Test Set Research Articles

Comparative analysis of machine learning algorithms for phishing website detection

As phishing assaults continue to pose a serious hazard in the digital world, trustworthy detection techniques are required. The effectiveness of machine learning techniques in detecting phishing websites is investigated in this study. The best-performing models were XGBoost and Multilayer Perceptrons (MLPs), which obtained test data accuracy of 90.4% and 90.3%, respectively. On the test data, the Random Forest and Decision Tree models showed competitive accuracies of 86.5% and 87.3%, respectively. SVMs, or support vector machines, performed admirably as well, obtaining an accuracy of 86.4% on the test set. Notably, with accuracy of 74.0% on the test data, the Autoencoder Neural Network demonstrated a restricted level of efficacy. These results highlight the effectiveness of XGBoost and MLPs in precisely detecting phishing websites, offering academics and practitioners in cybersecurity useful information.

Deep Learning Method for Rapid Simultaneous Multistructure Temporal Bone Segmentation.

To develop and validate a deep learning algorithm for the automated segmentation of key temporal bone structures from clinical computed tomography (CT) data sets. Cross-sectional study. A total of 325 CT scans from a clinical database. A state-of-the-art deep learning (DL) algorithm (SwinUNETR) was used to train a prediction model for rapid segmentation of 9 key temporal bone structures in a data set of 325 clinical CTs. The data set was manually annotated by a specialist to serve as the ground truth. The data set was randomly split into training (n = 260) and testing (n = 65) sets. The model's performance was objectively assessed through external validation on the test set using metrics including Dice, Balanced accuracy, Hausdorff distances, and processing time. The model achieved an average Dice coefficient of 0.87 for all structures, an average balanced accuracy of 0.94, an average Hausdorff distance of 0.79 mm, and an average processing time of 9.1 seconds per CT. The present DL model for the automated simultaneous segmentation of multiple structures within the temporal bone from CTs achieved high accuracy according to currently commonly employed objective analysis. The results demonstrate the potential of the method to improve preoperative evaluation and intraoperative guidance in otologic surgery.

Distinguishing between aldosterone-producing adenomas and non-functional adrenocortical adenomas using the YOLOv5 network.

You Only Look Once version 5 (YOLOv5), a one-stage deep-learning (DL) algorithm for object detection and classification, offers high speed and accuracy for identifying targets. To investigate the feasibility of using the YOLOv5 algorithm to non-invasively distinguish between aldosterone-producing adenomas (APAs) and non-functional adrenocortical adenomas (NF-ACAs) on computed tomography (CT) images. A total of 235 patients who were diagnosed with ACAs between January 2011 and July 2022 were included in this study. Of the 215 patients, 81 (37.7%) had APAs and 134 (62.3%) had NF-ACAs' they were randomly divided into either the training set or the validation set at a ratio of 9:1. Another 20 patients, including 8 (40.0%) with APA and 12 (60.0%) with NF-ACA, were collected for the testing set. Five submodels (YOLOv5n, YOLOv5s, YOLOv5m, YOLOv5l, and YOLOv5x) of YOLOv5 were trained and evaluated on the datasets. In the testing set, the mAP_0.5 value for YOLOv5x (0.988) was higher than the values for YOLOv5n (0.969), YOLOv5s (0.965), YOLOv5m (0.974), and YOLOv5l (0.983). The mAP_0.5:0.95 value for YOLOv5x (0.711) was also higher than the values for YOLOv5n (0.587), YOLOv5s (0.674), YOLOv5m (0.671), and YOLOv5l (0.698) in the testing set. The inference speed of YOLOv5n was 2.4 ms in the testing set, which was the fastest among the five submodels. The YOLOv5 algorithm can accurately and efficiently distinguish between APAs and NF-ACAs on CT images, especially YOLOv5x has the best identification performance.

