Generalized Linear Model Models Research Articles

Abstract 13566: Predictors of Progressive QRS Duration in Left Bundle Branch Block

Introduction: QRS duration in patients with left bundle branch block (LBBB) predicts adverse clinical outcomes and response to cardiac resynchronization therapy (CRT). Changes in LBBB QRS duration over time have not been reported. Hypothesis: QRS duration in patients with LBBB changes over time and may be predicted by patient demographics and comorbidities. Methods: We retrospectively identified 450 patients with LBBB on an index ECG (QRS ≥120 msec) recorded between 1992 and 2000 and examined any follow-up ECGs recorded after approximately 1, 2, 3, 5, and 10 years. Piecewise linear interpolations were performed, and separate generalized linear models (GLM) with Schwartz Bayesian Criterion effect selection were used to relate baseline clinical variables to changes in QRS duration at follow-up. Results: The cohort was 53.1% male, average age 66.7 yr, and mean index ECG QRS duration 144.7 msec. Although there was significant variability, the duration of the mean QRS increased as a tapering exponential function of time (p<0.0001) with progressively less increase at later follow-up. Three additional variables strongly and independently predicted progressive lengthening of the QRS within the overall GLM: heart failure with reduced ejection fraction (HFrEF) (p<0.0001), index ECG QRS duration (neg) (p<0.0001), and male gender (p=0.0063). In a subgroup of patients with an index ECG QRS <150 msec, these four GLM model variables also predicted the increase in QRS: time (p=0.0129), index QRS duration (p=0.0042), HFrEF (p=0.0001), and male gender (p=0.0084). (Equation, Figure below) ΔQRS(t;i) = 50.22 - 9.71*e -0.29*t - 0.344*QRS(0;i) + 7.62* 1 HFrEF (i) +5.12* 1 Male (i) Adjusted R 2 = 0.103 Conclusions: QRS duration progressively prolongs with time in patients with LBBB especially in patients with relatively narrow QRS and can be predicted by the presence of HFrEF and male gender. These predictors may be useful in better informing clinical prognosis and decisions regarding CRT.

Oxidative stress markers-driven prognostic model to predict post-discharge mortality in heart failure with reduced ejection fraction.

Current predictive models based on biomarkers reflective of different pathways of heart failure with reduced ejection fraction (HFrEF) pathogenesis constitute a useful tool for predicting death risk among HFrEF patients. The purpose of the study was to develop a new predictive model for post-discharge mortality risk among HFrEF patients, based on a combination of clinical patients' characteristics, N-terminal pro-B-type Natriuretic peptide (NT-proBNP) and oxidative stress markers as a potentially valuable tool for routine clinical practice. 116 patients with stable HFrEF were recruited in a prospective single-center study. Plasma levels of NT-proBNP and oxidative stress markers [superoxide dismutase (SOD), glutathione peroxidase (GPX), uric acid (UA), total bilirubin (TB), gamma-glutamyl transferase (GGT) and total antioxidant capacity (TAC)] were measured in the stable predischarge condition. Generalized linear model (GLM), random forest and extreme gradient boosting models were developed to predict post-discharge mortality risk using clinical and laboratory data. Through comprehensive evaluation, the most performant model was selected. During a median follow-up of 525 days (7-930), 33 (28%) patients died. Among the three created models, the GLM presented the best performance for post-discharge death prediction in HFrEF. The predictors included in the GLM model were age, female sex, beta blockers, NT-proBNP, left ventricular ejection fraction (LVEF), TAC levels, admission systolic blood pressure (SBP), angiotensin-converting enzyme inhibitors/angiotensin receptor II blockers (ACEI/ARBs) and UA levels. Our model had a good discriminatory power for post-discharge mortality [The area under the curve (AUC) = 74.5%]. Based on the retained model, an online calculator was developed to allow the identification of patients with heightened post-discharge death risk. In conclusion, we created a new and simple tool that may allow the identification of patients at heightened post-discharge mortality risk and could assist the treatment decision-making.

