Direct Regression Methods Research Articles

A task-unified network with transformer and spatial–temporal convolution for left ventricular quantification

Quantification of the cardiac function is vital for diagnosing and curing the cardiovascular diseases. Left ventricular function measurement is the most commonly used measure to evaluate the function of cardiac in clinical practice, how to improve the accuracy of left ventricular quantitative assessment results has always been the subject of research by medical researchers. Although considerable efforts have been put forward to measure the left ventricle (LV) automatically using deep learning methods, the accurate quantification is yet a challenge work as a result of the changeable anatomy structure of heart in the systolic diastolic cycle. Besides, most methods used direct regression method which lacks of visual based analysis. In this work, a deep learning segmentation and regression task-unified network with transformer and spatial–temporal convolution is proposed to segment and quantify the LV simultaneously. The segmentation module leverages a U-Net like 3D Transformer model to predict the contour of three anatomy structures, while the regression module learns spatial–temporal representations from the original images and the reconstruct feature map from segmentation path to estimate the finally desired quantification metrics. Furthermore, we employ a joint task loss function to train the two module networks. Our framework is evaluated on the MICCAI 2017 Left Ventricle Full Quantification Challenge dataset. The results of experiments demonstrate the effectiveness of our framework, which achieves competitive cardiac quantification metric results and at the same time produces visualized segmentation results that are conducive to later analysis.

ER-Net: Efficient Recalibration Network for Multi-View Multi-Person 3D Pose Estimation

Multi-view multi-person 3D human pose estimation is a hot topic in the field of human pose estimation due to its wide range of application scenarios. With the introduction of end-to-end direct regression methods, the field has entered a new stage of development. However, the regression results of joints that are more heavily influenced by external factors are not accurate enough even for the optimal method. In this paper, we propose an effective feature recalibration module based on the channel attention mechanism and a relative optimal calibration strategy, which is applied to the multi-view multi-person 3D human pose estimation task to achieve improved detection accuracy for joints that are more severely affected by external factors. Specifically, it achieves relative optimal weight adjustment of joint feature information through the recalibration module and strategy, which enables the model to learn the dependencies between joints and the dependencies between people and their corresponding joints. We call this method as the Efficient Recalibration Network (ER-Net). Finally, experiments were conducted on two benchmark datasets for this task, Campus and Shelf, in which the PCP reached 97.3% and 98.3%, respectively.

Calibration and Validation of CYGNSS Reflectivity through Wetlands’ and Deserts’ Dielectric Permittivity

The reflection of Global Navigation Satellite Systems (GNSS) signals, namely GNSS-Reflectometry (GNSS-R), has recently proven to be able to monitor land surface properties in the microwave spectrum, at a global scale, and with very low revisiting time. Moreover, this new technique has numerous additional advantages, including low cost, low power consumption, lightweight and small payloads, and near real-time massive data availability, as compared to conventional monostatic microwave remote sensing. However, the GNSS-R surface reflectivity values estimated through the bistatic radar equation, and the Fresnel coefficients have shown a lack of coincidence with real surface reflectivity data, mostly due to calibration issues. Previous studies have attempted to avoid this matter with direct regression methods between uncalibrated GNSS-R reflectivity data and external soil moisture content (SMC) products. However, calibration of GNSS-R reflectivity used in traditional inversion models is still a challenge, such as those to estimate SMC, freeze/thaw, or biomass. In this paper, a successful procedure for GNSS-R reflectivity calibration is established using data from the CYGNSS (Cyclone GNSS) constellation. The scale and bias parameters are estimated from the theoretical dielectric properties of water and dry sand, which are well-known and empirically validated values. We employ four calibration areas that provide maximum range limits of reflectivity, such as deserts and wetlands. The CYGNSS scale factor and the bias parameter resulted in a = 3.77 and b = 0.018, respectively. The derived scale and bias parameters are applied to the CYGNSS dataset, and the retrieved SMC values through the Fresnel reflection coefficients are in excellent agreement with the Soil Moisture Active Passive (SMAP) SMC product. Then, the SMAP SMC is used as a reference true value, and provides a standard linear regression with an R-square coefficient of 0.803, a root mean square error (RMSE) of 0.084, and a Pearson’s correlation coefficient of 0.896.

Left Ventricle Quantification Challenge: A Comprehensive Comparison and Evaluation of Segmentation and Regression for Mid-Ventricular Short-Axis Cardiac MR Data.

