Test Set Research Articles

Establishment and external validation of an early warning model of diabetic peripheral neuropathy based on random forest and logistic regression

ObjectiveThe primary objective of this study was to investigate the risk factors for diabetic peripheral neuropathy (DPN) and to establish an early diagnostic prediction model for its onset, based on clinical data and biochemical indices.MethodsRetrospective data were collected from 1,446 diabetic patients at the First Affiliated Hospital of Anhui University of Chinese Medicine and were split into training and internal validation sets in a 7:3 ratio. Additionally, 360 diabetic patients from the Second Affiliated Hospital were used as an external validation cohort. Feature selection was conducted within the training set, where univariate logistic regression identified variables with a p-value < 0.05, followed by backward elimination to construct the logistic regression model. Concurrently, the random forest algorithm was applied to the training set to identify the top 10 most important features, with hyperparameter optimization performed via grid search combined with cross-validation. Model performance was evaluated using ROC curves, decision curve analysis, and calibration curves. Model fit was assessed using the Hosmer-Lemeshow test, followed by Brier Score evaluation for the random forest model. Ten-fold cross-validation was employed for further validation, and SHAP analysis was conducted to enhance model interpretability.ResultsA nomogram model was developed using logistic regression with key features: limb numbness, limb pain, diabetic retinopathy, diabetic kidney disease, urinary protein, diastolic blood pressure, white blood cell count, HbA1c, and high-density lipoprotein cholesterol. The model achieved AUCs of 0.91, 0.88, and 0.88 for the training, validation, and test sets, respectively, with a mean AUC of 0.902 across 10-fold cross-validation. Hosmer-Lemeshow test results showed p-values of 0.595, 0.418, and 0.126 for the training, validation, and test sets, respectively. The random forest model demonstrated AUCs of 0.95, 0.88, and 0.88 for the training, validation, and test sets, respectively, with a mean AUC of 0.886 across 10-fold cross-validation. The Brier score indicates a good calibration level, with values of 0.104, 0.143, and 0.142 for the training, validation, and test sets, respectively.ConclusionThe developed nomogram exhibits promise as an effective tool for the diagnosis of diabetic peripheral neuropathy in clinical settings.

Prediction of pathological response and lymph node metastasis after neoadjuvant therapy in rectal cancer through tumor and mesorectal MRI radiomic features

Establishing predictive models for the pathological response and lymph node metastasis in locally advanced rectal cancer (LARC) treated with neoadjuvant chemoradiotherapy (nCRT) based on MRI radiomic features derived from the tumor and mesorectal compartment (MC). This study included 209 patients with LARC who underwent rectal MRI both before and after nCRT. The patients were divided into a training set (n = 146) and a test set (n = 63). Regions of interest (ROIs) for the tumor and MC were delineated on both pre- and post-nCRT MRI images. Radiomic features were extracted, and delta radiomic features were computed. The predictive endpoints were pathological complete response (pCR), pathological good response (pGR), and lymph node metastasis (LNM). Feature selection for various models involved sequentially removing features with a correlation coefficient > 0.9, and features with P-values ≥ 0.05 in univariate analysis, followed by LASSO regression on the remaining features. Logistic regression models were developed, and their performance was evaluated using the area under the receiver operating characteristic curve (AUC). Among the 209 LARC patients, the number of patients achieving pCR, pGR, and LNM were 44, 118, and 40, respectively. The optimal model for predicting each endpoint is the combined model that incorporates pre- and delta-radiomics features for both the tumor and MC. These models exhibited superior performance with AUC values of 0.874 (for pCR), 0.801 (for pGR), and 0.826 (for LNM), outperforming the MRI tumor regression grade (mrTRG) which yielded AUC values of 0.800, 0.715, and 0.603, respectively. The results demonstrate the potential utility of the tumor and MC radiomics features, in predicting treatment efficacy among LARC patients undergoing nCRT.