Genome-Wide Association Study of Leaf Rust Resistance at Seedling and Adult Plant Stages in a Global Barley Panel

Barley leaf rust caused by Puccinia hordei (Ph) is one of the major limiting biotic stresses of barley production worldwide and causes yield losses of up to 60%. A diversity panel of 316 barley genotypes (AM2017) composed of released cultivars, advanced breeding lines and landraces was screened for Ph resistance at the seedling stage using two isolates (SRT-SAT and SRT-MRC), while the adult plant stage resistance screening was conducted at the disease hotspot location of Sidi Allal Tazi (SAT) for the cropping seasons of 2017 and 2019. The phenotypic responses were combined with 36,793 single nucleotide polymorphism (SNP) markers in a genome-wide association study (GWAS) using the general linear model (GLM), mixed linear model (MLM), settlement of MLM under progressively exclusive relationship (SUPER), multiple-locus MLM (MLMM), fixed and random model circulating probability unification (FarmCPU), and Bayesian-information and linkage-disequilibrium iteratively nested keyway (BLINK) in GAPIT3, and MLM (K+Q), MLM (K+PCA), and GLM (Q) models in TASSEL to identify genomic regions linked to Ph resistance. Fourteen barley genotypes were resistant (R) at the seedling stage to both Ph isolates, SRT-SAT and SRT-MRC, and twelve genotypes were either resistant (R) or moderately resistant (MR) at the adult plant stage, whereas only one genotype was resistant at the seedling stage, and moderately resistant at the adult plant stage. The genome scan revealed 58 significant marker trait associations (MTA) among which 34 were associated with seedling resistance (SR) and 24 with adult plant resistance (APR). Common genomic regions conferring resistance to Ph were identified at both stages on chromosome 2H (106.53 cM and at 107.37 cM), and on chromosome 7H (126.7 cM). Among the 58 MTA identified, 26 loci had been reported in previous studies, while the remaining 32 loci were regarded as novel. Furthermore, the functional annotation of candidate genes (CGs) adjacent to 36 SNP markers with proteins involved in disease resistance further confirms that some of the SNP markers from our study could be associated with Ph resistance in barley. The resistant barley genotypes and some of the SNP markers from this study with high R2 and additive effects can be converted into high-throughput functional markers for accelerated selection and pyramiding of leaf rust resistance genes in North African barley germplasm.

이항 회귀모형의 연결함수 오지정 문제에서 확률화된 분위수 잔차 사용에 대한 연구

Logistic regression model is usually used when the response is Bernoulli or binomial to predict the probability of event of interest. Specifically, the logistic model is a generalized linear model(GLM) where the assumed link function is the inverse CDF of the logistic distribution. The link misspecification often occurs when the true link is not the logistic link function. In addition, residuals in the binomial GLM model are of less practical use for diagnostics because the response is not continuous. Randomized quantile residuals are an alternative option because they are defined to follow the standard normal distribution. In this study, we investigated the usage of the randomized quantile residual to diagnose the link misspecification. When the logistic regression is fitted to the data where the true data generating process is irrelevant to the logistic link function, we consider the normality test on the randomized quantile residuals from the misspecified logistic regression model and explore whether the link misspecification can be detected or not via extensive simulation studies. We have found that the randomized quantile residual is far from the normality especially when the fitted logistic link function fails to approximate the true link function so that their gap is big.

Pemodelan Jumlah Penduduk Miskin di Jawa Timur dengan Generalized Linier Model

Poverty as a development issue is an important thing to pay attention. Therefore, it is necessary to analyze to prevent an increase in the number of poor people in an area, one of which is by considering the things that affect poverty estimates and knowing the model of the number of poor people that can later be used as predictions. East Java is the region in Indonesia with the highest population in Indonesia in 2019. One way to find out the influencing factors and at the same time to make a model of the number of poor people in East Java is by modeling with the Generalized Linear Model (GLM) method. General Linear Model (GLM). GLMs aim to determine the causal relationship and the influence of the independent variable on the dependent variable. The GLMs model in this study uses the Gaussian family because the research response variable (Y) is an integer. The results show that of the seven estimated models, only one model that meets the significance is the model with the dependent variable population density (X3). So it can be concluded that between Area (X1), Open Unemployment Rate (X2), and Population Density (X3), population density has a significant effect on the number of poor people with a 95% confidence level.