Artificial Intelligence and Colposcopy: Automatic Identification of Cervical Squamous Cell Carcinoma Precursors

Background/Objectives: Proficient colposcopy is crucial for the adequate management of cervical cancer precursor lesions; nonetheless its limitations may impact its cost-effectiveness. The development of artificial intelligence models is experiencing an exponential growth, particularly in image-based specialties. The aim of this study is to develop and validate a Convolutional Neural Network (CNN) for the automatic differentiation of high-grade (HSIL) from low-grade dysplasia (LSIL) in colposcopy. Methods: A unicentric retrospective study was conducted based on 70 colposcopy exams, comprising a total of 22,693 frames. Among these, 8729 were categorized as HSIL based on histopathology. The total dataset was divided into a training (90%, n = 20,423) and a testing set (10%, n = 2270), the latter being used to evaluate the model’s performance. The main outcome measures included sensitivity, specificity, accuracy, positive predictive value (PPV), negative predictive value (NPV), and the area under the receiving operating curve (AUC-ROC). Results: The sensitivity was 99.7% and the specificity was 98.6%. The PPV and NPV were 97.8% and 99.8%, respectively. The overall accuracy was 99.0%. The AUC-ROC was 0.98. The CNN processed 112 frames per second. Conclusions: We developed a CNN capable of differentiating cervical cancer precursors in colposcopy frames. The high levels of accuracy for the differentiation of HSIL from LSIL may improve the diagnostic yield of this exam

Novel inflammatory and insulin resistance indices provide a clue in cerebral amyloid angiopathy

This study investigated the correlation of newly identified inflammatory and insulin resistance indices with cerebral amyloid angiopathy (CAA), and explored their potential to differentiate CAA from hypertensive arteriopathy (HA). We retrospectively analyzed 514 consecutive patients with cerebral small vessel disease (CSVD)-related haemorrhage, comparing the differences in novel inflammatory and insulin resistance indices between patients with CAA and HA. Univariate regression, LASSO and multivariate regression were used to screen variables and construct a classification diagnosis nomogram. Additionally, these biomarkers were explored in patients with mixed haemorrhagic CSVD. Inflammatory indices were higher in CAA patients, whereas insulin resistance indices were higher in HA patients. Further analysis identified neutrophil-to-lymphocyte ratio (NLR, OR 1.17, 95% CI 1.07–1.30, P < 0.001), and triglyceride–glucose index (TyG, OR = 0.56, 95% CI 0.36–0.83, P = 0.005) as independent factors for CAA. Therefore, we constructed a CAA prediction nomogram without haemorrhagic imaging markers. The nomogram yielded an area under the curve (AUC) of 0.811 (95% CI 0.764–0.865) in the training set and 0.830 (95% CI 0.718–0.887) in the test set, indicating an ability to identify high-risk CAA patients. These results show that CSVD patients can be phenotyped using novel inflammatory and insulin resistance indices, potentially allowing identification of high-risk CAA patients without haemorrhagic imaging markers.

Predicting Legal Systems: An Artificial Neural Network Approach with Statistical Analysis for Comparative Study of Civil Law and Common Law Countries

Abstract This study compares countries with common law with countries with civil law systems and investigates the possibility of predicting legal systems using artificial neural networks (ANNs). The OLS model, ANOVA, ANN, and Tensor Flow are used in the research to analyze the data. The goal is to find out how board characteristics and country legislative frameworks affect how European corporations disclose their social performance. The performance of a hidden layer with five nodes is best, according to the ANN model. The model’s accuracy throughout testing and validation is 0.750. The confusion matrix shows that, of the four observations in the test set, three were correctly categorized as “Civil law” and one was incorrectly categorized as “Common law.” To evaluate the model’s efficacy, evaluation metrics are computed. The model’s accuracy is 0.750, which represents a prediction success rate of 75%. For the “Civil law” class, the recall (true positive rate) is 1.0, indicating that all “Civil law” cases are correctly identified. Metrics for the “Common law” class, however, are not available due to the scant amount of data that is available. The prevalence of countries with common law and civil law systems is compared in the ANOVA analysis. As shown by the computed F-value of 0.482, there is less variance inside each legal system than there is between the two. There is no statistically significant difference in frequency between the two legal systems, according to the p-value of 0.495. Overall, the research’s conclusions imply that social performance disclosure between countries with common law and civil law systems differs only slightly. The neural network model’s network weights provide insight into the importance of different features in prediction.

