Fatigue compensation during FES using surface EMG

Muscle fatigue limits the effectiveness of FES when applied to regain functional movements in spinal cord injured (SCI) individuals. The stimulation intensity must be manually increased to provide more force output to compensate for the decreasing muscle force due to fatigue. An artificial neural network (ANN) system was designed to compensate for muscle fatigue during functional electrical stimulation (FES) by maintaining a constant joint angle. Surface electromyography signals (EMG) from electrically stimulated muscles were used to determine when to increase the stimulation intensity when the muscle’s output started to drop.

In two separate experiments on able-bodied subjects seated in hard back chairs, electrical stimulation was continuously applied to fatigue either the biceps (during elbow flexion) or the quadriceps muscle (during leg extension) while recording the surface EMG. An ANN system was created using processed surface EMG as the input, and a discrete fatigue compensation control signal, indicating when to increase the stimulation current, as the output. In order to provide training examples and test the systems’ performance, the stimulation current amplitude was manually increased to maintain constant joint angles. Manual stimulation amplitude increases were required upon observing a significant decrease in the joint angle. The goal of the ANN system was to generate fatigue compensation control signals in an attempt to maintain a constant joint angle.

On average, the systems could correctly predict 78.5% of the instances at which a stimulation increase was required to maintain the joint angle. The performance of these ANN systems demonstrates the feasibility of using surface EMG feedback in an FES control system.     

人工神经网络在蝙蝠回声定位叫声识别方面的应用

近年来,人工神经网络被不断应用于野生动物的声学研究中,本文概括地介绍了人工神经网络的概念以及这项新技术的研究方法,并且重点介绍了它在蝙蝠回声定位叫声识别方面的应用。     

Estimation of biomass concentrations in fermentation processes for recombinant protein production

Online biomass estimation for bioprocess supervision and control purposes is addressed. As the biomass concentration cannot be measured online during the production to sufficient accuracy, indirect measurement techniques are required. Here we compare several possibilities for the concrete case of recombinant protein production with genetically modified Escherichia coli bacteria and perform a ranking. At normal process operation, the best estimates can be obtained with artificial neural networks (ANNs). When they cannot be employed, statistical correlation techniques can be used such as multivariate regression techniques. Simple model-based techniques, e.g., those based on the Luedeking/Piret-type are not as accurate as the ANN approach; however, they are very robust. Techniques based on principal component analysis can be used to recognize abnormal cultivation behavior. For the cases investigated, a complete ranking list of the methods is given in terms of the root-mean-square error of the estimates. All techniques examined are in line with the recommendations expressed in the process analytical technology (PAT)-initiative of the FDA.     

Focus

A solid background in population ecology is valuable for anyone concerned with natural resources management. However, without the opportunity of field experience, it may be difficult for students to become familiar with some of the principles of population biology. Microcomputers and programmable pocket calculators allow the construction of simple simulation models, making it possible to carry out some ‘numerical experiments’ in the classroom which may play the same role in the learning process as in vivo experimentation

This work introduces a mathematical model based on the Lotka—Volterra equations and a computer program developed for a programmable calculator. They can be used to assess the influence of environmental heterogeneity and disturbances on the results of competitive relationships between two plant populations. As an example, the interaction between two tussock grass species of the Patagonian steppe, Stipa speciosa Trin. et Rupr. and Festuca pallescens (St Yves) Parodi (the second being less xerophytic than the first one), is analysed.     

An ANN model for treatment prediction in HBV patients

Two types of antiviral treatments, namely, interferon and nucleoside/nucleotide analogues are available for hepatitis infections. The selection of drug and dose determined using known pharmacokinetics and pharmacodynamics data is important. The lack of sufficient information for pharmacokinetics of a drug may not produce the desired results. Artificial neural network (ANN) provides a novel model-independent approach to pharmacokinetics and pharmacodynamics data. ANN model is created by supervised learning of 90 patients sample to predict the treatment strategy (lamivudine only and Lamivudine + Interferon) on the basis of viral load, liver function test, visit number, treatment duration, ethnic area, sex, and age. The model was trained with 68 (77.3%) samples and tested with 20 (22.7%) samples. The model produced 92% accuracy with 92.8% sensitivity and 83.3% specificity.     