Development and comparative analysis of ANN and SVR-based models with conventional regression models for predicting spray drift.

As monitoring of spray drift during application can be expensive, time-consuming, and labor-intensive, drift predicting models may provide a practical complement. Several mechanistic models have been developed as drift prediction tool for various types of application equipment. Nevertheless, mechanistic models are quite often intricate and complex with a large number of input parameters required. Quite often, the detailed data needed for such models are not readily available. In this study, two advanced machine learning models (artificial neural network (ANN) and support vector regression (SVR)) were developed for pesticide drift prediction and compared with three conventional regression-based models: multiple linear regression (MLR), generalized linear model (GLM), and generalized nonlinear least squares (GNLS). The models were evaluated in fivefold cross-validation and by external validation using the coefficient of determination (R2), root mean square error (RMSE), mean absolute error (MAE), and mean absolute bias (MAB). From regression-based models, GLM and GNLS models performed very well when evaluated by cross-validation with R2 = 0.96 and 0.95 and RMSE = 0.70 and 0.82 respectively, while MLR performed less with R2 of 0.65 and RMSE of 2.25. Simultaneously, ANN and SVR models performed very well with R2 = 0.98 and 0.97 and RMSE = 0.58 and 0.71 respectively. Overall, ANN model performed best compared to the other four models followed by SVR. A comparison was also made between the high-performing model, ANN, and two previously published empirical models. The ANN model outperformed the two previously published empirical models and can be used to predict pesticide drift. Therefore, the ANN model is a potentially promising new approach for predicting ground drift that merits further study. In conclusion, our work demonstrated that the new approach, ANN and SVR-based models, for pesticide drift modeling has better predictive power than conventional regression models. Their ability to model complex relationships is a clear benefit in pesticide drift modeling where the variability in pesticide drift is often affected by a number of variables and the relationships between drift and predictors are very complicated. We believe such insights will pave better way for the application of machine learning towards spray drift modeling.

Systemic inflammation accelerates the adverse effects of air pollution on metabolic syndrome: Findings from the China health and Retirement Longitudinal Study (CHARLS)

Long-term exposure to air pollution and systemic inflammation are associated with increased prevalence of metabolic syndrome (MetS); however, their joint effects in Chinese middle-aged and older adults is unknown. In this cross-sectional study, 11,838 residents aged 45 years and older from the China Health and Retirement Longitudinal Study (CHARLS) Wave 3 in 2015 were included. MetS was diagnosed using the Joint Interim Societies’ definition. C-Reactive Protein (CRP) was assessed to reflect systemic inflammation. Individual exposure to air pollutants (particulate matter with a diameter ≤2.5 μm (PM2.5) or ≤ 10 μm (PM10), sulfur dioxide (SO2), nitrogen dioxide (NO2), ozone (O3), and carbon monoxide (CO)) was evaluated using satellite-based spatiotemporal models according to participant residence at county-level. Generalized linear models (GLMs) were applied to examine the association between air pollution and MetS, and the modification effects of CRP between air pollution and MetS were estimated using interaction terms of CRP and air pollutants in the GLM models. The prevalence of MetS was 32.37%. The adjusted odd ratio (OR) of MetS was 1.192 (95% confidence interval (CI): 1.116, 1.272), 1.177 (95% CI: 1.103, 1.255), 1.158 (95% CI: 1.072, 1.252), 1.303 (95% CI: 1.211,1.403), 1.107 (95% CI: 1.046, 1.171) and 1.156 (95% CI:1.083, 1.234), per inter-quartile range increase in PM2.5 (24.04 μg/m3), PM10 (39.00 μg/m3), SO2 (19.05 μg/m3), NO2 (11.28 μg/m3), O3 (9.51 μg/m3) and CO (0.46 mg/m3), respectively. CRP was also associated with increased prevalence of MetS (OR = 1.049, 95% CI: 1.035, 1.064; per 1.90 mg/L increase in CRP). Interaction analysis suggested that high CRP levels enhanced the association between air pollution exposure and MetS. Long-term exposure to air pollution is associated with increased prevalence of MetS, which might be enhanced by systemic inflammation. Given the rapidly aging society and heavy burden of MetS, measures should be taken to improve air quality and reduce systemic inflammation.

