1,721,411 research outputs found
Potentiometrisch en calorimetrisch onderzoek van het zuur-base gedrag en van de complexvorming met Ag(1) van een aantal N-methyl-gesubstitueerde-dieethyleentriaminen
No full textNot availabl
Potentiometrisch en calorimetrisch onderzoek van het zuur-base gedrag en van de complexvorming met Ag(1) van een aantal N-methyl-gesubstitueerde-dieethyleentriaminen
No full textNot availabl
Bridging the gap between machine learning and evoked potentials in MS
No full textThe field of machine learning has seen a major resurgence in recent years, fueled
by the ever-increasing availability of both data and computing power. Healthcare in
particular stands to gain much from the adoption of such techniques. In this thesis,
we apply machine learning in a range of domains, focusing mainly on the field of
multiple sclerosis (MS).
MS is an incurable, chronic disease which affects the conduction of neurons in the
central nervous system. Evoked potentials (EP) are measurements that give insight
into the damage caused to these neural pathways. Motor evoked potentials (MEP)
specifically measure the level of degradation of signals transmitted from the brain to
muscles in the hands or the feet. These types of measurements have been suggested
to have prognostic value for the MS disease course.
In this work, we seek to further substantiate this claim by gathering and curating
a large, real-world dataset of MEPs. Through the application of various machine
learning techniques we confirm that MEPs have prognostic value. We also find that
there is potentially more information contained in the raw data which is not captured
by commonly used summarizing measures.
To facilitate further research on MEP measurements, we automate part of their
analysis which so far was done manually by experts. By eliminating the need for
expert knowledge in the interpretation of the measurements, the insights provided by
MEP measurements may benefit more patients. Furthermore, we opted to make the
MEP dataset publicly available so as to promote further research on these types of
measurements.
Besides the applications in MS we also present some of our work in other domains.
This includes our early work on the optimization of various reservoir computing algorithms using Bayesian methods, where we show significant improvements on several
benchmarks. Furthermore, we describe our research performed in collaboration with
the late Christian Van den Broeck who started the research line on applied machine
learning in the physics group at Hasselt University. In that work we use a deep learning approach to both identify bird species based on recordings of their songs with
remarkable accuracy, and to detect apples in images of orchards. For both applications we find that these techniques readily outperform classical techniques, using
off-the-shelf frameworks.Het onderzoeksveld van machinaal leren (machine learning) heeft een grote wederopleving gekend in de afgelopen jaren, gedreven door de steeds toenemende beschikbaarheid van zowel data als rekenkracht. Gezondheidszorg in het bijzonder heeft veel te
winnen bij het toepassen van dergelijke technieken. In deze thesis passen we machine
learning toe in verschillende domeinen, waarbij we ons voornamelijk richten op het
onderzoeksveld van multiple sclerose (MS).
MS is een ongeneeslijke, chronische ziekte die de geleiding van neuronen in het
centrale zenuwstelsel aantast. Ge¨evoceerde potentialen (EP) zijn metingen die inzicht geven over de opgelopen schade aan deze zenuwbanen. Motorisch ge¨evoceerde
potentialen (MEP) meten specifiek de mate van degradatie van signalen die van de
hersenen naar spieren in de handen of voeten worden gestuurd. Er is gesuggereerd
dat dit soort metingen prognostische waarde hebben voor het ziekteverloop van MS.
In dit werk trachten we deze bewering verder te onderbouwen door een grote,
realistische dataset van MEPs te verzamelen en te beheren. Door de toepassing
van verschillende machine learning-technieken bevestigen we dat MEPs prognostische
waarde hebben. We stellen ook vast dat er mogelijk meer informatie in de ruwe data
van de metingen zit die niet wordt opgepikt door typisch gebruikte samenvattende
maten.
Om verder onderzoek naar MEP-metingen te faciliteren, automatiseren we een
deel van hun analyse, hetgeen tot nu toe handmatig werd gedaan door experts. Door
de nood aan gespecialiseerde kennis bij de interpretatie van de metingen weg te nemen, kunnen de inzichten die door MEP-metingen worden geboden meer pati¨enten
ten goede komen. Verder hebben we ervoor gekozen om de MEP-dataset publiek
beschikbaar te stellen om verder onderzoek naar dit soort metingen te stimuleren.
