Time series classification for the prediction of dialysis in critically ill patients using echo statenetworks

Objective: Time series often appear in medical databases, but only few machine learning methods exist that process this kind of data properly. Most modeling techniques have been designed with a static data model in mind and are not suitable for coping with the dynamic nature of time series. Recurrent neural networks (RNNs) are often used to process time series, but only a few training algorithms exist for RNNs which are complex and often yield poor results. Therefore, researchers often turn to traditional machine learning approaches, such as support vector machines (SVMs), which can easily be set up and trained and combine them with feature extraction (FE) and selection (FS) to process the high-dimensional temporal data. Recently, a new approach, called echo state networks (ESNs), has been developed to simplify the training process of RNNs. This approach allows modeling the dynamics of a system based on time series data in a straightforwardway.The objective of this study is to explore the advantages of using ESN instead of other traditional classifiers combined with FE and FS in classification problems in the intensive care unit (ICU) when the input data consists of time series. While ESNs have mostly been used to predict the future course of a time series, we use the ESN model for classification instead. Although time series often appear in medical data, little medical applications of ESNs have been studiedyet.Methods and material: ESN is used to predict the need for dialysis between the fifth and tenth day after admission in the ICU. The input time series consist of measured diuresis and creatinine values during the first 3days after admission. Data about 830 patients was used for the study, of which 82 needed dialysis between the fifth and tenth day after admission. ESN is compared to traditional classifiers, a sophisticated and a simple one, namely support vector machines and the naive Bayes (NB) classifier. Prior to the use of the SVM and NB classifier, FE and FS is required to reduce the number of input features and thus alleviate the curse dimensionality. Extensive feature extraction was applied to capture both the overall properties of the time series and the correlation between the different measurements in the time series. The feature selection method consists of a greedy hybrid filter-wrapper method using a NB classifier, which selects in each iteration the feature that improves prediction the best and shows little multicollinearity with the already selected set. Least squares regression with noise was used to train the linear readout function of the ESN to mitigate sensitivity to noise and overfitting. Fisher labeling was used to deal with the unbalanced data set. Parameter sweeps were performed to determine the optimal parameter values for the different classifiers. The area under the curve (AUC) and maximum balanced accuracy are used as performance measures. The required execution time was also measured.Results: The classification performance of the ESN shows significant difference at the 5% level compared to the performance of the SVM or the NB classifier combined with FE and FS. The NB+FE+FS, with an average AUC of 0.874, has the best classification performance. This classifier is followed by the ESN, which has an average AUC of 0.849. The SVM+FE+FS has the worst performance with an average AUC of 0.838. The computation time needed to pre-process the data and to train and test the classifier is significantly less for the ESN compared to the SVM andNB.Conclusion: It can be concluded that the use of ESN has an added value in predicting the need for dialysis through the analysis of time series data. The ESN requires significantly less processing time, needs no domain knowledge, is easy to implement, and can be configured using rules ofthumb.     

Multi-month memory effects on early summer vegetative activity in semi-arid South Africa and their spatial heterogeneity

In semi-arid African regions (annual rainfall between 200 and 600 mm), variability of vegetative activity is mainly due to the rainfall of the current rainy season. In most of South Africa, the rainy season occurs from October to March. On average, vegetative activity lags rainfall by 1 to 2 months. The interannual variability in early summer (December to September) normalized difference vegetation index (NDVI) depends primarily on precipitation at the beginning (October to November) of the rainy season. However, once this primary control is removed, the residual interannual variability in NDVI highlights a double memory effect: a 1-year effect, referred to as Mem₁, and a 7- to 10-month effect, referred to as Mem₂. This article aims at better describing the influence of soil and vegetation characteristics on these two memory effects. The data sets used in this study are as follows: (1) a 19-year NDVI time series from National Oceanic and Atmospheric Administration (NOAA) satellites, (2) rainfall records from a network of 1160 rain-gauge stations compiled by the Water Research Commission (WRC), (3) vegetation types from Global Land Cover (GLC) 2000 and (4) soil characteristics from the soil and terrain database for Southern Africa (SOTERSAF). Results indicate that among 20–30% of NDVI variance that is not explained by the concurrent rainfall, one-third is explained by the two memory effects. Mem₁ is found to have maximum effect in the northwest of our study domain, near the Botswana boundary, in the South Kalahari. Associated conditions are open grasslands growing on Arenosols. Mem₁ is less important in the southeast, particularly in open grassland with shrubs growing on Cambisols. Thus, Mem₁ mainly depends on soil texture. Mem₂ is more widespread and its influence is the greatest in the centre, the south and the east of our domain. It is related to rainfall from January to April, which controls, beyond the intervening dry season, the interannual variations of NDVI (December to September) at the beginning of the next rainy season. Through these new findings, this article emphasizes again the high potential of remote-sensing techniques to monitor and understand the dynamics of semi-arid environments.     

