Data splitting for artificial neural networks using SOM-based stratified sampling.

Data splitting is an important consideration during artificial neural network (ANN) development where hold-out cross-validation is commonly employed to ensure generalization. Even for a moderate sample size, the sampling methodology used for data splitting can have a significant effect on the quality of the subsets used for training, testing and validating an ANN. Poor data splitting can result in inaccurate and highly variable model performance; however, the choice of sampling methodology is rarely given due consideration by ANN modellers. Increased confidence in the sampling is of paramount importance, since the hold-out sampling is generally performed only once during ANN development. This paper considers the variability in the quality of subsets that are obtained using different data splitting approaches. A novel approach to stratified sampling, based on Neyman sampling of the self-organizing map (SOM), is developed, with several guidelines identified for setting the SOM size and sample allocation in order to minimize the bias and variance in the datasets. Using an example ANN function approximation task, the SOM-based approach is evaluated in comparison to random sampling, DUPLEX, systematic stratified sampling, and trial-and-error sampling to minimize the statistical differences between data sets. Of these approaches, DUPLEX is found to provide benchmark performance with good model performance, with no variability. The results show that the SOM-based approach also reliably generates high-quality samples and can therefore be used with greater confidence than other approaches, especially in the case of non-uniform datasets, with the benefit of scalability to perform data splitting on large datasets.

A simple and inexpensive method for investigating microbiological, enzymatic, or inorganic catalysis using standard histology and microbiology laboratory equipment: assembly, mass transfer properties, hydrodynamic conditions and evaluation.

We introduce a generic, simple, and inexpensive method for performing microbiological, enzymatic, or inorganic catalysis with solids using standard histology and microbiology laboratory equipment. Histology cassettes were used to standardize hydrodynamic conditions and to protect the catalysts and their solid supports. Histology cassettes have the following advantages: they are readily available, inexpensive, solvent and acid resistant, automatable, and the slots in the cassette walls allow liquid to circulate freely. Standard Erlenmeyer flasks were used as reaction vessels. We developed a new camera to observe the movement and position of the histology cassettes as well as the liquid in the Erlenmeyer flasks. The camera produces a stable image of the rotating liquid in the Erlenmeyer flask. This visualization method revealed that in a 250 ml Erlenmeyer flask, stable operating conditions are achieved at a shaking frequency of 300 rpm and a fill volume of 30 ml. In vessels with vertical walls, such as beakers or laboratory bottles, the movement of the histology cassette is not reproducible. Mass transfer characterization using a biological model system and the chemical sulfite-oxidation method revealed that the histology cassette does not influence gas-liquid mass transfer.

[Osteosynthesis in facial bones: silicon nitride ceramic as material]. / Osteosynthese im Gesichtsschädelbereich: Siliziumnitritkeramik als Werkstoff.

BACKGROUND: The favorable properties of silicon nitride (Si3N4) ceramic, such as high stability and biocompatibility suggest its biomedical use as an implant material. The aim of this study was to test its suitability for osteosynthesis. MATERIALS AND METHODS: A Si3N4 prototype minifixation system was manufactured and implanted for osteosynthesis of artificial frontal bone defects in three minipigs. After 3 months, histological sections, CT and MRI scans were obtained. Finite element modeling (FEM) was used to simulate stresses and strains on Si3N4 miniplates and screws to calculate survival probabilities. RESULTS: Si3N4 miniplates and screws showed satisfactory intraoperative workability. There was no implant loss, displacement or fracture. Bone healing was complete in all animals and formation of new bone was observed in direct contact to the implants. CONCLUSIONS: Si3N4 ceramic showed a good biocompatibility outcome both in vitro and in vivo. This ceramic may serve as biomaterial for osteosynthesis, e.g. of the midface including reconstruction of the floor of the orbit and the skull base. Advantages compared to titanium are no risk of implantation to bone with mucosal attachment, no need for explantation, no interference with radiological imaging.

A distance-based uncertainty analysis approach to multi-criteria decision analysis for water resource decision making.

The choice among alternative water supply sources is generally based on the fundamental objective of maximising the ratio of benefits to costs. There is, however, a need to consider sustainability, the environment and social implications in regional water resources planning, in addition to economics. In order to achieve this, multi-criteria decision analysis (MCDA) techniques can be used. Various sources of uncertainty exist in the application of MCDA methods, including the selection of the MCDA method, elicitation of criteria weights and assignment of criteria performance values. The focus of this paper is on the uncertainty in the criteria weights. Sensitivity analysis can be used to analyse the effects of uncertainties associated with the criteria weights. Two existing sensitivity methods are described in this paper and a new distance-based approach is proposed which overcomes limitations of these methods. The benefits of the proposed approach are the concurrent alteration of the criteria weights, the applicability of the method to a range of MCDA techniques and the identification of the most critical criteria weights. The existing and proposed methods are applied to three case studies and the results indicate that simultaneous consideration of the uncertainty in the criteria weights should be an integral part of the decision making process.