Quality control of elbow joint radiography using a YOLOv8-based artificial intelligence technology

BackgroundTo explore an artificial intelligence (AI) technology employing YOLOv8 for quality control (QC) on elbow joint radiographs.MethodsFrom January 2022 to August 2023, 2643 consecutive elbow radiographs were collected and randomly assigned to the training, validation, and test sets in a 6:2:2 ratio. We proposed the anteroposterior (AP) and lateral (LAT) models to identify target detection boxes and key points on elbow radiographs using YOLOv8. These identifications were transformed into five quality standards: (1) AP elbow positioning coordinates (XA and YA); (2) olecranon fossa positioning distance parameters (S17 and S27); (3) key points of joint space (Y3, Y4, Y5 and Y6); (4) LAT elbow positioning coordinates (X2 and Y2); and (5) flexion angle. Models were trained and validated using 2,120 radiographs. A test set of 523 radiographs was used for assessing the agreement between AI and physician and to evaluate clinical efficiency of models.ResultsThe AP and LAT models demonstrated high precision, recall, and mean average precision for identifying boxes and points. AI and physicians showed high intraclass correlation coefficient (ICC) in evaluating: AP coordinates XA (0.987) and YA (0.991); olecranon fossa parameters S17 (0.964) and S27 (0.951); key points Y3 (0.998), Y4 (0.997), Y5 (0.998) and Y6 (0.959); LAT coordinates X2 (0.994) and Y2 (0.986); and flexion angle (0.865). Compared to manual methods, using AI, QC time was reduced by 43% for AP images and 45% for LAT images (p < 0.001).ConclusionYOLOv8-based AI technology is feasible for QC of elbow radiography with high performance.Relevance statementThis study proposed and validated a YOLOv8-based AI model for automated quality control in elbow radiography, obtaining high efficiency in clinical settings.Key PointsQC of elbow joint radiography is important for detecting diseases.Models based on YOLOv8 are proposed and perform well in image QC.Models offer objective and efficient solutions for QC in elbow joint radiographs.Graphical

Robot error compensation strategy based on error sensitivity

Abstract Due to the errors in manufacturing and assembly, there are differences between the actual model and the theoretical model of the robot, which affects the positioning accuracy of the robot end-effector. In order to improve the accuracy of robot end-effector position, a robot error compensation strategy based on error sensitivity is proposed.Firstly, the robot kinematic model is established by Denavit-Hartenberg (D-H) method, and the sensitivity of end-effector position error is analyzed. According to the influence degree of different kinematic parameters on the robot end-effector position accuracy in the whole workspace, different weights are given to different kinematic parameters. {\color{red}Secondly, the kinematic error model is established, and the redundancy of the error parameter matrix is analyzed to obtain an independent error model. Thirdly, based on the error sensitivity analysis, a Weighted Levenberg-Marquard (WLM) algorithm with adaptive penalty factor is proposed, and the kinematic parameters are iteratively identified.} Finally, an error compensation experiment is carried out by using a universal serial six-degree-of-freedom robot. The experimental results show that the maximum error, mean absolute error and root mean square error of the position error on the test set are reduced by 90.75%, 89.86% and 95.64% respectively. The research in this paper provides a theoretical basis for robot end error compensation.

A Complex Background SAR Ship Target Detection Method Based on Fusion Tensor and Cross-Domain Adversarial Learning

Synthetic Aperture Radar (SAR) ship target detection has been extensively researched. However, most methods use the same dataset division for both training and validation. In practical applications, it is often necessary to quickly adapt to new loads, new modes, and new data to detect targets effectively. This presents a cross-domain detection problem that requires further study. This paper proposes a method for detecting SAR ships in complex backgrounds using fusion tensor and cross-domain adversarial learning. The method is designed to address the cross-domain detection problem of SAR ships with large differences between the training and test sets. Specifically, it can be used for the cross-domain detection task from the fully polarised medium-resolution ship dataset (source domain) to the high-resolution single-polarised dataset (target domain). This method proposes a channel fusion module (CFM) based on the YOLOV5s model. The CFM utilises the correlation between polarised channel images during training to enrich the feature information of single-polarised images extracted by the model during inference. This article proposes a module called the cross-domain adversarial learning module (CALM) to reduce overfitting and achieve adaptation between domains. Additionally, this paper introduces the anti-interference head (AIH) which decouples the detection head to reduce the conflict of classification and localisation problems. This improves the anti-interference and generalisation ability in complex backgrounds. This paper conducts cross-domain experiments using the constructed medium-resolution SAR full polarisation dataset (SFPD) as the source domain and the high-resolution single-polarised ship detection dataset (HRSID) as the target domain. Compared to the best-performing YOLOV8s model among typical mainstream models, this model improves precision by 4.9%, recall by 3.3%, AP by 2.4%, and F1 by 3.9%. This verifies the effectiveness of the method and provides a useful reference for improving cross-domain learning and model generalisation capability in the field of target detection.