Exposure to multiple trace elements and thyroid cancer risk in Chinese adults: A case-control study.

The incidence of thyroid cancer (TC) has increased rapidly in last decades. Multiple trace elements in the external environment have important effects with thyroid function. However, the evidence for these on TC risk were rarely reported. A total of 585 newly diagnosed TC patients and 585 healthy controls were included in this study, and 14 urinary elements were measured to explain the fixed-exposure effect on TC risk. Conditional logistic regressions were used to reflect the multi-element associations, and Bayesian kernel machine regression (BKMR) was applied to show the tendency of mixed effects. Furthermore, the interaction effects were examined by Generalized linear model (GLM). The levels of lithium (Li), cobalt (Co), strontium (Sr), zinc (Zn) and copper (Cu) had negative effects with TC risk, nevertheless lead (Pb), arsenic (As) and chromide (Cr) showed positive effects. The BKMR and GLM models reflected the effect fluctuations of different elements, and there was a slight interaction effects between Li and Cr, Co, Zn and Pb. Further study is required to confirm these results in the future.

The Determinants of the Environmental Performance of EU Financial Institutions: An Empirical Study with a GLM Model

Within the last few decades, the issue of the environmental performance of European financial institutions has become a significant feature of their strategic plans. Financial institutions can contribute through their own activities and investments, and also through their relationship with economic sectors and consumers, in decreasing environmental footprint. The purpose of this research is to investigate the determinants that affect the environmental performance of European financial institutions. Financial markets have been selected as the main research field for this study, as it presents an opportunity for environmental policy and is useful in view of the need for a wider range of policy instruments. Moreover, on a more practical level, financial institutions can interact with the environment in several ways, such as investors, innovators, valuers, powerful stakeholders, and polluters. The study is based on a mixed methodology approach, which integrates: (i) bibliometric analysis based on R package and (ii) panel data analysis with the assistance of a generalized linear model (GLM). Findings show that socioeconomic, governance, and technology factors positively affect the environmental performance of European financial institutions. Moreover, the incorporation of alternative energy sources, such as renewable energy in the corporate function, is a requirement for greening the financial institutions. The above can guide financial institutions to develop the appropriate strategies for decreasing their environmental footprint, improving their operational efficiency, and becoming more attractive and competitive in the market.

MODELLING INDIAN OCEAN AIR TEMPERATURE USING ADDITIVE MODEL

In this study, we used the fluctuating air temperature dataset. The change is caused by data fluctuations, trend, seasonality, cyclicity and irregularities. The generalized additive model (GAM) data approach is used to describe these phenomena. The aim of this research is to find out the factors that affect the air temperature in the Indian Ocean, find a suitable model, and obtain the best model from three approximate methods, namely the Linear Model (LM), the Generalized Linear Model (GLM), and the GAM models, which use a dataset of factors that affect the temperature of the Indian Ocean (close to Aceh region). For the air temperature of α = 0.05, the significant effects are precipitation, relative humidity, sea surface temperature, and the wind speed. The LM, GLM and GAM models are quite feasible because they all meet and pass the classical hypothesis tests, namely the normality test, multicollinearity test, the heteroscedasticity test, and the autocorrelation test. The appropriate model is GAM model based on adaptive smoothers. Compared to the LM, GLM and GAM models, GAM model with the adaptive smoothers base gave smallest AIC values of 4552.890 and 2392.396 where modeling was without and with time variable respectively. Therefore, it can be said that the correct model used at air temperature is the GAM model for adaptive smoothers base.

Variability in the Spatiotemporal Distribution Patterns of Greater Amberjack in Response to Environmental Factors in the Taiwan Strait Using Remote Sensing Data

