Statistical Learning Models Research Articles

Predictive Modelling of Postpartum Haemorrhage Using Early Risk Factors: A Comparative Analysis of Statistical and Machine Learning Models.

Postpartum haemorrhage (PPH) is a significant cause of maternal morbidity and mortality worldwide, particularly in low-resource settings. This study aimed to develop a predictive model for PPH using early risk factors and rank their importance in terms of predictive ability. The dataset was obtained from an observational case-control study in northern Rwanda. Various statistical models and machine learning techniques were evaluated, including logistic regression, logistic regression with elastic-net regularisation, Random Forests, Extremely Randomised Trees, and gradient-boosted trees with XGBoost. The Random Forest model, with an average sensitivity of 80.7%, specificity of 71.3%, and a misclassification rate of 12.19%, outperformed the other models, demonstrating its potential as a reliable tool for predicting PPH. The important predictors identified in this study were haemoglobin level during labour and maternal age. However, there were differences in PPH risk factor importance in different data partitions, highlighting the need for further investigation. These findings contribute to understanding PPH risk factors, highlight the importance of considering different data partitions and implementing cross-validation in predictive modelling, and emphasise the value of identifying the appropriate prediction model for the application. Effective PPH prediction models are essential for improving maternal health outcomes on a global scale. This study provides valuable insights for healthcare providers to develop predictive models for PPH to identify high-risk women and implement targeted interventions.

An Enhanced Long-Term Wind Speed Prediction Using Dynamic Unified Ensemble Learning and Data Assimilation Techniques: A Case Study in Tamil Nadu, India

Reliable wind speed prediction is essential for effective grid management since wind energy is a key component of renewable energy generation. However, controlling wind speed just right is no easy feat since the wind flow changes constantly. This paper presents an innovative method for predicting wind speeds in the future. It uses state-of-the-art data assimilation methods and a dynamic unified ensemble learning model. For efficient wind energy planning and monitoring, WSP accuracy is crucial. Data from a single site limited the accuracy of WSPs in previous research. An improved accuracy of long-term wind speed predictions is achieved by the proposed model via the integration of data assimilation and ensemble learning. To increase forecast accuracy, the model uses sophisticated data assimilation methods like the Kalman filter to combine data from many sources. Specifically, the model employs the Stacked CNN + BiLSTM with Data Assimilator (SCBLSTM+DA) technique, which integrates Wind Speed (WS) data from adjacent areas with the CNN + BiLSTM-based Ensemble Learning Model (ELM) and the Four-Dimensional Variational and Ensemble Kalman Filter (4DVar/EnKF) Data Assimilation method. Using real-world wind speed data from nine meteorological stations in Tamil Nadu, India, we find that current prediction models, including both classical statistical and cutting-edge machine learning models, perform better. Further, unlike standalone models, the suggested model shows less susceptibility to changes in prediction time scales. Promising a solution to improve long-term wind speed predictions accuracy, this study has significant consequences for wind energy management and production.

Employing Graph Neural Networks for Predicting Electrode Average Voltages and Screening High-Voltage Sodium Cathode Materials.

For many years, humans have been relentlessly focused on enhancing battery longevity and boosting energy storage capacities. The performance and durability of a battery depend significantly on the material used for its electrodes. In this context, merging machine learning with density functional theory (DFT) calculations has emerged as a pivotal approach to advancing the exploration of battery crystal structures. We present a new method that combines a graph convolutional neural network (GNN) with a Transformer convolutional layer, which we call Transformer-GNN. To underscore its efficacy, we benchmarked Transformer-GNN against three established statistical machine learning models: Support Vector Machine, Random Forest, and XGBoost. We also developed a standard GNN, which we refer to as Basic-GNN. Additionally, we compared Basic-GNN with Transformer-GNN to highlight the improvements brought about by incorporating the Transformer convolutional layer. The Transformer-GNN model outperforms the other models, achieving the highest R2 value of 0.82 and the lowest mean squared error of 0.3161. Our findings demonstrate that the Transformer-GNN can profoundly understand battery crystal structures, thus forging the path toward more sophisticated and durable battery systems. Leveraging the GNN model's voltage predictions in tandem with the capacity data sourced from the database, we screened and pinpointed Na(NiO2)2 as a high-voltage (higher than 5 V), high-capacity sodium cathode material. We conducted DFT calculations on Na(NiO2)2 and revealed the migration mechanism of the Na ions.

