Comparison of Convolutional Neural Networks Based Method and LCmodel on the Quantification of in Vivo Magnetic Resonance Spectroscopy

Quantification of metabolites concentrations in institutional unit (IU) for between subject and long-term comparison is considered important strategy in the applications of magnetic resonance spectroscopy (MRS). The aim of this study is to investigate if metabolite concentrations quantified by convolutional neuronal network (CNN) based method associated with a proposed scaling procedure can reflect variations of the metabolite concentrations in institution unit (IU) at different brain regions with different signal-to-noise-ratio (SNR) and linewidth (LW). An error index based on standard error (SE) is proposed to indicate the confidence levels on the prediction for metabolites. In vivo MRS spectra were collected at 3 brain regions from 44 subjects at 3T system. Metabolite concentrations in IU quantified by LCModel and CNN from 44 subjects were compared. For in vivo spectra characterized under different spectral quality in terms of SNR and LW, line narrowing and noise free spectra were successfully exported by CNN. Concentrations of five metabolites quantified by CNN and LCModel are in similar range with statistically significant Pearson’s correlation coefficients (0.28~0.70). SE of the metabolites show positive correlation with Cramer-Rao lower bound (CRLB) (r=0.60) and with absolute CRLB (r=0.84). In conclusion, the CNN based method with the proposed scaling procedures can be used to quantify in vivo MRS spectra. The concentrations of five major metabolites were reported in IU, which are in the same range as those quantified using a routine MRS quantification procedures by LCModel. The SE can be used as error index indicating predicted uncertainties for metabolites with the information similar to the absolute CRLB.

The Relevance of Geotechnical-Unit Characterization for Landslide-Susceptibility Mapping with SHALSTAB

Given the increasing occurrence of landslides worldwide, the improvement of predictive models for landslide mapping is needed. Despite the influence of geotechnical parameters on SHALSTAB model outputs, there is a lack of research on models’ performance when considering different variables. In particular, the role of geotechnical units (i.e., areas with common soil and lithology) is understudied. Indeed, the original SHALSTAB model considers that the whole basin has homogeneous soil. This can lead to the under-or-overestimation of landslide hazards. Therefore, in this study, we aimed to investigate the advantages of incorporating geotechnical units as a variable in contrast to the original model. By using locally sampled geotechnical data, 13 slope-instability scenarios were simulated for the Jaguar creek basin, Brazil. This allowed us to verify the sensitivity of the model to different input variables and assumptions. To evaluate the model performance, we used the Success Index, Error Index, ROC curve, and a new performance index: the Detective Performance Index of Unstable Areas. The best model performance was obtained in the scenario with discretized geotechnical units’ values and the largest sample size. Results indicate the importance of properly characterizing the geotechnical units when using SHALSTAB. Hence, future applications should consider this to improve models’ predictivity.

Assessment of earthquake-induced landslide inventories and susceptibility maps using slope unit-based logistic regression and geospatial statistics

Inventories of seismically induced landslides provide essential information about the extent and severity of ground effects after an earthquake. Rigorous assessment of the completeness of a landslide inventory and the quality of a landslide susceptibility map derived from the inventory is of paramount importance for disaster management applications. Methods and materials applied while preparing inventories influence their quality, but the criteria for generating an inventory are not standardized. This study considered five landslide inventories prepared by different authors after the 2015 Gorkha earthquake, to assess their differences, understand the implications of their use in producing landslide susceptibility maps in conjunction with standard landslide predisposing factors and logistic regression. We adopted three assessment criteria: (1) an error index to identify the mutual mismatches between the inventories; (2) statistical analysis, to study the inconsistency in predisposing factors and performance of susceptibility maps; and (3) geospatial analysis, to assess differences between the inventories and the corresponding susceptibility maps. Results show that substantial discrepancies exist among the mapped landslides. Although there is no distinct variation in the significance of landslide causative factors and the performance of susceptibility maps, a hot spot analysis and cluster/outlier analysis of the maps revealed notable differences in spatial patterns. The percentages of landslide-prone hot spots and clustered areas are directly proportional to the size of the landslide inventory. The proposed geospatial approaches provide a new perspective to the investigators for the quantitative analysis of earthquake-triggered landslide inventories and susceptibility maps.

Neural networks for long-term earthquake prediction using modified meta-learning

Neural networks have been successfully applied for modeling time series. However, the results of long-term prediction are not satisfied. In this paper, the modified Meta-Learning is applied to the neural model. The normal Meta-Learning is modified by time-varying learning rates and adding a momentum term to improve convergence speed and robustness property. The stability of the learning process is proven. Finally, two experiments are presented to evaluate the proposed method. The first one shows an improvement in earthquakes prediction in the long-term, and the second one is a classical Benchmark problem. In both experiments, the modified Meta-Learning technique minimizes remarkably the mean square error index.