The environmental characteristics of the Taiwan Strait (TS) have been linked to variations in the abundance and distribution of greater amberjack (Seriola dumerili) populations. Greater amberjack is a commercially and ecologically valuable species in ecosystems, and its spatial distribution patterns are pivotal to fisheries management and conservation. However, the relationship between the catch rates of S. dumerili and the environmental changes and their impact on fish communities remains undetermined in the TS. The goal of this study was to determine the spatiotemporal distribution pattern of S. dumerili with environmental characteristics in the TS from south to north (20°N–29°N and 115°E–127°E), applying generalized additive models (GAMs) and spatiotemporal fisheries data from logbooks and voyage data recorders from Taiwanese fishing vessels (2014–2017) as well as satellite-derived remote sensing environmental data. We used the generalized linear model (GLM) and GAM to analyze the effect of environmental factors and catch rates. The predictive performance of the two statistical models was quantitatively assessed by using the root mean square difference. Results reveal that the GAM outperforms the GLM model in terms of the functional relationship of the GAM for generating a reliable predictive tool. The model selection process was based on the significance of model terms, increase in deviance explained, decrease in residual factor, and reduction in Akaike’s information criterion. We then developed a species distribution model based on the best GAMs. The deviance explained indicated that sea surface temperature, linked to high catch rates, was the key factor influencing S. dumerili distributions, whereas mixed layer depth was the least relevant factor. The model predicted a relatively high S. dumerili catch rate in the northwestern region of the TS in summer, with the area extending to the East China Sea. The target species is strongly influenced by biophysical environmental conditions, and potential fishing areas are located throughout the waters of the TS. The findings of this study showed how S. dumerili populations respond to environmental variables and predict species distributions. Data on the habitat preferences and distribution patterns of S. dumerili are essential for understanding the environmental conditions of the TS, which can inform future priorities for conservation planning and management.

Thyroid-Stimulating Hormone and Estimated Glomerular Filtration Rate

Background: Hypothyroidism has been identified as a comorbidity related to chronic kidney disease (CKD). The retrospective study investigated thyroid function and CKD, and assessed the relationship between thyroid-stimulating hormone (TSH) and urine albumin/creatinine ratio (ACR), and the slope of estimated glomerular filtration rate (eGFR), stratified by CKD grades. Methods: This retrospective cohort study was conducted in a community nephrology clinic established with clinical and demographic data, from April 1, 2015 until December 30, 2019. Hypothyroidism prevalence, eGFR slope and ACR were the outcomes of interest and were analyzed by using unconditional and adjusted generalized linear model (GLM) and logistic regression model. Results: Of the 312 subjects, 58.3% were male, 12.8% had hypothyroidism, and 43.3% had diabetes mellitus, with the median age of 73 years (interquartile range (IQR): 29 - 99). The hypothyroidism prevalence was 9.4%, 11.5%, 15%, and 17.5% for the CKD categories defined as grade 1 and 2 combined, grade 3, grade 4, and grade 5, respectively. The overall median eGFR slope was -0.0027 (IQR: -0.158 - 0.602). With GLM models, the adjusted odds ratio of 1.052 (95% confidence interval (CI): 1.006 - 1.100) was calculated for TSH level > 5 µIU/L (Q2), per unit mL/min/day decline in eGFR slope. The overall median urine ACR was 10.2 mg/mmol (IQR: 0.24 - 1,414). In a GLM model with urine ACR per unit mg/mmol, the adjusted odds ratio of TSH level of > 1.8 µIU/L (50th percentile) was 1.02 (95% CI: 1.01 - 1.23). Conclusions: The prevalence of hypothyroidism increased with worsening eGFR grades from 9.4% to 17.4% at baseline. The higher TSH levels were associated with faster decline in eGFR and higher levels of albuminuria. Furthermore, prospective studies are needed to evaluate the effect of hypothyroidism treated on renal function. J Endocrinol Metab. 2022;12(3):89-96 doi: https://doi.org/10.14740/jem817

Nodule vascularity as novel radiomics imaging endpoint for lung cancer diagnosis and prognosis.