[Biocompatibilty of silicon nitride ceramic in vitro. A comparative fluorescence-microscopic and scanning electron-microscopic study]. / Biokompatibilität von Siliziumnitrit-Keramik in der Zellkultur. Eine vergleichende fluoreszenzmikroskopische und rasterelektronenmikroskopische Untersuchung.

BACKGROUND: With regard to its favorable physical properties, silicon nitride ceramic is considered as biomaterial for human medical application. Minor controversy exists about the biocompatibility of the material. METHODS: Cytotoxicity testing, cell viability and morphology assessment was performed applying the L929-mice fibroblast cell culture model in a direct contact assay. Testing materials were silicon nitride ceramics of different surface properties and titanium alloy as a reference. Polyvinylchloride served as a negative control. The cells were stained with bisbenzimide and propidium iodine for double fluorochromasia viability testing, and evaluated by inversion-fluorescence microscopy. Scanning electron microscopy was applied to additionally investigate cell morphology. RESULTS: No cytotoxic effects were observed on the silicon nitride ceramic samples; moreover cell morphology remained the same as on titanium. Avital cells were present exclusively on PVC. The cell growth in the silicon nitride samples showed no significant differences compared to titanium. Cell counts on all polished surfaces showed significantly higher numbers. Scanning electron microscopy revealed typical fibroblast morphology with filiform extensions. CONCLUSION: The current results indicate a favorable biocompatibility of silicon nitride ceramic. Cell growth, viability and morphology are comparable to parameters of titanium. Polished surfaces appear to promote cell growth. Further in vivo studies are mandatory prior to human medical application. Owing to its favorable physiochemical properties, especially its superior resistance to mechanical stress, silicon nitride could serve as a biomaterial for osteosynthesis of bone with mucosal attachment.

Comparative investigation of the biocompatibility of various silicon nitride ceramic qualities in vitro.

There is a controversy about the biocompatibility of silicon nitride ceramics contained in the literature, which appears to be related to a factor of the individual chemical composition of different qualities of silicon nitride ceramics and of the different surface properties. This study attempts to investigate the cytotoxicity of different qualities of industrial silicon nitride ceramics applying an L929-cell culture model in a direct contact assay combined with a cell viability assessment. Five different qualities of industrial standard silicon nitride ceramics were chosen for in vitro testing. The chemical composition was determined by EDS analysis. Different biomedically approved aluminium oxide qualities, a titanium alloy, glass and polyvinylchloride (PVC) served as control materials. L929 mice fibroblasts were incubated directly on the materials for 24 h, stained with bisbenzimide and propidium iodine for double fluorochromasia viability testing, and evaluated by inversion-fluorescence microscopy to control cell morphology, viability and cell counts compared to empty well values. Scanning electron microscopy was applied to additionally investigate cell morphology. There was no observation of cytotoxic effects on the silicon nitride ceramic samples; moreover cell morphology remained the same as on aluminium oxide and titanium. Viability testing revealed the presence of avital cells exclusively on PVC, which served as a negative control. Cell counts on all polished surfaces showed significantly higher numbers, whereas some rough surface samples showed significantly lower numbers. We conclude that silicon nitride ceramics show no cytotoxic effects and should be considered for biomedical application owing to its favourable physiochemical properties, especially its superior resistance to mechanical stress, which would be useful for compression loaded conditions. Polished surfaces would appear to promote advanced biocompatibility.

Improved reliability of leucite reinforced glass by ion exchange.

OBJECTIVES: The aim of this study was to develop an adequate process-, proof- and calculation technique for the chemical strengthening of dental ceramic materials in order to evaluate the efficiency of this method from the engineering point of view. METHODS: Ion exchange examinations below glass temperature have been done for the dental all ceramic system IPS-Empress (Ivoclar). In addition experiments on specimens were undertaken, which were stored in artificial saliva for the period of one month after the ion exchange treatment. The efficiency of the ion-exchange treatments has been evaluated based on the chemical composition of the surface layer, the residual stress distribution, the flexural- and the burst-strength and surface hardness. In addition, Weibull statistical analysis of the four-point bending and burst test data provides new information about the failure mechanisms of chemically strengthened dental ceramics. RESULTS: The ion-exchange process with a potassium salt increases the reliability of the dental ceramic material "IPS-Empress". By the KNO3 or Tuf-Coat ion exchange treatment potassium ions will be implanted as clusters directly into the crystal lattice which results in an increase of the binding energy and hence, in a pronounced increase in the surface hardness. In particular the stress bearing capabilities of dental all ceramic restorations, which are subject to high bending stresses, i.e. dental crowns and dental bridges, are largely improved by the induced residual stresses. This increase of residual stresses did not relax significantly after a one month storage in artificial saliva. SIGNIFICANCE: Enhanced load limits and a more narrow strength distribution spectrum is considered desirable for increasing the clinical longevity of all ceramics.