Development of Machine Learning-Based Quantitative Structure-Activity Relationship Models for Predicting Plasma Half-Lives of Drugs in Six Common Food Animal Species

Abstract Plasma half-life is a crucial pharmacokinetic parameter for estimating extralabel withdrawal intervals of drugs to ensure the safety of food products derived from animals. This study focuses on developing a quantitative structure−activity relationship (QSAR) model incorporating multiple machine learning and artificial intelligence algorithms, and aims to predict the plasma half-lives of drugs in six food animals, including cattle, chickens, goats, sheep, swine, and turkeys. By integrating four machine learning algorithms with five molecular descriptor types, 20 QSAR models were developed using data from the Food Animal Residue Avoidance Databank (FARAD) Comparative Pharmacokinetic Database. The deep neural network (DNN) algorithm demonstrated the best prediction ability of plasma half-lives. The DNN model with all descriptors achieved superior performance with a high coefficient of determination (R 2) of 0.82±0.19 in 5-fold cross-validation on the training sets and a R 2 of 0.67 on the independent test set, indicating accurate predictions and good generalizability. The final model was converted to a user-friendly web dashboard to facilitate its wide application by the scientific community. This machine learning-based QSAR model serves as a valuable tool for predicting drug plasma half-lives and extralabel withdrawal intervals in six common food animals based on physicochemical properties. It also provides a foundation to develop more advanced models to predict the tissue half-life of drugs in food animals.

Machine learning elucidates electrophysiological properties predictive of multi- and single-firing human and mouse dorsal root ganglia neurons

Human and mouse dorsal root ganglia (hDRG and mDRG) neurons are important tools in understanding the molecular and electrophysiological mechanisms that underlie nociception and drive pain behaviors. One of the simplest differences in firing phenotypes is that neurons are single-firing (exhibit only one action potential) or multi-firing (exhibit 2 or more action potentials). To determine if single- and multi-firing hDRG neurons exhibit differences in intrinsic properties, firing phenotypes, and AP waveform properties, and if these properties could be used to predict multi-firing, we measured 22 electrophysiological properties by whole-cell patch-clamp electrophysiology of 94 hDRG neurons from 6 male and 4 female donors. We then analyzed the data using several machine learning models to determine if these properties could be used to predict multi-firing. We used 1000 iterations of Monte Carlo Cross Validation to split the data into different train and test sets and tested the Logistic Regression, k-Nearest Neighbors, Random Forest, Support Vector Classifier, and XGBoost machine learning models. All models tested had a greater than 80% accuracy on average, with Support Vector Classifier and XGBoost performing the best. We found that several properties correlated with multi-firing hDRG neurons and together could be used to predict multi-firing neurons in hDRG including a long decay time, a low rheobase, and long first spike latency. We also found that the hDRG models were able to predict multi-firing with 90% accuracy in mDRG neurons. Understanding these properties could be beneficial in the elucidation of targets on peripheral sensory neurons related to pain.Significance StatementThis study will improve understanding of the electrophysiological mechanisms of DRG neurons. Our machine learning algorithms show few species differences between mouse and human DRG neuron electrophysiology under baseline conditions. These are important findings for the study of neuronal excitability in the context of pain therapeutic development.

The dutch draw: constructing a universal baseline for binary classification problems

Abstract Novel prediction methods should always be compared to a baseline to determine their performance. Without this frame of reference, the performance score of a model is basically meaningless. What does it mean when a model achieves an $F_1$ of 0.8 on a test set? A proper baseline is, therefore, required to evaluate the ‘goodness’ of a performance score. Comparing results with the latest state-of-the-art model is usually insightful. However, being state-of-the-art is dynamic, as newer models are continuously developed. Contrary to an advanced model, it is also possible to use a simple dummy classifier. However, the latter model could be beaten too easily, making the comparison less valuable. Furthermore, most existing baselines are stochastic and need to be computed repeatedly to get a reliable expected performance, which could be computationally expensive. We present a universal baseline method for all binary classification models, named the Dutch Draw (DD). This approach weighs simple classifiers and determines the best classifier to use as a baseline. Theoretically, we derive the DD baseline for many commonly used evaluation measures and show that in most situations it reduces to (almost) always predicting either zero or one. Summarizing, the DD baseline is general, as it is applicable to any binary classification problem; simple, as it can be quickly determined without training or parameter tuning; and informative, as insightful conclusions can be drawn from the results. The DD baseline serves two purposes. First, it is a robust and universal baseline that enables comparisons across research papers. Second, it provides a sanity check during the prediction model’s development process. When a model does not outperform the DD baseline, it is a major warning sign.