Abstract PO5-07-01: A novel clinical risk score that can accurately predict the survival of young breast cancer patients: A UAE-based cohort study

Abstract Introduction: Breast cancer (BC) in women below the age of 40 accounts for approximately 7% of all BC cases. According to studies, BC in younger women is more likely to have negative tumor characteristics and outcomes compared to older women. The rate of early-onset breast cancer in the United Arab Emirates (UAE) is higher than in the Western population (24%, previously reported). Young breast cancer patients in the UAE have worse adverse features, including a higher grade, a larger tumor size, and lymph node involvement, as we reported previously. Therefore, combining several clinical indicators to predict recurrence risk in our unique, heterogeneous young breast cancer group is needed. Aim:The aim of this study is to develop a statistical machine learning model that can predict survival and the risk of recurrence in a young breast cancer population seen and treated in a large-volume tertiary cancer center in the United Arab Emirates (Tawam Hospital). Methods: Early-onset breast cancer patients from a retrospective observational cohort study were included in this study (N = 904). A broad range of clinical data was collected. A multivariate ridge Cox regression model was performed using age, histology, molecular type (ER,PR,HER-2 status), T-stage, N-Stage, M-stage, grade, and BMI, and the interaction term age*BMI at the time of diagnosis. The primary outcome was time-to-death. A novel breast cancer risk score (BCRS) was developed as the linear predictor of the fitted-ridge Cox regression model. Breast cancer patients were categorized into low, medium, and high risk using the 33th and 66th centiles of the novel BCRS. A univariate Cox regression model was performed with a categorized BCRS as a predictor of time-to-death. Hazard ratios (HR) with a 95% confidence interval (95% CI) were estimated. Kaplan-Meier (KM) curves were plotted and compared using the log-rank test. Time-dependent sensitivity and specificity were computed at five- and ten-year follow-ups, respectively. Statistical analyses were performed in R version 4.2.3. Results: The median age was 36 years (IQR: 32–38), and 64% were Arabic. Fifteen percent had an event with a median duration of follow-up of 15.4 years (95% CI: 13.6–17.1). The estimated HR of the categorized novel BCRS in a univariate Cox regression model was 1.968 (95% CI: 1.075- 3.601) and 8.040 (95% CI: 4.695–13.769) for the intermediate and high-risk groups as compared to the low-risk group, respectively. KM curves show a clear separation in terms of survival between the three risk groups, p< 0.0001 (Figure 1). A cutoff value of -1.733 gives a sensitivity of 93.8 and 90.6 with corresponding specificities of 31.42 and 31.31 at five- and ten-year follow-up, respectively. Conclusion: This is the first study in the region that uses clinical data that combines several clinical characteristics and aims to design a statistical machine learning model to predict the risk of death from breast cancer. Our novel model can be used in daily clinical practice to identify high risk breast cancer patients and aid clinical decision-making in patients with breast cancer. However, further validation studies are needed. Figure 1:KM curves of breast-cancer survival for patients with low-risk (BCRS in [-2.32, -1.63], black curve), intermediate-risk (BCRS in [-1.64, -1.25], red curve), and high-risk BCRS in [-1.26, 1.41], green curve). Citation Format: Aydah Al-Awadhi, Mohammed Hourani, Mawada Hussein, Fatima Alkindi, Lina Wahba, Abla AlAgha, Alaa Shoqeir, Mouza AlShebli, Atlal Abusanad, Amar Ahmad. A novel clinical risk score that can accurately predict the survival of young breast cancer patients: A UAE-based cohort study [abstract]. In: Proceedings of the 2023 San Antonio Breast Cancer Symposium; 2023 Dec 5-9; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2024;84(9 Suppl):Abstract nr PO5-07-01.