Naast de toepassingen binnen MS presenteren we ook een deel van ons werk in
andere domeinen. Dit omvat ons vroege werk rond de optimalisatie van verschillende
algoritmes binnen het veld van reservoir berekening (reservoir computing) met behulp van Bayesiaanse methodes, waarbij we aanzienlijke verbeteringen demonstreren
op verschillende benchmarks. Verder beschrijven we ons onderzoek dat werd uitgevoerd in samenwerking met wijlen Christian Van den Broeck die de onderzoekslijn
over toegepaste machine learning startte in de fysica-groep van Universiteit Hasselt.
In dat werk gebruiken we een diep leren (deep learning) aanpak om zowel vogelsoorten met opmerkelijke nauwkeurigheid te identificeren op basis van opnames van hun
zang, alsook om appels te detecteren in afbeeldingen van boomgaarden. Voor beide
toepassingen komen we tot de conclusie dat deze technieken, gebruik makende van
vrij beschikbare software, beter presteren dan klassieke technieken
Bridging the gap between machine learning and evoked potentials in MS
No full textThe field of machine learning has seen a major resurgence in recent years, fueled
by the ever-increasing availability of both data and computing power. Healthcare in
particular stands to gain much from the adoption of such techniques. In this thesis,
we apply machine learning in a range of domains, focusing mainly on the field of
multiple sclerosis (MS).
MS is an incurable, chronic disease which affects the conduction of neurons in the
central nervous system. Evoked potentials (EP) are measurements that give insight
into the damage caused to these neural pathways. Motor evoked potentials (MEP)
specifically measure the level of degradation of signals transmitted from the brain to
muscles in the hands or the feet. These types of measurements have been suggested
to have prognostic value for the MS disease course.
In this work, we seek to further substantiate this claim by gathering and curating
a large, real-world dataset of MEPs. Through the application of various machine
learning techniques we confirm that MEPs have prognostic value. We also find that
there is potentially more information contained in the raw data which is not captured
by commonly used summarizing measures.
To facilitate further research on MEP measurements, we automate part of their
analysis which so far was done manually by experts. By eliminating the need for
expert knowledge in the interpretation of the measurements, the insights provided by
MEP measurements may benefit more patients. Furthermore, we opted to make the
MEP dataset publicly available so as to promote further research on these types of
measurements.
Besides the applications in MS we also present some of our work in other domains.
This includes our early work on the optimization of various reservoir computing algorithms using Bayesian methods, where we show significant improvements on several
benchmarks. Furthermore, we describe our research performed in collaboration with
the late Christian Van den Broeck who started the research line on applied machine
learning in the physics group at Hasselt University. In that work we use a deep learning approach to both identify bird species based on recordings of their songs with
remarkable accuracy, and to detect apples in images of orchards. For both applications we find that these techniques readily outperform classical techniques, using
off-the-shelf frameworks.Het onderzoeksveld van machinaal leren (machine learning) heeft een grote wederopleving gekend in de afgelopen jaren, gedreven door de steeds toenemende beschikbaarheid van zowel data als rekenkracht. Gezondheidszorg in het bijzonder heeft veel te
winnen bij het toepassen van dergelijke technieken. In deze thesis passen we machine
learning toe in verschillende domeinen, waarbij we ons voornamelijk richten op het
onderzoeksveld van multiple sclerose (MS).
MS is een ongeneeslijke, chronische ziekte die de geleiding van neuronen in het
centrale zenuwstelsel aantast. Ge¨evoceerde potentialen (EP) zijn metingen die inzicht geven over de opgelopen schade aan deze zenuwbanen. Motorisch ge¨evoceerde
potentialen (MEP) meten specifiek de mate van degradatie van signalen die van de
hersenen naar spieren in de handen of voeten worden gestuurd. Er is gesuggereerd
dat dit soort metingen prognostische waarde hebben voor het ziekteverloop van MS.