Quantitative radiomics analysis of imaging features in adults and children Mycoplasma pneumonia

PurposeThis study aims to explore the value of clinical features, CT imaging signs, and radiomics features in differentiating between adults and children with Mycoplasma pneumonia and seeking quantitative radiomic representations of CT imaging signs.Materials and methodsIn a retrospective analysis of 981 cases of mycoplasmal pneumonia patients from November 2021 to December 2023, 590 internal data (adults:450, children: 140) randomly divided into a training set and a validation set with an 8:2 ratio and 391 external test data (adults:121; children:270) were included. Using univariate analysis, CT imaging signs and clinical features with significant differences (p &amp;lt; 0.05) were selected. After segmenting the lesion area on the CT image as the region of interest, 1,904 radiomic features were extracted. Then, Pearson correlation analysis (PCC) and the least absolute shrinkage and selection operator (LASSO) were used to select the radiomic features. Based on the selected features, multivariable logistic regression analysis was used to establish the clinical model, CT image model, radiomic model, and combined model. The predictive performance of each model was evaluated using ROC curves, AUC, sensitivity, specificity, accuracy, and precision. The AUC between each model was compared using the Delong test. Importantly, the radiomics features and quantitative and qualitative CT image features were analyzed using Pearson correlation analysis and analysis of variance, respectively.ResultsFor the individual model, the radiomics model, which was built using 45 selected features, achieved the highest AUCs in the training set, validation set, and external test set, which were 0.995 (0.992, 0.998), 0.952 (0.921, 0.978), and 0.969 (0.953, 0.982), respectively. In all models, the combined model achieved the highest AUCs, which were 0.996 (0.993, 0.998), 0.972 (0.942, 0.995), and 0.986 (0.976, 0.993) in the training set, validation set, and test set, respectively. In addition, we selected 11 radiomics features and CT image features with a correlation coefficient r greater than 0.35.ConclusionThe combined model has good diagnostic performance for differentiating between adults and children with mycoplasmal pneumonia, and different CT imaging signs are quantitatively represented by radiomics.

Cross-Regional Crop Classification Based on Sentinel-2

Accurate crop classification is of vital importance for agricultural water management. Most researchers have achieved crop classification by model optimization in the same temporal and regional domain by adjusting the value of input features. This study aims to improve the accuracy of crop classification across temporal and spatial domains. Sentinel-2 satellite imagery is employed for crop classification training and prediction in selected farming areas of Heilongjiang Province by calculating vegetation indices and constructing sequential input feature datasets. The HUNTS filtering method was used to mitigate the influence of cloud cover, which increased the stability and completeness of the input feature data across different years. To address the issue of shifts in the input feature values during cross-scale classification, this study proposes the hypothesis testing distribution method (HTDM). This method balances the distribution of input feature values in the test set even without knowing the crop distribution, thereby enhancing the accuracy of the classification test set. The results indicate that the HTDM significantly improves prediction accuracy in cases of substantial image quality variance. In 2022, the recognition accuracy for crop types at all farms processed by the HTDM was above 87%, showcasing the strong robustness of the HTDM.

An automated hybrid approach via deep learning and radiomics focused on the midbrain and substantia nigra to detect early-stage Parkinson’s disease

ObjectivesThe altered neuromelanin in substantia nigra pars compacta (SNpc) is a valuable biomarker in the detection of early-stage Parkinson’s disease (EPD). Diagnosis via visual inspection or single radiomics based method is challenging. Thus, we proposed a novel hybrid model that integrates radiomics and deep learning methodologies to automatically detect EPD based on neuromelanin-sensitive MRI, namely short-echo-time Magnitude (setMag) reconstructed from quantitative susceptibility mapping (QSM).MethodsIn our study, we collected QSM images including 73 EPD patients and 65 healthy controls, which were stratified into training-validation and independent test sets with an 8:2 ratio. Twenty-four participants from another center were included as the external validation set. Our framework began with the detection of the brainstem utilizing YOLO-v5. Subsequently, a modified LeNet was applied to obtain deep learning features. Meanwhile, 1781 radiomics features were extracted, and 10 features were retained after filtering. Finally, the classified models based on radiomics features, deep learning features, and the hybrid of both were established through machine learning algorithms, respectively. The performance was mainly evaluated using accuracy, net reclassification improvement (NRI), and integrated discrimination improvement (IDI). The saliency map was used to visualize the model.ResultsThe hybrid feature-based support vector machine (SVM) model showed the best performance, achieving ACC of 96.3 and 95.8% in the independent test set and external validation set, respectively. The model established by hybrid features outperformed the one radiomics feature-based (NRI: 0.245, IDI: 0.112). Furthermore, the saliency map showed that the bilateral “swallow tail” sign region was significant for classification.ConclusionThe integration of deep learning and radiomic features presents a potent strategy for the computer-aided diagnosis of EPD. This study not only validates the accuracy of our proposed model but also underscores its interpretability, evidenced by differential significance across various anatomical sites.