Comparison of methods to aggregate climate data to predict crop yield: an application to soybean

High-dimensional climate data collected on a daily, monthly, or seasonal time step are now commonly used to predict crop yields worldwide with standard statistical models or machine learning models. Since the use of all available individual climate variables generally leads to calculation problems, over-fitting, and over-parameterization, it is necessary to aggregate the climate data used as predictors. However, there is no consensus on the best way to perform this task, and little is known about the impacts of the type of aggregation method used and of the temporal resolution of weather data on model performances. Based on historical data from 1981 to 2016 of soybean yield and climate on 3447 sites worldwide, this study compares different temporal resolutions (daily, monthly, or seasonal) and dimension reduction techniques (principal component analysis (PCA), partial least square regression, and their functional counterparts) to aggregate climate data used as inputs of machine learning and linear regression (LR) models predicting yields. Results showed that random forest models outperformed and were less sensitive to climate aggregation methods than LRs when predicting soybean yields. With our models, the use of daily climate data did not improve predictive performance compared to monthly data. Models based on PCA or averages of monthly data showed better predictive performance compared to those relying on more sophisticated dimension reduction techniques. By highlighting the high sensitivity of projected impact of climate on crop yields to the temporal resolution and aggregation of climate input data, this study reveals that model performances can be improved by choosing the most appropriate time resolution and aggregation techniques. Practical recommendations are formulated in this article based on our results.

Statistical parameterized physics-based machine learning digital shadow models for laser powder bed fusion process

This paper presents a statistical physics-based machine learning model for predicting defects, such as surface roughness and lack-of-fusion porosity, in the laser powder bed fusion of metals (PBF-LB/M) additive manufacturing process. The statistical physics-based model is calibrated and validated against controlled single-track experiments and used for statistical prediction for multi-layer and multi-track cases for PBF-LB/M defects. A mechanistic reduced-order-based stochastic calibration process is introduced to capture the stochastic nature of the melt pool. The calibrated physics-based digital shadow model is demonstrated for predicting the surface roughness of the National Institute of Standards and Technology (NIST) overhang part X4, with a difference of 9.3% compared to the experimental results. By leveraging data obtained from both the physics-based model and experiments, a machine learning model has been trained for fast predictions (inference time of 0.4 ms) with high accuracy (error bound of 6.7%). This model can predict melt pool geometries under various processing conditions, offering a control strategy for the PBF-LB/M process. Further, the trained machine learning model is showcased to demonstrate a control application of melt pool geometries (width and depth) for specific processing parameters. These developed models (physics-based and machine learning) serve as a digital shadow of the PBF-LB/M process, offering predictive capabilities to build a digital twin model for process control, optimization, and online monitoring.

Mobile network traffic analysis based on probability-informed machine learning approach

The paper proposes an approach to the joint use of statistical and machine learning (ML) models to solve the problems of the precise reconstruction of historical events, real-time detection of ongoing incidents, and the prediction of future quality of service-related occurrences for prospective development of the modern networks. For forecasting, a regression version of the deep Gaussian mixture model (DGMM) is introduced. First, the preliminary clustering based on the finite normal mixtures is performed. This information is then used as an input for some supervised ML algorithm. It is the basic concept of the probability-informed ML approach in the field of telecommunications networks. Using the real-world datasets from a Portuguese mobile operator as well as public cellular traffic data, the article compares this approach with methods such as random forests, support vector machine regression, gradient boosting and LSTM. Vector autoregression, informed by the parameters of the generalized gamma (GG) distribution, which has also been successfully used to reconstruct past traffic patterns, is also used as a benchmark. We demonstrate that DGMM-based regression is 6.82–22.8 times faster than LSTM for the dataset. Moreover, DGMM-based regression can achieve better results for the most important traffic characteristics (average and total traffic, the number of users). For metrics MAPE and RMSE, it surpasses the results of statistical methods up to 46.7% (RMSE) and 91.5% (MAPE) (median increases are 28.0% and 80.1%, respectively), as well as for ML methods up to 13.0% (RMSE) and 35.7% (MAPE) (median increases are 0.39% and 2.5%, respectively). Thus, the use of a probability-informed approach for telecommunication data seems optimal for the computational speed and accuracy trade-off. Also, we introduce a novel statistical hypothesis testing method based on GG distribution for detecting suspected anomalies in traffic.