Automatic geometric digital twin of box girder bridge using a laser-scanned point cloud

Geometric modeling is a pivotal step in creating a digital twin for existing bridge structures. Its deficiency of automation makes geometric modeling step time-consuming and laborious. This paper presents a solution for automatically modeling box girder bridges, including external and internal structures, based on laser-scanned point cloud. The solution includes three vital methods: component segmentation, key points extraction of cross-section, and internal structure reconstruction. The results indicate that the established segmentation model, BCR-Net, exhibited better performance than PointNet++ in component segmentation, as demonstrated by mIoU of 0.9751 on the test set. The mean absolute error on the dimension of the pier, bent cap and external and internal structure of the box girder is 0.27 %, 0.38 %, 0.47 %, and 0.48 %, respectively. It means the proposed methods possessed excellent modeling accuracy while ensuring high efficiency than manual modeling, providing a promising solution for digital twin modeling of bridge structures.

Comprehensive maps of material stock dynamics reveal increasingly coordinated urban development in the Yangtze River Delta of China

Sustainable urban development critically depends on effectively managing the interplay between material stock (MS) and economic growth. This study combined convolutional neural network model and nighttime lights data to map building MS of Yangtze River Delta (YRD) urban agglomeration in China from 2000 to 2020 across 1 km × 1 km pixel scale, then uncovered the spatiotemporal dynamics of MS and its correlation with economic development. Our findings indicate that the model performed robustly on the test set (R2 > 0.88). YRD's MS surged over tenfold, reaching 20,772 teragram, primarily expanding along northwest-southeast developmental axes. Most YRD cities exhibited synchronized growth in material stock and GDP, suggesting an emergent pattern of sustainable urban expansion. However, cities at the developmental extremes highlighted the need for optimizing urban development strategies. By categorizing YRD cities into four distinct development modes, our study offers deep insights into the dynamics of urban development, underpinning targeted strategies that could guide cities towards more sustainable and resource-efficient growth trajectories.

Forest Change Monitoring Based on Block Instance Sampling and Homomorphic Hypothesis Margin Evaluation

Forests play a crucial role in maintaining the integrity of natural ecosystems. Accurate mapping of windfall damages following storms is essential for effective post-disaster management. While remote sensing image classification offers substantial advantages over ground surveys for monitoring changes in forests, it encounters several challenges. Firstly, training samples in classification algorithms are typically selected through pixel-based random sampling or manual regional sampling. This approach struggles with accurately modeling complex patterns in high-resolution images and often results in redundant samples. Secondly, the limited availability of labeled samples compromises the classification accuracy when they are divided into training and test sets. To address these issues, two innovative approaches are proposed in this paper. The first is a new sample selection method which combines block-based sampling with spatial features extracted by single or multiple windows. Second, a new evaluation criterion is proposed by using the homomorphic hypothesis margin map with out-of-bag (OOB) accuracy. The former can not only assess the confidence level of each pixel category but also make regional boundaries clearer, and the latter can replace the test set so that all samples can be used for change detection. The experimental results show that the OOB accuracy obtained by spatial features with whole block sampling was 7.2% higher than that obtained by spectral features with pixel-based sampling and 2–3% higher than that for block center sampling, of which the highest value reached 98.8%. Additionally, the feasibility of identifying storm-damaged forests using only post-storm images has been demonstrated.

Identifying severe community-acquired pneumonia using radiomics and clinical data: a machine learning approach

Evaluating Community-Acquired Pneumonia (CAP) is crucial for determining appropriate treatment methods. In this study, we established a machine learning model using radiomics and clinical features to rapidly and accurately identify Severe Community-Acquired Pneumonia (SCAP). A total of 174 CAP patients were included in the study, with 64 cases classified as SCAP. Radiomic features were extracted from chest CT scans using radiomics techniques, and screened to remove irrelevant features. Additionally, clinical indicators of patients were similarly screened and constituted the clinical feature set. Subsequently, eight common machine learning models were employed to complete the SCAP identification task. Specifically, interpretability analysis was conducted on the models. In the end, we screened out 15 radiomic features (such as LeastAxisLength, Maximum2DDiameterColumn and ZonePercentage) and two clinical features: Lymphocyte (p = 0.041) and Albumin (p = 0.044). Using radiomic features as inputs in model predictions yielded the highest AUC of 0.85 on the test set. When using the clinical feature set as model inputs, the AUC was 0.82. Combining the two sets of features as model inputs, Ada Boost achieved the best performance with an AUC of 0.89. Our study demonstrates that combining radiomics and clinical data using machine learning methods can more accurately identify SCAP patients.