e20580 Background: The vascularization of lung nodules has been proven as severe risk factor for malignancy, and in lung cancer, indication of worse prognosis (1,2). For this reason, we developed a novel imagining endpoint based on the vasculature surrounding a lung mass and we tested this endpoint for the prediction of malignancy for lung nodules. Methods: The vasculature of the nodules (both arteries and veins) has been computed using the surface intersection between the nodule and the vascular structure 3D meshes. Both 3D structures were obtained by converting the segmentations of the nodule and of the vessels to meshes with a marching cubes algorithm. Nodule and vessels segmentation has been obtained with an in-house deep learning segmentation model. The features considered are the numbers of intersections, the total area of intersection and the mean area of intersection. These features have been used to predict nodule malignancy on thoracic CT scans from the Lung Image Database Consortium image collection (3). Quality controls on clinical data completeness and imaging parameters resulted in a cohort of 894 scans (715 for training and 179 for testing), from the original 1018 cases. The malignancy status is defined as high risk and low risk, based on the consensus classification of a panel of four radiologists. Firstly, an univariate analysis is performed to assess the variability of the features grouped by the malignancy score by using Mann-Whitney and ANOVA tests. After, seven combinations of features have been used to train generalized linear models (GLM) to predict nodule malignancy. To compare the models, the Area Under the Curve (AUC) is used as the main performance metric. Results: Univariate analysis of each feature grouped by the malignancy outcome showed that all the three features have good univariate discriminative power between high risk and low risk categories ( p value ≤ 0.05), with nb_connections as the most predictive singular feature ( p value of 1.343277 × 10-36). All the GLM models developed showed a good performance (AUC equal or higher than 0.7), with the best model in testing based on the combination of mean_area and sum_area (AUC of 0.84). Conclusions: The radiomics vascularity endpoint has been proven capable of predicting nodule malignancy with very good performance. The singular feature that is most related to malignancy is the number of vessels intersecting the nodule while the total area of intersection followed by the number of intersections are the most useful to model risk of malignancy. Wang et al., Lung Cancer 114: 38–43, 2017. Hamanaka et al., Diagn Pathol 10,17, 2015. G. Armato et al., Med. Phys., 38: 915-931, 2011.

Impact of model choice in predicting urban forest storm damage when data is uncertain

Research that illuminates causes of urban forest storm damage is valuable for planning and management. However, logistical and safety concerns often delay post-storm surveys in urban areas; thus, surveys may include observations with unverified sources of damage. While this uncertainty is often ignored, it can make up a high proportion of the number of damaged trees. The goal of this research was to improve understanding of techniques for modeling storm damage in urban forests. Using urban forest storm damage inventories collected in Florida, post-Hurricane Irma (2017), we tested how different imputation methods, modeling procedures, and damage frequency levels could impact overall model results. We utilized machine learning algorithms Random Forest (RF) and k-Nearest Neighbors (KNN), and generalized linear models (GLM). We found that GLM and RF models gave overall unbiased predictions of damage across all methods and rarity levels, while KNN consistently under-predicted damage. Damage frequency influenced some measures of performance but did not impact variable significance. Imputation methods all identified consistent variables of most significance within each model procedure; however, there was variation among variables ranked moderately important. While both GLM and RF models identified plot tree basal area as highly significant damage predictors, they otherwise disagreed on individual variable importance. These findings suggest that while explanatory models for urban forest storm damage can be achieved by examining linear relationships, more complex relationships such as the ones identified by RF models can have equal explanatory power with different subsets of predictor variables.

Mortality due to COVID-19 in Spain and its association with environmental factors and determinants of health

BackgroundThe objective of this study was to identify which air pollutants, atmospheric variables and health determinants could influence COVID-19 mortality in Spain. This study used information from 41 of the 52 provinces in Spain (from Feb. 1, to May 31, 2021). Generalized Linear Models (GLM) with Poisson link were carried out for the provinces, using the Rate of Mortality due to COVID-19 (CM) per 1,000,000 inhabitants as dependent variables, and average daily concentrations of PM10 and NO2 as independent variables. Meteorological variables included maximum daily temperature (Tmax) and average daily absolute humidity (HA). The GLM model controlled for trend, seasonalities and the autoregressive character of the series. Days with lags were established. The relative risk (RR) was calculated by increases of 10 g/m3 in PM10 and NO2 and by 1 ℃ in the case of Tmax and 1 g/m3 in the case of HA. Later, a linear regression was carried out that included the social determinants of health.ResultsStatistically significant associations were found between PM10, NO2 and the CM. These associations had a positive value. In the case of temperature and humidity, the associations had a negative value. PM10 being the variable that showed greater association, with the CM followed of NO2 in the majority of provinces. Anyone of the health determinants considered, could explain the differential geographic behavior.ConclusionsThe role of PM10 is worth highlighting, as the chemical air pollutant for which there was a greater number of provinces in which it was associated with CM. The role of the meteorological variables—temperature and HA—was much less compared to that of the air pollutants. None of the social determinants we proposed could explain the heterogeneous geographical distribution identified in this study.