Mutation Patterns Predict Drug Sensitivity in Acute Myeloid Leukemia.

The inherent genetic heterogeneity of acute myeloid leukemia (AML) has challenged the development of precise and effective therapies. The objective of this study was to elucidate the genomic basis of drug resistance or sensitivity, identify signatures for drug response prediction, and provide resources to the research community. We performed targeted sequencing, high-throughput drug screening, and single-cell genomic profiling on leukemia cell samples derived from patients with AML. Statistical approaches and machine learning models were applied to identify signatures for drug response prediction. We also integrated large public datasets to understand the co-occurring mutation patterns and further investigated the mutation profiles in the single cells. The features revealed in the co-occurring or mutual exclusivity pattern were further subjected to machine learning models. We detected genetic signatures associated with sensitivity or resistance to specific agents, and identified five co-occurring mutation groups. The application of single-cell genomic sequencing unveiled the co-occurrence of variants at the individual cell level, highlighting the presence of distinct subclones within patients with AML. Using the mutation pattern for drug response prediction demonstrates high accuracy in predicting sensitivity to some drug classes, such as MEK inhibitors for RAS-mutated leukemia. Our study highlights the importance of considering the gene mutation patterns for the prediction of drug response in AML. It provides a framework for categorizing patients with AML by mutations that enable drug sensitivity prediction.

Developing statistical and machine learning models for predicting CO2 solubility in live crude oils

CO2 solubility in crude oil primarily affects the viscosity and swelling capacity of oil and determines the displacement efficiency of CO2 flooding during the enhanced oil recovery process. Dissolution of CO2 into crude oil also ensures safe and permanent storage of CO2 in depleted formations via solubility trapping. This study employed statistical and machine learning algorithms to develop precise and stable predictive models for estimating CO2 solubility in crude oils with dissolved gas. Recognizing the limitations of existing experimental data on CO2 solubility for live oils (crude oils with dissolved gas), we introduced a novel approach employing statistical and machine learning techniques for developing accurate and robust predictive models. Our methodology combined regularization, k-fold cross-validation, and statistical significance tests to ensure the robustness of the models, practices not considered in the previous studies. A linear regression model with L2 regularization is developed as the base model and compared with a 2nd order polynomial model and an ensemble feed-forward neural network (FNN) model integrated with gradient boosting (GB). The incorporation of GB into the neural network capitalizes on the additive sequence of FNN models, thereby enhancing the accuracy and reliability of predictions. Moreover, an improved user-accessible and easily implementable correlation is proposed to estimate CO2 solubility in live crude oils at diverse operating conditions as a function of reservoir temperature, saturation pressure, bubble-point pressure, oil molecular weight, and specific gravity. The performances of the developed model and the correlation are compared against the literature models and correlations. The FNN model with an overall R2 score of 0.9978, average absolute relative deviation (AARD) of 1.54 %, and root mean squared error (RMSE) of 0.0083 outperformed the previously published models and correlations. The new correlation, while being relatively simpler than previously published correlations, showed similar predictive performance with an overall R2 score of 0.9817, AARD of 4.98 %, and RMSE of 0.0225. The model and the correlation can be used to estimate CO2 solubility in live crude oils with 0.67 to 0.97 specific gravity for reservoir temperatures and pressures in the range of 28 to 124 °C and 3.23 to 32.76 MPa, respectively.

Exploration and demonstration of explainable machine learning models in prosthetic rehabilitation-based gait analysis.