Preoperative prediction of lymph node metastasis in endometrial cancer patients via an intratumoral and peritumoral multiparameter MRI radiomics nomogram

IntroductionWhile lymph node metastasis (LNM) plays a critical role in determining treatment options for endometrial cancer (EC) patients, the existing preoperative methods for evaluating the lymph node state are not always satisfactory. This study aimed to develop and validate a nomogram based on intra- and peritumoral radiomics features and multiparameter magnetic resonance imaging (MRI) features to preoperatively predict LNM in EC patients.MethodsThree hundred and seventy-four women with histologically confirmed EC were divided into training (n = 220), test (n = 94), and independent validation (n = 60) cohorts. Radiomic features were extracted from intra- and peritumoral regions via axial T2-weighted imaging (T2WI) and apparent diffusion coefficient (ADC) mapping. A radiomics model (annotated as the Radscore) was established using the selected features from different regions. The clinical parameters were statistically analyzed. A nomogram was developed by combining the Radscore and the most predictive clinical parameters. Decision curve analysis (DCA) and the net reclassification index (NRI) were used to assess the clinical benefit of using the nomogram.ResultsNine radiomics features were ultimately selected from the intra- and peritumoral regions via ADC mapping and T2WI. The nomogram combining the Radscore, serum CA125 level, and tumor area ratio achieved the highest AUCs in the training, test and independent validation sets (nomogram vs. Radscore vs. clinical model: 0.878 vs. 0.850 vs. 0.674 (training), 0.877 vs. 0.838 vs. 0.668 (test), and 0.864 vs. 0.836 vs. 0.618 (independent validation)). The DCA and NRI results revealed the nomogram had greater diagnostic performance and net clinical benefits than the Radscore alone.ConclusionThe combined intra- and peritumoral region multiparameter MRI radiomics nomogram showed good diagnostic performance and could be used to preoperatively predict LNM in patients with EC.

Machine cross-domain remaining useful life prediction via contrastive adversarial variational recurrent method

The performance degradation process of the machine is non-stationary which varies with the operating environment and is reflected as a temporal shift phenomenon. Models or methods that assume the test and training sets have the same distribution will no longer be suitable for solving problems where domain shifts exist. This brings new challenges for accurately evaluating the machine’s remaining useful life (RUL). This paper studies a contrastive adversarial variational recurrent method for machine RUL prediction under different service conditions. In this new approach, the variational recurrent networks are developed to extract distribution features and latent spaces of raw data, meanwhile, the adversarial strategies are applied to enable the learning of domain-invariant features to make the model achieve cross-domain task processing. To supervise the learning process of the model to learn more mutual information between the extracted features and the raw input data, a contrastive loss is also introduced in the proposed method. Sufficient experiments were conducted to verify the feasibility and superiority of the suggested approach, including 12 sets of cross-scenario tests on the turbofan engine dataset C-MAPSS from NASA. Experimental findings confirm that the proposed method performs satisfactorily and competitively even compared to current state-of-the-art works.

Towards a fully automatic workflow for investigating the dynamics of lung cancer cachexia during radiotherapy using cone beam computed tomography