Spectral technology and multispectral imaging for estimating the photosynthetic pigments and SPAD of the Chinese cabbage based on machine learning

• The estimation models constructed by spectra and multispectral imaging combined with various machine learning algorithms were compared. • A one-dimensional convolutional neural network model based on spectral data was established. • A two-dimensional convolutional neural network model based on multispectral images was established. • Spectra and multispectral images were combined with deep learning to achieve the estimation of multiple indicators, such as photosynthetic pigments and SPAD. The contents of photosynthetic pigment, which directly affect the growth of crops, could be evaluated with spectral and multispectral imaging technologies in an accurate and rapid way. For commodity germplasm resources on the market, the estimation of photosynthetic pigments and soil and plant analyzer development (SPAD) value was accomplished using the two techniques combined with machine learning. The spectrometer used in this study employed 781 bands from 320 nm to 1100 nm, and a multispectral imaging camera was used to acquire images in visible and near-infrared. Convolutional neural network (CNN), multiple linear regression (MLR) and generalized linear model (GLM) were used to establish the machine learning models, which established by preprocessed spectral data or 4-channel multispectral images. For estimating photosynthetic pigments (chlorophyll a , chlorophyll b , total chlorophyll and carotenoids), the GLM model established by spectral data was the optimal among all the models. For the SPAD optimal estimation model, the GLM model established by the spectral data and CNN model established by the multispectral images were fair. The R 2 and RMSE of the CNN model validation set in estimating SPAD were 0.87 and 2.31, respectively. The R 2 and RMSE of the GLM model validation set in estimating SPAD were 0.88 and 2.39, respectively. By combining two techniques with different machine learning methods, a comprehensive analysis of photosynthetic pigments and SPAD was accomplished in this paper.

Local tuning of radiomics-based model for predicting pathological response to neoadjuvant chemoradiotherapy in locally advanced rectal cancer

PurposeThis study aims to further enhance a validated radiomics-based model for predicting pathologic complete response (pCR) after chemo‑radiotherapy in locally advanced rectal cancer (LARC) for use in clinical practice.MethodsA generalized linear model (GLM) to predict pCR in LARC patients previously trained in Europe and validated with an external inter-continental cohort (59 patients), was first examined with further 88 intercontinental patient datasets to assess its reproducibility; then new radiomics and clinical features, and validation methods were investigated to build a new model for enhancing the pCR prediction for patients admitted to our department. The patients were divided into training group (75%) and validation group (25%) according to their demographic. The least absolute shrinkage and selection operator (LASSO) logistic regression was used to reduce the dimensionality of the extracted features of the training group and select the optimal ones; the performance of the reference GLM and enhanced models was compared through the area under curve (AUC) of the receiver operating characteristics.ResultsThe value of AUC of the reference model was 0.831 (95% CI, 0.701–0.961), and 0.828 (95% CI, 0.700–0.956) in the original and new validation cohorts, respectively, showing a reproducibility in the applicability of the GLM model. Eight features were found to be significant with LASSO and used to establish an enhanced model. The AUC of the enhanced model of 0.926 (95% CI, 0.859–0.993) for training, and 0.926 (95% CI, 0.767–1.00) for the validation group shows better performance than the reference model.ConclusionsThe GLM model shows good reproducibility in predicting pCR in LARC; the enhanced model has the potential to improve prediction accuracy and may be a candidate in clinical practice.