Quantitative gait analysis is important for understanding the non-typical walking patterns associated with mobility impairments. Conventional linear statistical methods and machine learning (ML) models are commonly used to assess gait performance and related changes in the gait parameters. Nonetheless, explainable machine learning provides an alternative technique for distinguishing the significant and influential gait changes stemming from a given intervention. The goal of this work was to demonstrate the use of explainable ML models in gait analysis for prosthetic rehabilitation in both population- and sample-based interpretability analyses. Models were developed to classify amputee gait with two types of prosthetic knee joints. Sagittal plane gait patterns of 21 individuals with unilateral transfemoral amputations were video-recorded and 19 spatiotemporal and kinematic gait parameters were extracted and included in the models. Four ML models-logistic regression, support vector machine, random forest, and LightGBM-were assessed and tested for accuracy and precision. The Shapley Additive exPlanations (SHAP) framework was applied to examine global and local interpretability. Random Forest yielded the highest classification accuracy (98.3%). The SHAP framework quantified the level of influence of each gait parameter in the models where knee flexion-related parameters were found the most influential factors in yielding the outcomes of the models. The sample-based explainable ML provided additional insights over the population-based analyses, including an understanding of the effect of the knee type on the walking style of a specific sample, and whether or not it agreed with global interpretations. It was concluded that explainable ML models can be powerful tools for the assessment of gait-related clinical interventions, revealing important parameters that may be overlooked using conventional statistical methods.

Targeted-BEHRT: Deep Learning for Observational Causal Inference on Longitudinal Electronic Health Records.

Observational causal inference is useful for decision-making in medicine when randomized clinical trials (RCTs) are infeasible or nongeneralizable. However, traditional approaches do not always deliver unconfounded causal conclusions in practice. The rise of "doubly robust" nonparametric tools coupled with the growth of deep learning for capturing rich representations of multimodal data offers a unique opportunity to develop and test such models for causal inference on comprehensive electronic health records (EHRs). In this article, we investigate causal modeling of an RCT-established causal association: the effect of classes of antihypertensive on incident cancer risk. We develop a transformer-based model, targeted bidirectional EHR transformer (T-BEHRT) coupled with doubly robust estimation to estimate average risk ratio (RR). We compare our model to benchmark statistical and deep learning models for causal inference in multiple experiments on semi-synthetic derivations of our dataset with various types and intensities of confounding. In order to further test the reliability of our approach, we test our model on situations of limited data. We find that our model provides more accurate estimates of relative risk [least sum absolute error (SAE) from ground truth] compared with benchmark estimations. Finally, our model provides an estimate of class-wise antihypertensive effect on cancer risk that is consistent with results derived from RCTs.

Performance Differences of a Touch-Based Serial Reaction Time Task in Healthy Older Participants and Older Participants With Cognitive Impairment on a Tablet: Experimental Study.

Digital neuropsychological tools for diagnosing neurodegenerative diseases in the older population are becoming more relevant and widely adopted because of their diagnostic capabilities. In this context, explicit memory is mainly examined. The assessment of implicit memory occurs to a lesser extent. A common measure for this assessment is the serial reaction time task (SRTT). This study aims to develop and empirically test a digital tablet-based SRTT in older participants with cognitive impairment (CoI) and healthy control (HC) participants. On the basis of the parameters of response accuracy, reaction time, and learning curve, we measure implicit learning and compare the HC and CoI groups. A total of 45 individuals (n=27, 60% HCs and n=18, 40% participants with CoI-diagnosed by an interdisciplinary team) completed a tablet-based SRTT. They were presented with 4 blocks of stimuli in sequence and a fifth block that consisted of stimuli appearing in random order. Statistical and machine learning modeling approaches were used to investigate how healthy individuals and individuals with CoI differed in their task performance and implicit learning. Linear mixed-effects models showed that individuals with CoI had significantly higher error rates (b=-3.64, SE 0.86; z=-4.25; P<.001); higher reaction times (F1,41=22.32; P<.001); and lower implicit learning, measured via the response increase between sequence blocks and the random block (β=-0.34; SE 0.12; t=-2.81; P=.007). Furthermore, machine learning models based on these findings were able to reliably and accurately predict whether an individual was in the HC or CoI group, with an average prediction accuracy of 77.13% (95% CI 74.67%-81.33%). Our results showed that the HC and CoI groups differed substantially in their performance in the SRTT. This highlights the promising potential of implicit learning paradigms in the detection of CoI. The short testing paradigm based on these results is easy to use in clinical practice.