Abstract Objective. Cachexia is a devastating condition, characterized by involuntary loss of muscle mass with or without loss of adipose tissue mass. It affects more than half of patients with lung cancer, diminishing treatment effects and increasing mortality. Cone-beam computed tomography (CBCT) images, routinely acquired during radiotherapy treatment, might contain valuable anatomical information for monitoring body composition changes associated with cachexia. For this purpose, we propose an automatic artificial intelligence (AI)-based workflow, consisting of CBCT to CT conversion, followed by segmentation of pectoralis muscles. &amp;#xD;Approach. Data from 140 stage III non-small cell lung cancer patients was used. Two deep learning models, cycle-consistent generative adversarial network (CycleGAN) and contrastive unpaired translation (CUT), were used for unpaired training of CBCT to CT conversion, to generate synthetic CT (sCT) images. The no-new U-Net (nnU-Net) model was used for automatic pectoralis muscle segmentation. To evaluate tissue segmentation performance in the absence of ground truth labels, an uncertainty metric (UM) based on Monte Carlo dropout was developed and validated. &amp;#xD;Main results. Both CycleGAN and CUT restored the Hounsfield unit fidelity of the CBCT images compared to the planning CT (pCT) images and visually reduced streaking artefacts. The nnU-Net model achieved a Dice Similarity Coefficient (DSC) of 0.93, 0.94, 0.92 for the CT, sCT and CBCT images, respectively, on an independent test set. The UM showed a high correlation with DSC with a correlation coefficient of -0.84 for the pCT dataset and -0.89 for the sCT dataset. &amp;#xD;Significance. This paper shows a proof-of-concept for automatic AI-based monitoring of the pectoralis muscle area of lung cancer patients during radiotherapy treatment based on CBCT images, which provides an unprecedented time resolution of muscle mass loss during cachexia progression. Ultimately, the proposed workflow could provide valuable information for early intervention of cachexia, ideally resulting in improved cancer treatment outcome.

Web tension AI modeling and reconstruction for digital twin of roll-to-roll system

AbstractDigital twins (DT) are gaining attention as an emerging technology in Smart manufacturing systems. These DTs comprise various units that enable simulation, monitoring, and prediction of the manufacturing process. This study introduces a predictive model for web tension and a tension reconstruction algorithm for the DT of the roll-to-roll (R2R) system. The observed web tension signals from tension sensors decomposed into a mean component, a sinusoidal wave, and a random noise. Utilizing deep neural networks, the predictive model integrated various sub-models to forecast statistical (mean, standard deviation) and frequency domain (main frequency, signal-to-noise ratio) features of the web tension signal. Through fivefold cross-validation, 23 model architectures were optimized, with selected architectures ranging from 16-32-32-1 to 16-32-64-32-1 nodes per layer. Overall, R2 scores on the test set ranged from approximately 52 to 100%. The proposed reconstruction algorithm generated tension signals from the model’s predictions that closely resemble the original tension signals, indicating credible reconstructions. The proposed predictive model and reconstruction algorithm were integrated into the DT of the R2R system, offering a valuable tool for monitoring and optimizing the R2R process.

Deep-neural-network model for predicting ground motion parameters using earthquake horizontal-to-vertical spectral ratios

This study proposed a deep-neural-network (DNN) model for seismic ground motion prediction by utilizing a unified strong motion database by the National Research Institute for Earth Science and Disaster Resilience, and earthquake horizontal-to-vertical spectral ratio (EHVR) database in Japan. The model aims to enhance the accuracy of predictions by incorporating the EHVRs for complementing site effects, and utilizing existing ground motion prediction equations (GMPE) as the base model for source and propagation path effects. The hybrid approach enables the prediction of peak ground accelerations (PGAs), peak ground velocities (PGVs), and 5% damped absolute acceleration response spectra (SAs). After classifying the training and test sets from the database, the trained DNN models were applied on the test set to evaluate the performance of the predicted results. The accuracy assessment by the residuals, R-squared ( R2), and root mean square error (RMSE) between the predicted and observed values in the test set revealed the superior performance of the proposed model compared with the traditional GMPE with proxy-based site effects such as V S30s especially in predicting both the spectral amplitude and shape of SAs.

Construction and validation of a risk prediction model for extrauterine growth restriction in preterm infants born at gestational age less than 34 weeks

ObjectiveThis study aimed to develop and validate a model for predicting extrauterine growth restriction (EUGR) in preterm infants born ≤34 weeks gestation.MethodsPreterm infants from Guangxi Maternal and Child Health Hospital (2019–2021) were randomly divided into training (80%) and testing (20%) sets. Collinear clinical variables were excluded using Pearson correlation coefficients. Predictive factors were identified using Lasso regression. Random forest (RF), support vector machine (SVM), and logistic regression (LR) models were then built and evaluated using the confusion matrix, area under the curve (AUC), and the F1 score. Additionally, calibration curves and decision curve analysis (DCA) were plotted to assess the performance and practical utility of the models.ResultsThe study included 387 infants, with no significant baseline differences between training (n = 310) and testing (n = 77) sets. LR identified gestational age, birth weight, premature rupture of membranes, patent ductus arteriosus, cholestasis, and neonatal sepsis as key EUGR predictors. The RF model (19 variables) demonstrated an accuracy of greater than 90% during training, and superior AUC (0.62), F1 score (0.80), and accuracy (0.72) in testing compared to other models.ConclusionsGestational age, birth weight, premature rupture of membranes, patent ductus arteriosus, cholestasis, and neonatal sepsis are significant EUGR predictors in preterm infants ≤34 weeks. The model shows promise for early EUGR prediction in clinical practice, potentially enhancing screening efficiency and accuracy, thus saving medical resources.