Automatic quantification of the left ventricle (LV) from cardiac magnetic resonance (CMR) images plays an important role in making the diagnosis procedure efficient, reliable, and alleviating the laborious reading work for physicians. Considerable efforts have been devoted to LV quantification using different strategies that include segmentation-based (SG) methods and the recent direct regression (DR) methods. Although both SG and DR methods have obtained great success for the task, a systematic platform to benchmark them remains absent because of differences in label information during model learning. In this paper, we conducted an unbiased evaluation and comparison of cardiac LV quantification methods that were submitted to the Left Ventricle Quantification (LVQuan) challenge, which was held in conjunction with the Statistical Atlases and Computational Modeling of the Heart (STACOM) workshop at the MICCAI 2018. The challenge was targeted at the quantification of 1) areas of LV cavity and myocardium, 2) dimensions of the LV cavity, 3) regional wall thicknesses (RWT), and 4) the cardiac phase, from mid-ventricle short-axis CMR images. First, we constructed a public quantification dataset Cardiac-DIG with ground truth labels for both the myocardium mask and these quantification targets across the entire cardiac cycle. Then, the key techniques employed by each submission were described. Next, quantitative validation of these submissions were conducted with the constructed dataset. The evaluation results revealed that both SG and DR methods can offer good LV quantification performance, even though DR methods do not require densely labeled masks for supervision. Among the 12 submissions, the DR method LDAMT offered the best performance, with a mean estimation error of 301 mm 2 for the two areas, 2.15 mm for the cavity dimensions, 2.03 mm for RWTs, and a 9.5% error rate for the cardiac phase classification. Three of the SG methods also delivered comparable performances. Finally, we discussed the advantages and disadvantages of SG and DR methods, as well as the unsolved problems in automatic cardiac quantification for clinical practice applications.

Comparison of vapour-exchange methods for predicting hourly twig fuel moisture contents of larch and birch stands in the Daxinganling Region, China

The Daxinganling Region, in Heilongjiang Province, China, has experienced considerable losses to forest resources owing to a recent increase in forest fire frequency. This necessitates the predictions of fire behaviour in this region, which requires assessment of fuel moisture content (FMC) at appropriate space and time intervals. Here, we compare and extrapolate results of vapour-exchange methods for predicting hourly moisture contents of downed wood. The FMC of larch and birch twig fuels was measured at 1-h intervals at different slope positions with respect to meteorological factors. The twigs were categorised based on diameter and degree of decay. The Nelson and Simard methods (which consider equilibrium moisture content and the associated time lag) and direct regression method (which allows direct attainment of the FMC) were used. Both Nelson and Simard methods predicted hourly twig moisture content more accurately than the direct regression method. The Simard method was the most accurate, with the lowest mean absolute error and mean relative error as well as relatively small variation coefficients based on results of the extrapolation analysis. Further, the resulting models applied to needle-leaved and lightly decomposed fuels performed better than for the broad-leaved and badly decomposed fuels on evaluation parameters.

Regional flood frequency analysis for flood index estimation in hydrologic regions with limited flood data

The objective of this study is to establish a regional relationship between mean annual peak flood and the catchment area based on the annual peak flood frequency analysis. Annual peak flood time series for various gauging sites of six basins in north part of Iraq are used in analysis. The procedure of analysis is approached by scaling a regional flood frequency distribution (index-flood of catchment) which is defined here as the mean annual maximum flood discharge and estimated by linear regression using physiographic and climatic catchment properties. The index-flood method was found to have slightly better prediction accuracies over the direct-regression method. The coefficient of determination R2 between reference and estimated index-floods is relatively high in most cases for all the gauging stations. All regions give very high R 2 correlation and the best model is the model (2), which is also associated with minimum value of RMSE.

A Direct Regression Scene Text Detector With Position-Sensitive Segmentation

Direct regression methods have demonstrated their success on various multi-oriented benchmarks for scene text detection due to the high recall rate for small targets and the direct regression for text boxes. However, too many false positive candidates and inaccurate position regression still limit the performance of these methods. In this paper, we propose an end-to-end method by introducing position-sensitive segmentation into the direct regression method to overcome these shortcomings. We generate the ground truth of position-sensitive segmentation maps based on the information of text boxes so that the position-sensitive segmentation module can be trained synchronously with the direct regression module. Besides, more information about the relative position of text is provided for the network through the training of position-sensitive segmentation maps, which improves the expressiveness of the network. We also introduce spatial pyramid of position-sensitive segmentation into the proposed method considering the huge differences in sizes and aspect ratios of scene texts and we propose position-sensitive COI(Corner area of Interest) pooling into the proposed method to speed up the inference. Experiments on datasets ICDAR2015, MLT-17 and COCO-Text demonstrate that the proposed method has a comparable performance with state-of-the-art methods while it is more efficient. We also provide abundant ablation experiments to demonstrate the effectiveness of these improvements in our proposed method.