Machine learning models for classifying non-specific neck pain using craniocervical posture and movement

ObjectivePhysical therapists and clinicians commonly confirm craniocervical posture (CCP), cervical retraction, and craniocervical flexion as screening tests because they contribute to non-specific neck pain (NSNP). We compared the predictive performance of statistical machine learning (ML) models for classifying individuals with and without NSNP using datasets containing CCP and cervical kinematics during pro- and retraction (CKdPR). DesignExploratory, cross-sectional design. Setting and participantsIn total, 773 public service office workers (PSOWs) were screened for eligibility (NSNP, 441; without NSNP, 332). MethodsWe set up five datasets (CCP, cervical kinematics during the protraction, cervical kinematics during the retraction, CKdPR and combination of the CCP and CKdPR). Four ML algorithms–random forest, logistic regression, Extreme Gradient boosting, and support vector machine–were trained. Main outcome measuresModel performance were assessed using area under the curve (AUC), accuracy, precision, recall and F1-score. To interpret the predictions, we used Feature permutation importance and SHapley Additive explanation values. ResultsThe random forest model in the CKdPR dataset classified PSOWs with and without NSNP and achieved the best AUC among the five datasets using the test data (AUC, 0.892 [good]; F1, 0.832). The random forest model in the CCP dataset had the worst AUC among the five datasets using the test data [AUC, 0.738 (fair); F1, 0.715]. ConclusionML performance was higher for the CKdPR dataset than for the CCP dataset, suggesting that ML algorithms are more suitable than classical statistical methods for developing robust models for classifying PSOWs with and without NSNP.

FedAGAT: Real-time traffic flow prediction based on federated community and adaptive graph attention network

Predicting traffic flow is vital for optimizing intelligent transportation systems (ITS) and reducing congestion by forecasting traffic patterns accurately. However, current centralized TFP systems have limitations, e.g., slow training, high communication costs, and privacy concerns. To address these challenges, this study proposes FedAGAT, a system for short-term traffic flow prediction (TFP) based on federated community and adaptive spatial-temporal graph attention networks (AGAT). The FedAGAT system allows for local data processing and sharing only model updates with the server. This enables real-time, scalable, and secure TFP - essential capabilities for efficient ITS. AGAT is employed to capture intricate spatial-temporal dependencies in traffic flow, while federated learning facilitates decentralized learning, enhancing privacy. The FedAGAT prediction process involves four steps: dividing the local subnetwork using spectral community detection, locally training based on global parameters, uploading updated parameters, and creating a global model prediction based on the aggregated parameters. To evaluate the performance of FedAGAT, two real-world traffic datasets were utilized for benchmarking against seven statistical and deep learning models. Results demonstrate that FedAGAT provides relatively higher accuracy for short- and mid-term forecasting horizons. Moreover, FedAGAT predictions closely match real traffic flow values, and the overall performance is comparable to a global model, while requiring less time.

A Systematic Review on Digital Soil Mapping Approaches in Lowland Areas

Digital soil mapping (DSM) around the world is mostly conducted in areas with a certain relief characterized by significant heterogeneities in soil-forming factors. However, lowland areas (e.g., plains, low-relief areas), prevalently used for agricultural purposes, might also show a certain variability in soil characteristics. To assess the spatial distribution of soil properties and classes, accurate soil datasets are a prerequisite to facilitate the effective management of agricultural areas. This systematic review explores the DSM approaches in lowland areas by compiling and analysing published articles from 2008 to mid-2023. A total of 67 relevant articles were identified from Web of Science and Scopus. The study reveals a rising trend in publications, particularly in recent years, indicative of the growing recognition of DSM’s pivotal role in comprehending soil properties in lowland ecosystems. Noteworthy knowledge gaps are identified, emphasizing the need for nuanced exploration of specific environmental variables influencing soil heterogeneity. This review underscores the dominance of agricultural cropland as a focus, reflecting the intricate relationship between soil attributes and agricultural productivity in lowlands. Vegetation-related covariates, relief-related factors, and statistical machine learning models, with random forest at the forefront, emerge prominently. The study concludes by outlining future research directions, highlighting the urgency of understanding the intricacies of lowland soil mapping for improved land management, heightened agricultural productivity, and effective environmental conservation strategies.