How reliable are the consumers? Comparison of sensory profiles from consumers and experts

This study compares expert and consumer sensory profiles for the same 12 perfumes in different ways: the discriminatory ability and reproducibility are analyzed through ANOVA and the panelists’ consensus through the correlation coefficients. Next, the two product spaces are first analyzed separately for each panel, and then compared through multiple factor analysis. Finally, the two panels are compared using the confidence ellipses methodology. These analyses show that the two panels give similar results with respect to the important criteria for panels (discrimination, consensus, reproducibility). The comparison of the two products spaces shows high similarity. From the confidence ellipses, it can be concluded that no significant differences exist for a given product between the two panels. Hence, in this particular case, the use of consumers appears to be a good alternative to the classical sensory profile provided by a trained panel.     

SPOT5-HRS digital elevation models and the monitoring of glacier elevation changes in North-West Canada and South-East Alaska

Monitoring the response of land ice to climate change requires accurate and repeatable topographic surveys. The SPOT5-HRS (High Resolution Stereoscopic) instrument covers up to 120 km by 600 km in a single pass and has the potential to accurately map the poorly known topography of most glaciers and ice caps. The acquisition of a large HRS archive over ice-covered regions is planned by the French Space Agency (CNES) and Spotimage, France during the 2007–2008 International Polar Year (IPY). Here, we report on the accuracy and value of HRS digital elevation model (DEM) over ice and snow surfaces.

A DEM is generated by combining tools available from CNES with the PCI Orthoengine^SE software, using HRS images acquired in May 2004 over South-East Alaska (USA) and northern British Columbia (Canada). The DEM is evaluated through comparison with shuttle radar topographic mission (SRTM) DEM and ICESAT data, on and around the glaciers. A horizontal shift of 50 m is found between the HRS and SRTM DEMs and is attributed to errors in the SRTM DEM. Over ice-free areas, HRS elevations are 7 m higher than those of SRTM, with a standard deviation of ± 25 m for the difference between the two DEMs. The 7-m difference is partly attributed to the differential penetration of the electromagnetic waves (visible for HRS; microwave for SRTM) in snow and vegetation.

We also report on the application of sequential DEMs (SRTM DEM in February 2000 and HRS DEM in May 2004) for the monitoring of glacier elevation changes. We map the topographic changes induced by a surge of one tributary of Ferris Glacier. Maximum surface lowering of 42 (± 10) m and rising of 77 (± 10) m are observed in the 4 years time interval. Thinning rates up to 10 (± 2.5) m/yr are observed at low altitudes and confirm the ongoing wastage of glaciers in South-East Alaska.     