In dit werk trachten we deze bewering verder te onderbouwen door een grote,
realistische dataset van MEPs te verzamelen en te beheren. Door de toepassing
van verschillende machine learning-technieken bevestigen we dat MEPs prognostische
waarde hebben. We stellen ook vast dat er mogelijk meer informatie in de ruwe data
van de metingen zit die niet wordt opgepikt door typisch gebruikte samenvattende
maten.
Om verder onderzoek naar MEP-metingen te faciliteren, automatiseren we een
deel van hun analyse, hetgeen tot nu toe handmatig werd gedaan door experts. Door
de nood aan gespecialiseerde kennis bij de interpretatie van de metingen weg te nemen, kunnen de inzichten die door MEP-metingen worden geboden meer pati¨enten
ten goede komen. Verder hebben we ervoor gekozen om de MEP-dataset publiek
beschikbaar te stellen om verder onderzoek naar dit soort metingen te stimuleren.
Naast de toepassingen binnen MS presenteren we ook een deel van ons werk in
andere domeinen. Dit omvat ons vroege werk rond de optimalisatie van verschillende
algoritmes binnen het veld van reservoir berekening (reservoir computing) met behulp van Bayesiaanse methodes, waarbij we aanzienlijke verbeteringen demonstreren
op verschillende benchmarks. Verder beschrijven we ons onderzoek dat werd uitgevoerd in samenwerking met wijlen Christian Van den Broeck die de onderzoekslijn
over toegepaste machine learning startte in de fysica-groep van Universiteit Hasselt.
In dat werk gebruiken we een diep leren (deep learning) aanpak om zowel vogelsoorten met opmerkelijke nauwkeurigheid te identificeren op basis van opnames van hun
zang, alsook om appels te detecteren in afbeeldingen van boomgaarden. Voor beide
toepassingen komen we tot de conclusie dat deze technieken, gebruik makende van
vrij beschikbare software, beter presteren dan klassieke technieken
Characterization of sulfur compounds in supercritical coal extracts by gas chromatography-mass spectrometry
No full textThe organosulfur compounds (OSC) in the supercritical extracts obtained from flame coal (subA) and orthocoking coal (mvb) were identified by gas chromatography-mass spectrometry. Supercritical fluid extraction was carried out with three different solvents, i.e., toluene, toluene/2-propanol and toluene/THF mixtures. at 360 degreesC and 10 MPa in an apparatus with continuous flow of solvent. The extraction yield was in the range of 11.4-39.9 wt.% depending on the type of solvent and coal. For flame coal, diphenyl sulfide and disulfide, thiophene, benzothiophene, dibenzothiophene and benzonaphtothiophene and their C-1-C-4 alkyl derivatives were detected, whereas for orthocoking coal only polycyclic aromatic sulfur heterocycles (PASH) containing two to five rings and their alkyl derivatives were found. Ligand exchange chromatography was applied to separate the PASH fraction. (C) 2002 Elsevier Science B.V. All rights reserved
Impacts of sonication and post-desulfurization on organic sulfur species by reductive pyrolysis
No full textA comparative research on the effect of the ultrasound irradiation and peroxyacetic acid (PAA) desulfurization on sulfur forms in high sulfur coal samples (gathered from Tabas mine in Iran) was studied by reductive pyrolysis method. The total sulfur reduction after chemical desulfurization by PAA for sonicated samples is achieved in a range of 49-58%. Moreover, the studied sonicated-desulfurized samples showed that the pyritic and sulfate sulfur were mainly attacked by PAA. The maximum organic sulfur reduction was obtained for longer sonication treatment times (15 and 20 min) being around 42-44%. For the first time the role of advanced oxidation process (AOP) in the sonication treatment of coal samples has been considered. The profiles of m/z 48 (SO+) and 64 (SO2+) obtained by Atmospheric Pressure-Temperature Program Reduction on-line coupled with MS (AP-TPR/MS) experiments for sonicated and sonicated-desulfurized samples exhibited identical trends over the whole temperature range. This achievement demonstrated the presence of high amounts of oxidized sulfur functionalities as a function of sonication treatment and post-chemical PAA desulfurization. Furthermore, AP-TPR "off-line" coupled with TD-GC/MS showed quantitative changes in the refractory sulfur forms as a result of sonication settings before and after desulfurization.The Iranian authors like to thank the Tabas coal mine in helping the sampling and also the science and research branch of Islamic Azad University and Hasselt University for financial support of this research project