Development of machine learning models for the prediction of erosion wear of hybrid composites

AbstractThis article reports on development of an adaptive framework for predicting the erosion performance of polymer composites using certain statistical and machine learning (ML) models. For this, ramie‐epoxy composites reinforced with variations (0–30 wt%) of sponge iron slag (an iron industry waste) are considered. The composites are fabricated and then subjected to high temperature solid particle erosion wear trials following Taguchi's L27 orthogonal array. The effects of different control factors on the erosion rate in an interactive environment are appraised by analysis of variance (ANOVA) which reveals the filler content as the most significant factor contributing 66.21%, followed by impact velocity (22.86%) and impingement angle (2.28%). A regression model based on the input–output parameters obtained from experimentation is constructed for prediction of erosion rate. Further, four predictive models using different machine learning algorithms are also proposed to predict the erosion rate of the composites. The feasibility and performance of each ML model is assessed using appropriate performance metrics. Among all the models, the gradient boosting machine model is found to be the most reliable model exhibiting the highest prediction accuracy and least errors.Highlights Development of novel class of composites reinforced with sponge iron slag. Database creation based on erosion wear experimentation on the composites. Data‐driven modeling for prediction of erosion rates using machine learning. Comparison of performance of different models and identifying the best one.

Reaction norm for genomic prediction of plant growth: modeling drought stress response in soybean

Key messageWe proposed models to predict the effects of genomic and environmental factors on daily soybean growth and applied them to soybean growth data obtained with unmanned aerial vehicles.Advances in high-throughput phenotyping technology have made it possible to obtain time-series plant growth data in field trials, enabling genotype-by-environment interaction (G × E) modeling of plant growth. Although the reaction norm is an effective method for quantitatively evaluating G × E and has been implemented in genomic prediction models, no reaction norm models have been applied to plant growth data. Here, we propose a novel reaction norm model for plant growth using spline and random forest models, in which daily growth is explained by environmental factors one day prior. The proposed model was applied to soybean canopy area and height to evaluate the influence of drought stress levels. Changes in the canopy area and height of 198 cultivars were measured by remote sensing using unmanned aerial vehicles. Multiple drought stress levels were set as treatments, and their time-series soil moisture was measured. The models were evaluated using three cross-validation schemes. Although accuracy of the proposed models did not surpass that of single-trait genomic prediction, the results suggest that our model can capture G × E, especially the latter growth period for the random forest model. Also, significant variations in the G × E of the canopy height during the early growth period were visualized using the spline model. This result indicates the effectiveness of the proposed models on plant growth data and the possibility of revealing G × E in various growth stages in plant breeding by applying statistical or machine learning models to time-series phenotype data.

An open science automatic workflow for multi-model species distribution estimation

AbstractIntegrated Environmental Assessment systems and ecosystem models study the links between anthropogenic and climatic pressures on marine ecosystems and help understand how to manage the effects of the unsustainable exploitation of ocean resources. However, these models have long implementation times, data and model interoperability issues and require heterogeneous competencies. Therefore, they would benefit from simplification, automatisation, and enhanced integrability of the underlying models. Artificial Intelligence can help overcome several limitations by speeding up the modelling of crucial functional parts, e.g. estimating the environmental conditions fostering a species’ persistence and proliferation in an area (the species’ ecological niche) and, consequently, its geographical distribution. This paper presents a full-automatic workflow to estimate species’ distributions through statistical and machine learning models. It embeds four ecological niche models with complementary approaches, i.e. Artificial Neural Networks, Maximum Entropy, Support Vector Machines, and AquaMaps. It automatically estimates the optimal model parametrisations and decision thresholds to distinguish between suitable- and unsuitable-habitat locations and combines the models within one ensemble model. Finally, it combines several ensemble models to produce a species richness map (biodiversity index). The software is open-source, Open Science compliant, and available as a Web Processing Service-standardised cloud computing service that enhances efficiency, integrability, cross-domain reusability, and experimental reproduction and repetition. We first assess workflow stability and sensitivity and then demonstrate effectiveness by producing a biodiversity index for the Mediterranean based on $$\sim $$ ∼ 1500 species data. Moreover, we predict the spread of the invasive Siganus rivulatus in the Mediterranean and its current and future overlap with the native Sarpa salpa under different climate change scenarios.