Reference database of total retinal vessel surface area derived from volume-rendered optical coherence tomography angiography

Optical coherence tomography angiography (OCTA) enables three-dimensional, high-resolution, depth-resolved flow to be distinguished from non-vessel tissue signals in the retina. Thus, it enables the quantification of the 3D surface area of the retinal vessel signal. Despite the widespread use of OCTA, no representative spatially rendered reference vessel surface area data are published. In this study, the OCTA vessel surface areas in 203 eyes of 107 healthy participants were measured in the 3D domain. A Generalized Linear Model (GLM) model analysis was performed to investigate the effects of sex, age, spherical equivalent, axial length, and visual acuity on the OCTA vessel surface area. The mean overall vessel surface area was 54.53 mm2 (range from 27.03 to 88.7 mm2). OCTA vessel surface area was slightly negatively correlated with age. However, the GLM model analysis identified axial length as having the strongest effect on OCTA vessel surface area. No significant correlations were found for sex or between left and right eyes. This is the first study to characterize three-dimensional vascular parameters in a population based on OCTA with respect to the vessel surface area.

Equalization Reserves for Reinsurance and Non-Life Undertakings in Switzerland

Equalization reserves is an insurance liability with features of own capital. By law, Swiss reinsurance and non-life undertakings must hold equalization reserves within their statutory accounts. Regarding Swiss solvency modeling, the equalization reserves are set to zero. Swiss reinsurance and non-life undertakings define the upper limit and the corresponding transfer rule to the equalization reserves; however, this information is not disclosed. The goal of the study is to find a relationship between the equalization reserves and the publicly available technical account items, applying a generalized additive model (GAM). Thereafter, we transform the continuous variables into discrete ones, and we apply a generalized linear model (GLM). The study is based on published data from 1997 to 2018, whereby we restate the implicitly published equalization reserves. For reinsurance undertakings, the GAM model captures the relationship better than the GLM one; for non-life undertakings, the GLM model performs better. For reinsurance undertakings, the equalization reserves depend on the equalization reserves of the previous year, on the calendar year, on the legal form, on the technical result, on the administration and commission costs and on other costs. For non-life undertakings, the equalization reserves depend on the net claims payments, on the equalization reserves of the previous year, on the net change in claims reserves without change in equalization reserves, on the calendar year and on the net earned premium. Furthermore, we look at the need for equalization reserves: do the undertakings accumulate and release the equalization reserves? Further, the impact of taxes on the equalization reserves is looked at. The concept of equalization reserves avoids the misuse of tax optimization. We conclude that the discussion about disclosure of equalization reserves will restart. In addition, the definition of the upper limit of the equalization reserves could be widened by linking the equalization reserves to the insurance/reserving risk from the capital modeling.

133 Quantitative Imaging Measurements Marginally Improve Risk Prediction for Children with Mild Traumatic Brain Injuries and Intracranial Injuries

INTRODUCTION: The KIIDS-TBI model was recently developed and externally validated in two large, multicenter populations to risk-stratify children with mild traumatic brain injuries (mTBI) and intracranial injuries. However, the KIIDS-TBI model only evaluates the presence/absence of imaging findings, rather than quantitative measures (e.g., hematoma size). METHODS: We included children <18 years who presented to one of five centers within 24 hours of TBI, had GCS scores of 13-15, and had intracranial injuries on neuroimaging. The dataset was split into training (75%) and test (25%) cohorts. We used generalized linear models (GLM) and recursive partitioning (RP) to predict the composite outcome of neurosurgery, intubation >24 hours due to TBI, or death due to TBI. Each model’s performance was compared to the KIIDS-TBI model across three decision-making risk cutoffs (<1%, <3%, and <5% predicted risk). RESULTS: There were 1,126 children included in the training and 374 in the test dataset. The RP model included epidural hematoma size, amount of skull fracture depression, amount of midline shift, and extra-axial hematoma size. The GLM model included these imaging variables and added clinical predictors, such as GCS Score. The GLM (76-90%) and RP (79-87%) models showed similar specificity across the three risk cutoffs in the test dataset, but the GLM model had higher sensitivity (89-96% for GLM; 89% for RP). By comparison, the KIIDS-TBI model had slightly higher sensitivity (93-100%) but lower specificity (27-82%) across the three risk cutoffs. CONCLUSION: Despite having obvious face value in many aspects of neurosurgical decision-making, quantitative imaging measures only marginally improved efforts to identify low risk patients in a large dataset of children with mTBI and intracranial injuries. Given the tradeoff between improved specificity but decreased sensitivity, these results do not support the addition of quantitative imaging measures to the KIIDS-TBI tool.