Development and validation of a risk scoring system to identify patients with lupus nephritis in electronic health record data

ObjectiveAccurate identification of lupus nephritis (LN) cases is essential for patient management, research and public health initiatives. However, LN diagnosis codes in electronic health records (EHRs) are underused, hindering efficient...

Cost prediction for water reuse equipment using interpretable machine learning models

During the initial stages of engineering projects management, the accurate estimation of equipment costs plays a crucial role in determining project approval decisions. However, there is a significant research gap regarding the estimation of investment costs for water reuse equipment. Additionally, cost predictions for such construction projects often suffer from limitations, including low accuracy, limited generalizability, and inefficiency. Advanced machine learning (ML) methods, renowned for their ability to model complex decision-making processes, offer powerful solutions. In this study, four traditional models and four ensemble models were employed to predict the cost of water reuse equipment. The results demonstrated that the ensemble models exhibited significantly superior predictive performance to that of traditional models with the three boosting ensemble models achieving the best performance (traditional models: training R2 = 29.77 %–94.85 %, testing R2 = 30.62 %–71.72 %; boosting ensemble models: training R2 = 97.42 %, testing R2 = 82.16 %–93.79 %). Furthermore, this study simplified the features of predictive models and identified the key variables that influence the cost of water reuse equipment using Shapley additive explanations (SHAP) method. The retrained ensemble models re-constructed based on the selected variables achieved significant predictive performance, with the Gradient Boosting Decision Tree (GBDT) outperforming the other models (training set R2 = 97.37 %, testing set R2 = 93.86 %). The water quantity, inflow conductivity, outflow conductivity, and recovery rate emerged as critical factors affecting the cost of water reuse equipment. Overall, the methods proposed in this study can enhance the versatility of cost prediction processes in environmental engineering scenarios, particularly those concerning the construction costs of water treatment equipment.

Portfolio management based on a reinforcement learning framework

AbstractPortfolio management is crucial for investors. We propose a dynamic portfolio management framework based on reinforcement learning using the proximal policy optimization algorithm. The two‐part framework includes a feature extraction network and a full connected network. First, the majority of the previous research on portfolio management based on reinforcement learning has been dedicated to discrete action spaces. We propose a potential solution to the problem of a continuous action space with a constraint (i.e., the sum of the portfolio weights is equal to 1). Second, we explore different feature extraction networks (i.e., convolutional neural network [CNN], long short‐term memory [LSTM] network, and convolutional LSTM network) combined with our system, and we conduct extensive experiments on the six kinds of assets, including 16 features. The empirical results show that the CNN performs best in the test set. Last, we discuss the effect of the trading frequency on our trading system and find that the monthly trading frequency has a higher Sharpe ratio in the test set than other trading frequencies.

Improved forecasting of the compressive strength of ultra‐high‐performance concrete (UHPC) via the CatBoost model optimized with different algorithms