Radiographic Imaging for the Diagnosis and Treatment of Patients with Skeletal Class III Malocclusion.

Skeletal Class III malocclusion is one type of dentofacial deformity that significantly affects patients' facial aesthetics and oral health. The orthodontic treatment of skeletal Class III malocclusion presents challenges due to uncertainties surrounding mandibular growth patterns and treatment outcomes. In recent years, disease-specific radiographic features have garnered interest from researchers in various fields including orthodontics, for their exceptional performance in enhancing diagnostic precision and treatment effect predictability. The aim of this narrative review is to provide an overview of the valuable radiographic features in the diagnosis and management of skeletal Class III malocclusion. Based on the existing literature, a series of analyses on lateral cephalograms have been concluded to identify the significant variables related to facial type classification, growth prediction, and decision-making for tooth extractions and orthognathic surgery in patients with skeletal Class III malocclusion. Furthermore, we summarize the parameters regarding the inter-maxillary relationship, as well as different anatomical structures including the maxilla, mandible, craniofacial base, and soft tissues from conventional and machine learning statistical models. Several distinct radiographic features for Class III malocclusion have also been preliminarily observed using cone beam computed tomography (CBCT) and magnetic resonance imaging (MRI).

Knowledge-based quality assurance of a comprehensive set of organ at risk contours for head and neck radiotherapy.

Manual review of organ at risk (OAR) contours is crucial for creating safe radiotherapy plans but can be time-consuming and error prone. Statistical and deep learning models show the potential to automatically detect improper contours by identifying outliers using large sets of acceptable data (knowledge-based outlier detection) and may be able to assist human reviewers during review of OAR contours. This study developed an automated knowledge-based outlier detection method and assessed its ability to detect erroneous contours for all common head and neck (HN) OAR types used clinically at our institution. We utilized 490 accurate CT-based HN structure sets from unique patients, each with forty-two HN OAR contours when anatomically present. The structure sets were distributed as 80% for training, 10% for validation, and 10% for testing. In addition, 190 and 37 simulated contours containing errors were added to the validation and test sets, respectively. Single-contour features, including location, shape, orientation, volume, and CT number, were used to train three single-contour feature models (z-score, Mahalanobis distance [MD], and autoencoder [AE]). Additionally, a novel contour-to-contour relationship (CCR) model was trained using the minimum distance and volumetric overlap between pairs of OAR contours to quantify overlap and separation. Inferences from single-contour feature models were combined with the CCR model inferences and inferences evaluating the number of disconnected parts in a single contour and then compared. In the test dataset, before combination with the CCR model, the area under the curve values were 0.922/0.939/0.939 for the z-score, MD, and AE models respectively for all contours. After combination with CCR model inferences, the z-score, MD, and AE had sensitivities of 0.838/0.892/0.865, specificities of 0.922/0.907/0.887, and balanced accuracies (BA) of 0.880/0.900/0.876 respectively. In the validation dataset, with similar overall performance and no signs of overfitting, model performance for individual OAR types was assessed. The combined AE model demonstrated minimum, median, and maximum BAs of 0.729, 0.908, and 0.980 across OAR types. Our novel knowledge-based method combines models utilizing single-contour and CCR features to effectively detect erroneous OAR contours across a comprehensive set of 42 clinically used OAR types for HN radiotherapy.