Low rank coals sulphur functionality study by AP-TPR/TPO coupled with MS and potentiometric detection and by XPS
No full textAtmospheric pressure-temperature programmed reduction (AP-TPR) and X-ray photoelectron spectroscopy (XPS) techniques were applied to low rank coals sulphur study. Coal samples were pyrolysed in a flow of water vapor (WV). It was demonstrated that this treatment influenced mainly aliphatic sulphur. Samples were characterised by two methods and data were interpreted within the limits of the techniques. XPS measurements registered sulphur 2p spectra with two main signals for organic and inorganic sulphur compounds. The AP-TPR set-up, with potentiometric detection of the formed H2S as S2- using an ion selective Ag2S-electrode, gives quantitative data about the presence of different sulphur species. The AP-TPR equipment on-line coupled with a mass spectrometer (MS) gives extra qualitative information about different reductive and oxidative organic sulphur forms. Using MS not only H2S but also SO2, COS, CS2, and all other volatile sulphur and organic compounds can be monitored, giving more information for the initial presence of the different sulphur forms and to the mechanisms involved in the pyrolytic process. This AP-TPR-MS experiment is subsequently followed by AP-TPO-MS measurement (in an oxidated atmosphere) to study sulphur presence in the residue (tar and char) in the reactor. Comparing all these AP-TPR profiles results in a better assignment of the different signals to specific sulphur functionalities. (C) 2003 Elsevier B.V. All rights reserved
Adsorption of Cibacron Yellow F-4G dye onto activated carbons obtained from peanut hull and rice husk: kinetics and equilibrium studies
No full textThe use of biomass residues from agricultural processes for the production of cheap and competitive activated carbon (AC) is an excellent option to minimize the costs of existing procedures to remove dyes from wastewater. In this study, the potential use of AC obtained from peanut hulls and rice husks for adsorption of Cibacron Yellow F-4G (CYF-4G) is examined. The activated peanut hull (PHAC) and rice husk (RHAC) were characterized by TGA, FTIR, BET and elemental analysis. The effects of different process variables as well as the dose of adsorbent, dye concentration and pH were evaluated. A decrease in amount of dye adsorbed per unit adsorbent mass was observed when increasing CYF-4G concentration. The results showed an optimal dye adsorption value at a pH of 2.0. The adsorption kinetics of CYF-4G are governed by the pseudo-second-order model. In addition, adsorption fits the Langmuir isotherm better than Freundlich's. Adsorption capacities of AC prepared from agricultural waste show that PHAC performs better than RHAC to remove CYF-4G.Yperman, J (corresponding author), Univ Hasselt, Res Grp Appl & Analyt Chem, Agoralaan Gebouw D, BE-3590 Diepenbeek, Belgium.
[email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]
The effect of demineralisation on characteristics and adsorption behaviour of activated carbons prepared from swine manure
No full textActivated carbons (ACs) from swine manure (SM), i.e. preliminary de-watered SM cake (Mc) and solid digestate (Md) were prepared by pyrolysis and water vapour activation. Demineralisation on obtained chars after carbonization was applied as well. Additionally, the adsorption capacity of demineralised ACs towards chromium in aqueous solutions was investigated and compared to non-demineralised ACs. Demineralisation caused a decrease in mineral matter content with more than 50% and affected characteristics and adsorption behaviour of these ACs. As a result, ACs with better developed porous texture were produced. Prepared ACs with reduced ash content demonstrated higher adsorption capacity toward Cr compared to non-demineralised ACs. The adsorption kinetics were investigated by applying two kinetic models, i.e. Lagergren pseudo-first order and pseudo-second order. The pseudo-second order kinetic model provided a better fit. Better fits are obtained for the Langmuir isotherm model and therefore a monolayer coverage chemisorption of the Cr(VI) on the ACs surface is suggested.FWO; BAS-IOCh bilateral research projec
Er zit meer in mest dan je denkt
No full textBiosorb wordt uitgevoerd met financiële steun van Vlaams Agentschap voor Innoveren en Ondernemen (VLaio
- …
