solution. One advantage of the proposed unsupervised feature selection algorithm is that it is dimensionality-unbiased. In addition, the proposed unsupervised feature selection algorithm leverages the consensus across multiple clustering solutions. Experimental results on several real data sets demonstrate that the proposed unsupervised feature selection algorithm is often able to obtain a better feature subset when compared with other existing unsupervised feature selection algorithms.     

Determination of N-nitrosodimethylamine at part-per-trillion levels in drinking waters and contaminated groundwaters

The carcinogen N-nitrosodimethylamine (NDMA) may be quantitated routinely at ultratrace (ng/L) levels in drinking water or contaminated groundwater. The aqueous sample is passed through a preconditioned Empore C18 filter disk to remove neutral nonpolar species and then extracted continuously overnight with highest purity dichloromethane. The latter is then concentrated to 1 mL, and a large aliquot (up to 200 microL) is loaded onto a dual-stage carbon sorbent trap, after which the solvent is removed with ultrapure helium. The concentrated residues are then injected onto a gas chromatographic column using a short-path thermal desorber. NDMA is selectively detected using a chemiluminescent nitrogen detector (CLND) operated in its nitrosamine-selective mode. The reporting limit for this procedure, evaluated using two independent statistically unbiased protocols, is 2 ng of NDMA/L. A related procedure, employing an automatic sampler instead of the short-path thermal desorber, provides convenient analysis of heavily contaminated samples and exhibits a reporting limit (same protocols cited previously) of 110 ng of NDMA/L. When the two methods are used together in a "two-tiered" protocol, NDMA concentrations spanning 4 orders of magnitude (ng/L to microgram/L levels) may be measured routinely. The low-level procedure employing only the short-path thermal desorber was applied successfully to three sources of drinking water, where NDMA concentrations ranged between 2 and 10 ng of NDMA/L. The two-tiered protocol was applied to a series of contaminated groundwaters whose NMDA concentrations ranged between approximately 10-7000 ng of NDMA/L. The results agreed with those obtained from an independent collaborating laboratory, which used a different analytical procedure.     

A competitive associative memory model and its dynamics

Conventional associative memory networks perform "noncompetitive recognition" or "competitive recognition in distance". In this paper a "competitive recognition" associative memory model is introduced which simulates the competitive persistence of biological species. Unlike most of the conventional networks, the proposed model takes only the prototype patterns as its equilibrium points, so that the spurious points are effectively excluded. Furthermore, it is shown that, as the competitive parameters vary, the network has a unique stable equilibrium point corresponding to the winner competitive parameter and, in this case, the unique stable equilibrium state can be recalled from any initial key.     

Effectiveness of slow release L-DOPA/benserazide in treatment of end-of-dose akinesia in Parkinson disease

In an open label study 63 patients with idiopathic Parkinson's disease suffering from end-of-dose akinesia were switched from a treatment with a L-DOPA standard formulation to a combined therapy of L-DOPA standard in the morning and L-DOPA slow release (levodopa, benserazide, Madopar Depot) at the remaining single doses. Substitution of L-DOPA standard by L-DOPA slow release took on average 2-4 weeks. Patients were subsequently treated for 6 months. Due to a lower bioavailability of the slow release formulation--the latter is based on the "hydrodynamically balanced system" (HBS)--, the patients remained initially on their time schedule of drug intake but received a higher dose of L-DOPA slow release compared to the preceding L-DOPA standard therapy. In 20 centers 37 men and 26 women were included into the study. 27 males and 20 females completed the 6 month treatment period. Before switching, the patients received 438 +/- 213 mg a day L-DOPA standard, after conversion, the average dose was 617 +/- 323 mg L-DOPA slow release and 107 +/- 95 mg L-DOPA standard a day. Fluctuations during the day and at night which were rated according to a newly developed clinical 5-point rating scale were significantly improved by the treatment regimen from 2.8 +/- 0.9 to 1.4 +/- 1.2. Additionally, parkinsonian symptoms were significantly reduced during the ON-phase as there was a significant decrease of the Webster rating score from 12.0 +/- 4.6 to 7.1 +/- 4.0. Quality of life as measured by subjective ratings of the patients improved. The tolerability of the new formulation of L-DOPA was rated to be good in 51.1% and very good in 48.9%. The results of this open label study suggest that the combination of L-DOPA standard in the morning and L-DOPA slow release formulation at the following time points can be an efficient therapy in parkinsonian patients who suffer form L-DOPA related end-of-dose motor akinesia.     