RISK PREDICTION AND DIAGNOSIS OF WATER SEEPAGE IN OPERATIONAL SHIELD TUNNELS BASED ON RANDOM FOREST

Water seepage (WS) is a paramount defect during tunnel operation and directly affects the operational safety of tunnels. Effectively predicting and diagnosing WS are problems that urgently need to be solved. This paper presents a standard and an evaluation index system for WS grades and constructs a sample dataset from monitoring recoreds for demonstration purposes. First, we use bootstrap resampling to build a random forest (RF) seepage risk prediction model. Second, the optimal branch and parameters are selected by the 5-fold cross-validation method to establish the RF prediction training model. Additionally, to illustrate the effectiveness of the method, the operational stage of Wuhan Metro Line 3 in China is taken as a case study. The results conclude that the segment spalling area, crack width, and loss rate of the rebar cross-section have a strong influence on WS. Finally, the test data are predicted, and the prediction result error index is calculated. Compared with the predictions of some traditional machine learning methods, such as support vector machines and artificial neural networks, RF prediction has the highest accuracy and is the closest to the true value, which demonstrates the accuracy of the model and its application potential.

Evaluating the capability of hybrid data-driven approaches to forecast monthly streamflow using hydrometric and meteorological variables

Streamflow forecasting, as one of the most important issues in hydrological studies, plays a vital role in several aspects of water resources management such as reservoir operation, water allocation, and flood forecasting. In this study, wavelet-gene expression programming (WGEP) and wavelet-M5 prime (WM5P) techniques, as two robust artificial intelligence (AI) models, were applied for forecasting the monthly streamflow in Khoshkroud and Polroud Rivers located in two basins with the same names. Results of hybrid AI techniques were compared with those achieved by two stand-alone models of GEP and M5P. Seven combinations of hydrological (H) and meteorological (M) variables were considered to investigate the effect of climatic variables on the performance of the proposed techniques. Moreover, the performance of both stand-alone and hybrid models were evaluated by statistical criteria of correlation of coefficient, root-mean-square error, index of agreement, the Nash–Sutcliffe model efficiency coefficient, and relative improvement. The statistical results revealed that there is a dependency between ‘the M5P and GEP performance’ and ‘the geometric properties of basins (e.g., area, shape, slope, and river network density)’. It was found that a preprocessed technique could increase the performance of M5P and GEP models. Compared to the stand-alone techniques, the hybrid AI models resulted in higher performance. For both basins, the performance of the WM5P model was higher than the WGEP model, especially for extreme events. Overall, the results demonstrated that the proposed hybrid AI approaches are reliable tools for forecasting the monthly streamflow, while the meteorological and hydrometric variables are taken into account.

Minimisation of Overestimation of River Flows in 1D- Hydrodynamic Modeling

One-dimensional hydrodynamic models overestimate river cross-section derived from freely available SRTM DEMs. The present study aims to minimize the overestimation of river flow. DEM-extracted cross-sections obtained from 30 m and 90 m resolutions show higher elevation values than the actual river cross sections of Krishna and Bhima rivers, India. To minimize the overestimation of the river flow, DEM-extracted cross-sections are modified using known cross-section of the river. The corrections for cross sections extracted from DEM, are obtained by subtracting the DEM-derived cross-sections from a known cross-section of the river. Monsoons flows that occurred in years 2006 and 2009 in Krishna and Bhimariver have been used for modeling. The MIKE HYDRO River model performance with modified DEM-extracted cross-sections of river improves as the correlation coefficient, root mean square error, index of agreement, Nash Sutcliffe efficiency & Percentage deviation in peak (%) values are improved.

Compound Grey-Logistic Model and Its Application

Logistic regression model is widely used in ecology and in the analysis of social economic systems, because of its good adaptability. In order to improve the measurement accuracy of logistic model, this paper proposes a new method. A compound grey-logistic model is developed to carry out the grey transformation of the original data. Practice shows that the grey transformation data has better simulation accuracy; at the same time, grey transformation can reduce the observation noise of the original data. Mean absolute percentage error index has been used to evaluate the accuracy of prediction model, and information entropy can be used to evaluate the change of information entropy of forecasting data. In this paper, three cases are used to verify the applicability of grey-logistic model. From the perspective of the type of original data, the three cases represent three different data conditions: sufficient data, insufficient data, and fragmentary data. The cases represent different related fields: market share data, economic growth data, and R&D output data. The results show that the proposed grey-logistic method can effectively carry out the population growth analysis.