Automata networks as preprocessing technique of artificial neural network in estimating primary production and dominating phytoplankton levels in a reservoir: An experimental work

Artificial Neural Networks (ANN) is computational architectures that can be used for estimating primary production levels and dominating phytoplankton species in reservoirs. Automata Networks (AN) were applied as a pre-processing method with subsequent ANN model development for Demirdöven Dam Reservoir. The primary purpose of using pre-processing technique was to distinguish the suitable and appropriate constituents of the input parameters' matrix, to eliminate redundancy, to enhance prediction power and calculation efficiency. The data were collected monthly over two years. The applications have yielded following results: The correlation coefficients (r values) between predicted and observed counts were as high as 0.83, 0.87, 0.83 and 0.88 for Cyclotella ocellata, Sphaerocystis schroeteri, Staurastrum longiradiatum counts, and Chlorophyll-a (Chl-a) concentrations respectively with AN. The performance of AN based pre-processing technique was compared with the performance of a well-known pre-processing technique, namely Principle Component Analysis(PCA), experimentally. r values between the predicted and observed C. ocellata, S. schroeteri and S. longiradiatum counts, and (Chl-a) were as high as 0.80, 0.86, 0.81 and 0.86 respectively with PCA.     

Mapping tropical forest fractional cover from coarse spatial resolution remote sensing imagery

At regional to global scales the only feasible approach to mapping and monitoring forests is through the use of coarse spatial resolution remotely sensed imagery. Significant errors in mapping may arise as such imagery may be dominated by pixels of mixed land cover composition which cannot be accommodated by conventional mapping approaches. This may lead to incorrect assessments of forest extent and thereby processes such as deforestation which may propagate into studies of environmental change. A method to unmix the class composition of image pixels is presented and used to map tropical forest cover in part of the Mato Grosso, Brazil. This method is based on an artificial neural network and has advantages over other techniques used in remote sensing. Fraction images depicting the proportional class coverage in each pixel were produced and shown to correspond closely to the actual land cover. The predicted and actual forest cover were, for instance, strongly correlated (up to r = 0.85, significant at the 99% level of confidence) and the predicted extent of forest over the test site much closer to the actual extent than that derived from a conventional approach to mapping from remotely sensed imagery.     

Comparative modeling of the conformational stability of chymotrypsin inhibitor 2 protein mutants using amino acid sequence autocorrelation (AASA) and amino acid 3D autocorrelation (AA3DA) vectors and ensembles of Bayesian-regularized genetic neural networks

Predicting protein stability changes upon point mutation is important for understanding protein structure and designing new proteins. Autocorrelation vector formalism was extended to amino acid sequences and 3D conformations for encoding protein structural information with modeling purpose. Protein autocorrelation vectors were weighted by 48 amino acid/residue properties selected from the AAindex database. Ensembles of Bayesian-regularized genetic neural networks (BRGNNs) trained with amino acid sequence autocorrelation (AASA) vectors and amino acid 3D autocorrelation (AA3DA) vectors yielded predictive models of the change of unfolding Gibbs free energy change (ΔΔG) of chymotrypsin Inhibitor 2 protein mutants. The ensemble predictor described about 58 and 72% of the data variances in test sets for AASA and AA3DA models, respectively. Optimum sequence and 3D-based ensembles exhibit high effects on relevant structural (volume, solvent-accessible surface area), physico-chemical (hydrophilicity/hydrophobicity-related) and thermodynamic (hydration parameters) properties.     

Modelling evolutionary cell behaviour using neural networks: Application to tumour growth

In this paper, we present a modelling framework for cellular evolution that is based on the notion that a cell’s behaviour is driven by interactions with other cells and its immediate environment. We equip each cell with a phenotype that determines its behaviour and implement a decision mechanism to allow evolution of this phenotype. This decision mechanism is modelled using feed-forward neural networks, which have been suggested as suitable models of cell signalling pathways. The environmental variables are presented as inputs to the network and result in a response that corresponds to the phenotype of the cell. The response of the network is determined by the network parameters, which are subject to mutations when the cells divide. This approach is versatile as there are no restrictions on what the input or output nodes represent, they can be chosen to represent any environmental variables and behaviours that are of importance to the cell population under consideration. This framework was implemented in an individual-based model of solid tumour growth in order to investigate the impact of the tissue oxygen concentration on the growth and evolutionary dynamics of the tumour. Our results show that the oxygen concentration affects the tumour at the morphological level, but more importantly has a direct impact on the evolutionary dynamics. When the supply of oxygen is limited we observe a faster divergence away from the initial genotype, a higher population diversity and faster evolution towards aggressive phenotypes. The implementation of this framework suggests that this approach is well suited for modelling systems where evolution plays an important role and where a changing environment exerts selection pressure on the evolving population.     