Stability and distribution of orally administered epidermal growth factor in neonatal pigs

Stability and distribution of orally administered epidermal growth factor (EGF) were examined in newborn and 5-day-old pigs. Forty-five minutes after oral administration of iodine-125 labeled EGF, 60 and 50% of the radioactivity administered were recovered from the internal organs in newborn and 5-day-old pigs, respectively. In both age groups, over 95% of the recovered radioactivity was found in the gastrointestinal tract, of which 78-86% was found in the luminal contents with the remaining found in the gastrointestinal wall. Within the gastrointestinal tract, 65-71% of radioactivity was found in the stomach, 27-30% in the proximal and mid small intestine and 3-4% was found in the distal part of the small intestine. There were no significant differences in the overall distribution of orally administered radioactivity between two age groups. Based on liquid chromatography and trichloroacetic acid precipitation, a substantial amount of EGF recovered from the luminal contents (63-86%) and the gastrointestinal wall (42-81%) remained "intact". The receptor binding ability of the EGF recovered from the gastric contents was 96-102% comparable to the native EGF tracer. The receptor binding ability remained high (40-58%) in the proximal small intestinal lumen and it decreased to 15% in the distal small intestinal lumen in newborn pigs. In 5-day-old pigs, EGF recovered from the small intestinal contents had 5 to 24% receptor binding ability when compared with native EGF tracer. The receptor binding ability of the EGF recovered from all other organs was below 5% with an exception of the gastric wall, from which recovered EGF retained 9 to 26% receptor binding ability. These results indicate that most of orally ingested EGF remained in the gastrointestinal tract in neonatal pigs 45 min after oral ingestion, and significant amount of the ingested EGF remained biologically active. It suggests that milk-borne EGF can survive in the gastrointestinal tract and may play a role in regulating gut development in neonatal animals.     

Identification of volatile organic compounds emitted in the field by oilseed rape (Brassica napus ssp. oleifera) over the growing season

There is currently a trend gaining acceptance that explicitly recognizes the need to set priorities for making rational decisions in health policy, respecting at the same time the underlying purpose of health care - to improve people's health. This trend in health policy is referred to as "health targets" although definitions vary considerably. Having been initiated by the World Health Organization in the late 1970s, the international policy on health targets is in the process of renewal to become Health for All in the Twenty-First Century. The new program highlights 10 global targets, from the reduction of worldwide burden of diseases to improvement of access to comprehensive, essential quality care. Countries such as the United Kingdom and Australia have adopted and implemented such programs which basically include cardiovascular, cancer, accidents, and mental health targets. For many countries, however, there are several difficulties in establishing similar programs. One of them is the unavailability of reliable data, although political factors and structure of health systems also play a key role in the creation of health targets. Overall, health targets appear to be a key element in building a strong public health policy, taking into consideration not only the cost of healthcare, but also the outcomes in improving health which is the ultimate goal of health care systems.     

Acute vasoconstriction after subarachnoid hemorrhage

OBJECTIVE: Decreased cerebral blood flow (CBF) and cerebral ischemia occurring immediately after subarachnoid hemorrhage (SAH) may be caused by acute microvascular constriction. However, CBF can also be influenced by changes in intracranial pressure (ICP) and cerebral perfusion pressure (CPP). The goal of these experiments was to assess the significance of acute vasoconstriction after SAH and its relationship to changes in CBF, ICP, CPP, and extracellular glutamate concentrations. METHODS: Three experiments were performed using the endovascular filament technique to produce SAH. In the first experiment, CBF, ICP, and CPP were measured for 60 minutes after SAH (n = 21) and were correlated with the 24-hour mortality rate. In the second experiment, rats undergoing SAH (n = 23) or a sham procedure (n = 7) were perfused 60 minutes after SAH for measurement of the circumference and wall thickness of the internal carotid and anterior cerebral arteries and correlation with CBF, ICP, and CPP. In the third experiment (n = 11), extracellular glutamate concentrations determined by hippocampal and cortical microdialysis and high performance liquid chromatography were correlated with physiological changes. RESULTS: CBF reductions to less than 40% of baseline for 60 minutes after SAH predicted 24-hour mortality with 100% accuracy and were used to define "lethal" SAH. In contrast, ICP and CPP 60 minutes after SAH were not correlated with the mortality rate. The vascular circumference was significantly smaller in lethal than in sublethal SAH or sham-operated rats (P < 0.001). Vessel measurements were correlated with both CBF and hemorrhage size (P < 0.01). Extracellular glutamate concentration increased to 600% of baseline after lethal SAH in both hippocampus and cortex and was inversely correlated with CBF (r = 0.9, P < 0.001) but did not increase after sublethal SAH. CONCLUSION: Acute vasoconstriction after SAH occurs independently of changes in ICP and CPP and is associated with decreased CBF, larger hemorrhage size, persistent elevations of extracellular glutamate, and poor outcome. Acute vasoconstriction seems to contribute directly to ischemic brain injury after SAH. Further evaluations of pharmacological agents with the potential to reverse acute vasoconstriction may increase CBF and improve outcome.     

Auditory system plasticity in children after long periods of complete deafness

Separation and identification of proteins by two-dimensional (2-D) electrophoresis can be used for protein-based gene expression analysis. In this report single protein spots, from polyvinylidene difluoride blots of micropreparative E. coli 2-D gels, were rapidly and economically identified by matching their amino acid composition, estimated pI and molecular weight against all E. coli entries in the SWISS-PROT database. Thirty proteins from an E. coli 2-D map were analyzed and identities assigned. Three of the proteins were unknown. By protein sequencing analysis, 20 of the 27 proteins were correctly identified. Importantly, correct identifications showed unambiguous "correct" score patterns. While incorrect protein identifications also showed distinctive score patterns, indicating that protein must be identified by other means. These techniques allow large-scale screening of the protein complement of simple organisms, or tissues in normal and disease states. The computer program described here is accessible via the World Wide Web at URL address (http:@expasy.hcuge.ch/).