Posture and muscular behaviour in emergency braking: An experimental approach

In the field of numerical crash simulations in road safety research, there is a need to accurately define the initial conditions of a frontal impact for the car occupant. In particular, human models used to simulate such impacts barely take into account muscular contracting effects. This study aims to quantify drivers’ behaviour in terms of posture and muscular activity just before a frontal impact.Experiments on volunteers were performed in order to define these conditions, both on a driving simulator and on a real moving car. Brake pedal loads, lower limbs kinematics and muscle activation were recorded.Coupling instantaneous data from both experimental protocols (simulator versus Real car), a standard emergency braking configuration could be defined as (1) joint flexion angles of 96°, 56° and 13° for the right hip, knee and ankle respectively; (2) a maximum brake pedal load of 780 N; (3) a muscular activation of 55% for the anterior thigh, 26% for the posterior thigh, 18% for the anterior leg and 43% for the posterior leg.The first application of this research is the implementation of muscle tone in human models designed to evaluate new safety systems.     

Composition and insecticidal activity of essential oil from Pistacia lentiscus L. against Ectomyelois ceratoniae Zeller and Ephestia kuehniella Zeller (Lepidoptera: Pyralidae)

This study reports investigations on the chemical constituents and fumigant toxicity of Pistacia lentiscus L. (Anacardiaceae) essential oil against two major stored-date insects, carob moth Ectomyelois ceratoniae Zeller and Mediterranean flour moth Ephestia kuehniella Zeller. Results showed that P. lentiscus essential oil contained terpinene-4-ol (23.32%), α-terpineol (7.12%) and β-caryophyllene (22.62%) as major compounds. Fumigant toxicity tests showed that P. lentiscus oil was more toxic to E. kuehniella (LC₅₀ = 1.84 μl/l, LC₉₅ = 5.14 μl/l) than E. ceratoniae (LC₅₀ = 3.29 μl/l, LC₉₅ = 14.24 μl/l). The fecundity and hatching rate of both insects decreased with increases in concentration or exposure time to the oil. At 136 μl/l air, fecundities and hatching rates were respectively 35 eggs/female and 42.86% for E. ceratoniae and 78 eggs/female and 29.49% for E. kuehniella.     

Net N immobilisation during the biodegradation of mucilage in soil as affected by repeated mineral and organic fertilisation

Repeated fertilisation of crops with manure commonly increases both the amount and the quality of soil organic matter, the mineralisation of which strongly determines the availability of N to plants. In the rhizosphere, the microbial mineralisation/immobilisation of N is enhanced due to rhizodeposition (release of organic compounds from roots). In this work, we studied N transformations during incubation of maize root mucilage in soils that had been previously fertilized with composted pig manure or with ammonium nitrate for 7 years. Our work revealed mucilage was rapidly mineralised (average half-life of 3 days), inducing a rapid N immobilisation of 94 mg N g⁻¹ of mucilage C, followed by a slow remineralisation. Fertilisation with manure induced a persistent stimulation of the soil organic matter mineralisation, leading to an enhanced content of soil inorganic N (+23% in 58 days of incubation). Due to this stimulation of microbial activity in the soil fertilized with manure, the decrease in inorganic N during the biodegradation of mucilage lasted a shorter time as compared to the mineral fertilisation. However, the type of fertilisation did not significantly change the amount of N immobilised.     

Clinical Determination of Dry Body Weight

While nephrologists wait for the ideal, non invasive, inexpensive, precise, and reproducible tool to evaluate extracellular volume (ECV), they need to exert their clinical acumen in the quest of that holy grail, dry weight (DW). Estimation of DW using a clinical approach based on blood pressure (BP) and ECV is feasible and reliable as shown by successful experiences in various dialysis modes over more than three decades. But a need still exists to resolve difficulties associated with accurate assessment of BP (methods and circumstances of measurement, and the confounding effects of antihypertensive drugs) and ECV (evaluation of weight changes unrelated to ECV, lack of specificity and sensitivity of clinical symptoms, lag time, confusion in terminology). An essential point in clinical assessment of DW is that a normal BP is at the same time the target and the crucial index of DW achievement. For this reason, a trialand‐error “probe” process has to be used at intervals to make sure that the dry weight target point is correctly estimated. The various “non clinical” methods proposed for dry weight assessment increase the complexity and the cost of hemodialysis. They are, in the present state of things, more clinical research than practice tools. They do not replace clinical judgment.