Prediction of amino acid profiles in feed ingredients:: Genetic algorithm calibration of artificial neural networks

Linear regression (LR) has been used to predict the amino acid (AA) profiles of feed ingredients, given proximate analysis (PA) input. Artificial neural networks (ANN) have also been trained to predict AA levels, generally with better results. Past projects have indicated that ANN more effectively identified the complex relationship between nutrients and feed ingredients than did LR. It was shown that the maximum R² value, a measurement of the amount of variability explained by the model, was highest when a general regression neural network (GRNN) with iterative calibration (GRNNIT) was used to train the ANN. This was in comparison to LR, Ward backpropagation (WBP) or 3-layer backpropagation (3BP) architectures. The current study investigated the potential of a new, advanced method of calibration using the genetic algorithm (GA) to optimize GRNN smoothing values. Calibration of an ANN allows the neural network to generalize well and therefore provide good results on new data. A GRNN architecture (NeuroShell 2^® Software) with GA calibration (GRNNGA) was used to train an ANN to predict AA levels in maize, soya bean meal (SBM), meat and bone meal, fish meal and wheat, based on proximate analysis input. Within the GRNNGA architecture, ANN were trained with either an Euclidean or City Block distance metric and a (0,1), (−1,1), (logistic) or (tanh) input scale. Predictive performance was judged on the basis of the maximum R² value. In general, maximum R² values were higher when the GA calibration was used in comparison to LR. For example, the highest methionine (MET) R² value for SBM was 0.54 (LR), 0.81 (3BP), 0.87 (WBP), 0.92 (GRNNIT) and 0.98 (GRNNGA). Genetic algorithm calibration of GRNN architecture led to further improvements in ANN performance for AA level predictions in most of the cases studied. Exceptions were the TSAA level in SBM (0.94 with GRNNIT vs. 0.90 with GRNNGA) and the TRY level in maize (0.88 with GRNNIT vs. 0.61 with GRNNGA).     

A comparison of methods for predicting vegetation type

Predictive modeling of vegetation patterns has wide application in vegetation science. In this paper I discuss three methods of predictive modeling using data from the alpine treeline ecotone as a case study. The study area is a portion of Glacier National Park, Montana. Parametric general linear models (GLM), artificial neural networks (ANN) and classification tree (CT) methods of predicting vegetation type are compared to determine the relative strength of each predictive approach and how they may be used in concert to increase understanding of important vegetation – environment relations. For each predictive method, vegetation type within the alpine treeline ecotone is predicted using a suite of environmental indicator variables including elevation, moisture potential, solar radiation potential, snow potential index, and disturbance history. Results from each of the predictive methods are compared against the real vegetation types to determine the relative accuracy of the methods.When the entire data field is examined (i.e., not evaluated by smaller spatial aggregates of data) the ANN procedure produces the most accurate predictions (=0.571); the CT predictions are the least accurate (=0.351). The predicted patterns of vegetation on the landscape are considerably different using the three methods. The GLM and CT methods produce large contiguous swaths of vegetation types throughout the study area, whereas the ANN method produces patterns with much more heterogeneity and smaller patches.When predictions are compared to reality at catchment scale, it becomes evident that the accuracy of each method varies depending upon the specific situation. The ANN procedure remains the most accurate method in the majority of the catchments, but both the GLM and PCT produce the most accurate classifications in at least one basin each.The variability in predictive ability of the three methods tested here indicates that there may not be a single best predictive method. Rather it may be important to use a suite of predictive models to help understand the environment – vegetation relationships. The ability to use multiple predictive methods to determine which spatial subunits of a landscape are outliers is important when identifying locations useful for climate change monitoring studies.     

基于人工神经网络的叶脉信息提取--植物活体机器识别研究I

叶片的识别是识别植物的重要组成部分,特别在野外识别植物活体尤其重要.叶脉的脉序是植物的内在特征,包含有重要的遗传信息.但由于叶脉本身的多样性,利用单一特征的图像处理方法难以有效地提取叶脉.为了充分利用图像的信息,本文提出了一种基于人工神经网络的叶脉提取方法.该方法利用边缘梯度、局部对比度和邻域统计特征等10个参数来描述像素的邻域特征,并将其作为神经网络的输入层.实验结果表明,与传统方法相比,经过训练的神经网络能够更准确地提取叶脉图像,为进一步的叶片识别打下了良好的基础.     