Machine learning and multiple linear regression models can predict ascorbic acid and polyphenol contents, and antioxidant activity in strawberries.

Strawberry is a rich source of antioxidants, including ascorbic acid (ASA) and polyphenols, which have numerous health benefits. Antioxidant content and activity are often determined manually using laboratory equipment, which is destructive and time-consuming. This study constructs a prediction model for antioxidant compounds utilizing machine learning (ML) and multiple linear regression based on environmental, plant growth and agronomic fruit quality-related parameters as well as antioxidant levels. These were studied in three farms at two-week intervals during two years of cultivation. During the ML model screening, artificial neural network (ANN)-boosted models displayed a moderate coefficient of determination (R2) at 0.68-0.78 and relative root mean square error (RRMSE) at 3.8-4.8% in polyphenols and total ASA levels, as well as a high R2 of 0.96 and low RRMSE at <3.0% in antioxidant activity. Additionally, we developed variable selection models regarding the antioxidant activity, and variables two and five (environmental parameters and leaf length, respectively) with high accuracy were selected. The linear regression analysis between the actual and predicted data of antioxidants in the ANN-boosted models revealed high fitness with all parameters in almost all training, validation and test sets. Furthermore, environmental parameters are essential in developing such reliable models. We conclude that ANN-boosted, stepwise and double-Lasso regression models can predict antioxidant compounds with enhanced accuracy, and the relevant parameters can be easily acquired on-site without the need for any specific equipment. © 2024 Society of Chemical Industry.

Deep learning-based prediction of the dose-volume histograms for volumetric modulated arc therapy of left-sided breast cancer.

The advancements in artificial intelligence and computational power have made deep learning an attractive tool for radiotherapy treatment planning. Deep learning has the potential to significantly simplify the trial-and-error process involved in inverse planning required by modern treatment techniques such as volumetric modulated arc therapy (VMAT). In this study, we explore the ability of deep learning to predict organ-at-risk (OAR) dose-volume histograms (DVHs) of left-sided breast cancer patients undergoing VMAT treatment based solely on their anatomical characteristics. The predicted DVHs could be used to derive patient-specific dose constraints and dose objectives, streamlining the treatment planning process, standardizing the quality of the plans, and personalizing the treatment planning. This study aimed to develop a deep learning-based framework for the prediction of organ-specific dose-volume histograms (DVH) based on structures delineated for left-sided breast cancer treatment. We used a dataset of 249 left-sided breast cancer patients treated with tangential VMAT fields. We extracted delineated structures and dose distributions for each patient and derived slice-by-slice DVHs for planning target volume (PTV) and organs-at-risk. The patients were divided into training (70%, n=174), validation (10%, n=24), and test (20%, n=51) sets. Collected data were used to train a deep learning model for the prediction of the DVHs based on the delineated structures. The developed deep learning model comprised a modified DenseNet architecture followed by a recurrent neural network. In the independent test set (n=51), the point-wise differences in the slice-by-slice DVHs between the clinical and predicted DVHs were small; the mean squared errors were 3.53, 1.58, 2.28, 3.37, and 1.44 [×10-4] for PTV, heart, ipsilateral lung, contralateral lung, and contralateral breast, respectively. With the derived cumulative DVHs, the mean absolute difference±standard deviation of mean doses between the clinical and the predicted DVH were 0.08± 0.04Gy, 0.24± 0.22Gy, 0.73± 0.46Gy, 0.07± 0.06Gy, and 0.14± 0.14Gy for PTV, heart, ipsilateral lung, contralateral lung, and contralateral breast, respectively. The deep learning-based approach enabled automatic and reliable prediction of the DVH based on delineated structures. The predicted DVHs could potentially serve as patient-specific clinical goals used to aid treatment planning and avoid suboptimal plans or to derive optimization objectives and constraints for automated treatment planning.