There and back again – estimating equivalence scales with measurement error

ABSTRACTUsing income satisfaction data from the German Socio-Economic Panel, we find large differences in the equivalence weight of a partner when it is being estimated by direct and reverse regressions. We argue that neither of the two models will produce consistent estimates when there is stochastic error in satisfaction and measurement error in incomes. We propose a correction of mismeasured incomes using a constructed alternative income measure. The corrected results are relatively close to those obtained from direct regression. We do not find evidence that previous studies, using the direct regression method, severely suffer from measurement error in incomes.

Predicting hourly litter moisture content of larch stands in Daxinganling Region, China using three vapour-exchange methods

Fuel moisture affects fuel ignition potential and fire behaviour. To accurately model fire behaviour, predict fuel ignition potential and plan fuel reduction, fuel moisture content must be assessed regularly and often. To establish models for Daxinganling Region, which has the most severe forest fires in China, hourly measurements were taken of moisture content in litter beds of larch stands sampled under different shading and slope conditions. Models were established using three vapour-exchange methods. The Nelson and Simard methods employed a direct timelag method using a timelag concept and the Nelson and Simard equilibrium moisture content (EMC) functions and estimating model parameters directly from fuel moisture and weather observation data in the field. The direct regression method used equations directly derived from linear regression of fuel moisture and field weather variation. The mean absolute error and mean relative error were determined for the Nelson (0.78%, 4.98%), Simard (1.04%, 5.57%) and direct regression (1.48%, 9.01%) methods. Only the models established using the direct timelag methods met the 1% accuracy requirement using either the Nelson or Simard EMC function, confirming the suitability and robustness of the direct timelag methods. The Simard and Nelson methods had similar accuracy, but Simard was more robust and only needed estimation of one parameter and hence is recommended for predicting litter moisture in this region.

Moisture sorption isotherms and net isosteric heat of sorption for spray-dried pure orange juice powder

The aim of this study was to measure the moisture sorption isotherms of spray-dried pure orange-juice powders. These isotherms have been fitted using eleven different sorption equations. The GAB model was found to be the best fit. The GAB parameters were estimated both by direct and indirect regression methods. The net isosteric heat of sorption was also determined using the Clausius–Clapeyron equation and was successfully used as a limiting condition for estimating the six unknown parameters of the GAB equation using the direct regression method. The net isosteric heat of sorption decreased from 9.05 kJ/mol at 0.05 kg/kg dry basis to approximately 0.0 kJ/mol at 0.7 kg/kg dry basis. These values, as expected, were comparable to the enthalpy difference between the monolayer and multilayer moisture (8.16 kJ/mol), and the difference between the latent heat of condensation for pure water and the heat of sorption of multilayer moisture (=−0.88 ± 0.69 kJ/mol).

Indirect modeling of hourly meteorological time series for winter road maintenance

Indirect models provide an intuitive and effective means for incorporating knowledge on structured class definitions or functional relationships into classification and regression models for complex data. We transfer the approach of generalized indirect classification to the regression context, and show in a case study that the predictive performance of direct regression methods can be significantly improved by incorporating classification models for intermediate categorical variables to construct indirect regression models. We apply this approach to model winter road maintenance on Ontario highways as a function of hourly time series of meteorological data. Two ordinal variables representing different salt application modes are used as intermediate variables. Stepwise linear regression as a direct regression technique is outperformed by an indirect linear regression using bagging and penalized linear discriminant analysis (PLDA) as intermediate models. In addition, direct bagging is outperformed by indirect bagging with classification trees and PLDA as intermediate models. The latter indirect method achieves the best performance of all four approaches examined, reducing the median bootstrapped RMSE significantly by 6.9–10.4% compared to the other three methods. The bootstrap is performed at the day level to account for temporal autocorrelation. In addition, permutation-based variable accuracy importance measures are applied to assess the utility of different variables and data sources. Copyright © 2010 John Wiley & Sons, Ltd and Crown in the right of Canada.

COMPARISON OF SIX EXTREME FLOOD ESTIMATION TECHNIQUES FOR UNGAUGED WATERSHEDS IN COASTAL BRITISH COLUMBIA

A comparison of six techniques for the estimation of peak flow from ungauged watersheds in coastal British Columbia is carried out. The techniques compared are a new physically-based stochastic-deterministic procedure (Loukas and Quick, 1994a), the British Columbia Ministry of Environment and Parks method, a statistical method based on the Bayes theorem and proposed by Russell (1982), and the well known regional techniques, the Index Flood method, the method of Direct Regression of Quantiles, and the method of Regression of the Distribution Parameters. The techniques are applied to a coastal British Columbia watershed, the Sarita River watershed which is located on the west coast of the Vancouver Island. Estimated the peak instantaneous and daily flows are compared with observed flows. The analysis shows that the stochastic-deterministic procedure, the B.C. Environment method, and the Bayesian method give acceptable results. In particular, the stochastic-deterministic procedure requires very limited infor...