Late Pleistocene-Holocene radiolarian paleotemperatures in the Norwegian Sea based on artificial neural networks

Artificial Neural Networks (ANN) were trained by using an extensive radiolarian census dataset from the Nordic (Greenland, Norwegian, and Iceland) Seas. The regressions between observed and predicted Summer Sea Temperature (SST) indicate that lower error margins and better correlation coefficients are obtained for 100 m (SST₁₀₀) compared to 10 m (SST₁₀) water depth, and by using a subset of species instead of all species. The trained ANNs were subsequently applied to radiolarian data from two Norwegian Sea cores, HM 79-4 and MD95-2011, for reconstructions of SSTs through the last 15,000 years. The reconstructed SST is quite high during the Bølling-Allerød, when it reaches values only found later during the warmest phase of the Holocene. The climatic transitions in and out of the Younger Dryas are very rapid and involve a change in SST₁₀₀ of 6.2 and 6.8 °C, taking place over 440 and 140 years, respectively. SST₁₀₀ remains at a maximum during the early Holocene, and this Radiolarian Holocene Optimum Temperature Interval (RHOTI) predates the commonly recognized middle Holocene Climatic Optimum (HCO). During the 8.2 ka event, SST₁₀₀ decreases by ca. 3 °C, and this episode marks the establishment of a cooling trend, roughly spanning the middle Holocene (until ca. 4.2 ka). Successively, since then and through the late Holocene, SST₁₀₀ follows instead a statistically significant warming trend. The general patterns of the reconstructed SSTs agree quite well with previously obtained results based on application of Imbrie and Kipp Transfer Functions (IKTF) to the same two cores for SST₀. A statistically significant cyclic component of our SST record (period of 278 years) has been recognized. This is close to the de Vries or Suess cycle, linked to solar variability, and documented in a variety of other high-resolution Holocene records.     

Spatial and Temporal Variability in Sediment Denitrification Within an Agriculturally Influenced Reservoir

Reservoirs are intrinsically linked to the rivers that feed them, creating a river–reservoir continuum in which water and sediment inputs are a function of the surrounding watershed land use. We examined the spatial and temporal variability of sediment denitrification rates by sampling longitudinally along an agriculturally influenced river–reservoir continuum monthly for 13 months. Sediment denitrification rates ranged from 0 to 63 μg N₂O g ash free dry mass of sediments (AFDM)⁻¹ h⁻¹ or 0–2.7 μg N₂O g dry mass of sediments (DM)⁻¹ h⁻¹ at reservoir sites, vs. 0–12 μg N₂O gAFDM⁻¹ h⁻¹ or 0–0.27 μg N₂O gDM⁻¹ h⁻¹ at riverine sites. Temporally, highest denitrification activity traveled through the reservoir from upper reservoir sites to the dam, following the load of high nitrate (NO₃⁻-N) water associated with spring runoff. Annual mean sediment denitrification rates at different reservoir sites were consistently higher than at riverine sites, yet significant relationships among theses sites differed when denitrification rates were expressed per gDM vs. per gAFDM. There was a significant positive relationship between sediment denitrification rates and NO₃⁻-N concentration up to a threshold of 0.88 mg NO₃⁻ -N l⁻¹, above which it appeared NO₃⁻-N was no longer limiting. Denitrification assays were amended seasonally with NO₃⁻-N and an organic carbon source (glucose) to determine nutrient limitation of sediment denitrification. While organic carbon never limited sediment denitrification, all sites were significantly limited by NO₃⁻-N during fall and winter when ambient NO ₃⁻-N was low.     

A cultured human neural network operates a robotic actuator

The development of bio-electronic prostheses, hybrid human-electronics devices and bionic robots has been the aim of many researchers. Although neurophysiologic processes have been widely investigated and bio-electronics has developed rapidly, the dynamics of a biological neuronal network that receive sensory inputs, store and control information is not yet understood. Toward this end, we have taken an interdisciplinary approach to study the learning and response of biological neural networks to complex stimulation patterns. This paper describes the design, execution, and results of several experiments performed in order to investigate the behavior of complex interconnected structures found in biological neural networks. The experimental design consisted of biological human neurons stimulated by parallel signal patterns intended to simulate complex perceptions. The response patterns were analyzed with an innovative artificial neural network (ANN), called ITSOM (Inductive Tracing Self Organizing Map). This system allowed us to decode the complex neural responses from a mixture of different stimulations and learned memory patterns inherent in the cell colonies. In the experiment described in this work, neurons derived from human neural stem cells were connected to a robotic actuator through the ANN analyzer to demonstrate our ability to produce useful control from simulated perceptions stimulating the cells. Preliminary results showed that in vitro human neuron colonies can learn to reply selectively to different stimulation patterns and that response signals can effectively be decoded to operate a minirobot. Lastly the fascinating performance of the hybrid system is evaluated quantitatively and potential future work is discussed.