Cefepime precision dosing tool: From Standard to Precise Dose Using Nonparametric Population Pharmacokinetics

Cefepime is the second most common cephalosporin used in U.S. hospitals. We aim to develop and validate cefepime population pharmacokinetic (PK) model and integrate into precision dosing tool for implementation. Two datasets (680 patients) were used to build cefepime PK model in Pmetrics, and three datasets (34 patients) were used for the validation. A separate application dataset (115 patients) was used for the implementation and validation of a precision dosing tool. The model support points and covariates were used to generate the optimal initial dose (OID). Cefepime PK was described by a two-compartment model including weight and creatinine clearance (CrCl) as covariates. The median rate of elimination was 0.30 hr −1 (adults) and 0.96 hr −1 (pediatrics), central volume of distribution 13.85 L, and rate of transfer from the central to the peripheral compartments 1.22 hr −1 and from the peripheral to the central compartments 1.38 hr −1 . After integration in BestDose, the observed vs. predicted cefepime concentration fit using the application dataset was excellent (R 2 >0.98) and the median difference between observed and what BestDose predicted in a second occasion was 4%. For OID, cefepime 0.5-1g 4-hour infusion q8-24hr with CrCl<70 mL/min was needed to achieve a target range of free trough:MIC 1-4 at MIC 8 mg/L, while continuous infusion was needed for higher CrCl and weight values. In conclusion, we developed and validated a cefepime model for clinical application. The model was integrated in a precision dosing tool for implementation and the median concentration prediction bias was 4%. OID algorithm was provided.

A Graph Neural Network gas concentration prediction model based on Spatio-Temporal data

For the problem of low prediction accuracy caused by traditional neural network gas concentration prediction models which did not consider temporal and spatial characteristics of gas data, this paper proposed a Spatial-Temporal graph neural network gas prediction model based on Spatial-Temporal data. Its essence was the integration of graph convolutional network and WaveNet network. In spatial dimension, graph convolutional network was used to aggregate the information of neighbor nodes, and adaptively adjusts the spatial association strength of each node according to the attention mechanism to captured the spatial characteristics of gas data. In temporal dimension, WaveNet network model was introduced, Dilated Causal Convolution was used to extract the temporal characteristics of gas data on temporal dimensions. According to the distance between gas sensors in the mine, the gas data spatial structure was constructed by Thresholded Gaussian kernel function. Experiment with the measured gas temporal and spatial data, using Mean Absolute Error (MAE) as an indicator of predictive accuracy. The experimental results show that the prediction model mentioned in this paper is significantly improved compared with the prediction accuracy of other predictive models.

Air Pollutant Concentration Prediction Based on a CEEMDAN-FE-BiLSTM Model

The concentration series of PM2.5 (particulate matter ≤ 2.5 μm) is nonlinear, nonstationary, and noisy, making it difficult to predict accurately. This paper presents a new PM2.5 concentration prediction method based on a hybrid model of complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) and bi-directional long short-term memory (BiLSTM). The new method was applied to predict the same kind of particulate pollutant PM10 and heterogeneous gas pollutant O3, proving that the prediction method has strong generalization ability. First, CEEMDAN was used to decompose PM2.5 concentrations at different frequencies. Then, the fuzzy entropy (FE) value of each decomposed wave was calculated, and the near waves were combined by K-means clustering to generate the input sequence. Finally, the combined sequences were put into the BiLSTM model with multiple hidden layers for training. We predicted the PM2.5 concentrations of Seoul Station 116 by the hour, with values of the root mean square error (RMSE), the mean absolute error (MAE), and the symmetric mean absolute percentage error (SMAPE) as low as 2.74, 1.90, and 13.59%, respectively, and an R2 value as high as 96.34%. The “CEEMDAN-FE” decomposition-merging technology proposed in this paper can effectively reduce the instability and high volatility of the original data, overcome data noise, and significantly improve the model’s performance in predicting the real-time concentrations of PM2.5.

Rapid Algae Identification and Concentration Prediction Based on Discrete Excitation Fluorescence Spectra

In this paper, an algal identification and concentration determination method based on discrete excitation fluorescence spectra is proposed for online algae identification and concentration prediction. The discrete excitation fluorescence spectra of eight species of harmful algae from four algal categories were assessed. After determining typical excitation wavelengths according to the distribution of photosynthetic pigments and eliminating strongly correlated wavelengths by applying the hierarchical clustering, seven characteristic excitation wavelengths (405, 435, 470, 490, 535, 555, and 590 nm) were selected. By adding the ratios between feature points (435 and 470 nm, 470 and 490 nm, as well as 535 and 555 nm), standard feature spectra were established for classification. The classification accuracy in pure samples exceeded 95%, and that of dominant algae species in a mixed sample was 77.4%. Prediction of algae concentration was achieved by establishing linear regression models between fluorescence intensity at seven characteristic excitation wavelengths and concentrations. All models performed better at low concentrations, not exceeding the threshold concentration of red tide algae outbreak, which provides a proximate cell density of dominant algal species.

Research on time series characteristics of the gas drainage evaluation index based on lasso regression

The evaluation of the coal mine gas drainage effect is affected by many factors, such as flow rate, wind speed, drainage negative pressure, concentration, and temperature. This paper starts from actual coal mine production monitoring data and based on the lasso regression algorithm, features selection of multiple parameters of the preprocessed gas concentration time series to construct gas concentration feature selection based on the algorithm. The three-time smoothing index method is used to fill in the missing values. Aiming at the problem of different dimensions in the gas concentration time series, the MinMaxScaler method is used to normalize the data. The lasso regression algorithm is used to perform feature selection on the multivariable gas concentration time series, and the gas concentration time series selected by the lasso feature and the gas concentration time series without feature selection are input. The performance of the ANN algorithm for gas concentration prediction is compared and analyzed. The optimal α value and L1 norm are selected based on the grid search method to determine the strong explanatory gas concentration time series feature set of the working face, and an experimental comparison of the gas concentration prediction results before and after the lasso feature selection is performed. We verify the effectiveness of the algorithm.

A Long Short-Term Memory Neural Network for Daily NO2 Concentration Forecasting

NO2 pollution has caused serious impact on people's production and life, and the management task is very difficult. Accurate prediction of NO2 concentration is of great significance for air pollution management. In this paper, a NO2 concentration prediction model based on long short-term memory neural network (LSTM) is constructed with daily NO2 concentration in Beijing as the prediction target and atmospheric pollutants and meteorological factors as the input indicators. Firstly, the parameters and architecture of the model are adjusted to obtain the optimal prediction model. Secondly, three different sets of input indicators are built on the basis of the optimal prediction model to enter the model learning. Finally, the impact of different input indicators on the accuracy of the model is judged. The results show that the LSTM model has high application value in NO2 concentration prediction. The maximum temperature and O3 among the three input indicators improve the prediction accuracy while the NO2 historical low-frequency data reduce the prediction accuracy.