AbstractThis paper focuses on the applicability of CatBoost models constructed using various optimization techniques for improved forecasting the compressive strength of ultra‐high‐performance concrete (UHPC). Phasor particle swarm optimization (PPSO), dwarf mongoose optimization (DMO), and atom search optimization (ASO), which have been very popular recently, are preferred as optimization algorithms. A comprehensive and reliable data set is used to develop the CatBoost models, which include 785 test results with 15 input features. The performance of the CatBoost models (PPSO‐CatBoost, DMO‐CatBoost, and ASO‐CatBoost) optimized with different algorithms is thoroughly assessed by means of various statistical metrics and error analysis to determine the model with the best forecasting capability, and this model is compared with the models obtained from previous studies. In addition, Shapley additive exPlanations (SHAP) analysis is used to ensure the interpretability of the forecasting models and to overcome the “black box” problem of machine learning (ML) models. The obtained results demonstrate that all CatBoost models outstandingly forecast the compressive strength of UHPC. Among these models, the DMO‐CatBoost model stands out compared to the other models in various performance metrics, such as high coefficient of determination (R2) values, low root mean squared error (RMSE), mean absolute percentage error (MAPE), and mean absolute error (MAE) values, along with a smaller error ratio. In other words, the RMSE, R2, MAPE, and MAE values of the DMO‐CatBoost model for the training set are 3.67, 0.993, 0.019, and 2.35, respectively, whereas those for the test set are 6.15, 0.978, 0.038, and 4.51. Additionally, the performance ranking of the algorithms used to optimize the hyperparameters of the CatBoost model is as follows: DMO &gt; PPSO &gt; ASO. On the other hand, SHAP analysis showed that age, fiber dosage, and cement dosage significantly influence the compressive strength of UHPC. These findings can guide structural engineers in the design and optimization of UHPC, thus assisting them in developing strategies to improve the strength properties of the material. Finally, based on the best forecasting model developed in this work, a graphical user interface has been developed to easily forecast the compressive strength of UHPC in practical applications without additional tools or software.

Different machine learning methods based on maxillary sinus in sex estimation for northwestern Chinese Han population.

Sex estimation is a critical aspect of forensic expertise. Some special anatomical structures, such as the maxillary sinus, can still maintain integrity in harsh environmental conditions and may be served as a basis for sex estimation. Due to the complex nature of sex estimation, several studies have been conducted using different machine learning algorithms to improve the accuracy of sex prediction from anatomical measurements. In this study, linear data of the maxillary sinus in the population of northwest China by using Cone-Beam Computed Tomography (CBCT) were collected and utilized to develop logistic, K-Nearest Neighbor (KNN), Support Vector Machine (SVM) and random forest (RF) models for sex estimation with R 4.3.1. CBCT images from 477 samples of Han population (75 males and 81 females, aged 5-17 years; 162 males and 159 females, aged 18-72) were used to establish and verify the model. Length (MSL), width (MSW), height (MSH) of both the left and right maxillary sinuses and distance of lateral wall between two maxillary sinuses (distance) were measured. 80% of the data were randomly picked as the training set and others were testing set. Besides, these samples were grouped by age bracket and fitted models as an attempt. Overall, the accuracy of the sex estimation for individuals over 18 years old on the testing set was 77.78%, with a slightly higher accuracy rate for males at 78.12% compared to females at 77.42%. However, accuracy of sex estimation for individuals under 18 was challenging. In comparison to logistic, KNN and SVM, RF exhibited higher accuracy rates. Moreover, incorporating age as a variable improved the accuracy of sex estimation, particularly in the 18-27 age group, where the accuracy rate increased to 88.46%. Meanwhile, all variables showed a linear correlation with age. The linear measurements of the maxillary sinus could be a valuable tool for sex estimation in individuals aged 18 and over. A robust RF model has been developed for sex estimation within the Han population residing in the northwestern region of China. The accuracy of sex estimation could be higher when age is used as a predictive variable.

Automatic 3D left atrial strain extraction framework on cardiac computed tomography

Background and objectiveStrain analysis provides insights into myocardial function and cardiac condition evaluation. However, the anatomical characteristics of left atrium (LA) inherently limit LA strain analysis when using echocardiography. Cardiac computed tomography (CT) with its superior spatial resolution, has become critical for in-depth evaluation of LA function. Recent studies have explored the feasibility of CT-derived strain; however, they relied on manually selected regions of interest (ROIs) and mainly focused on left ventricle (LV). This study aimed to propose a first-of-its-kind fully automatic deep learning (DL)-based framework for three-dimensional (3D) LA strain extraction on cardiac CT. MethodsA total of 111 patients undergoing ECG-gated contrast-enhanced CT for evaluating subclinical atrial fibrillation (AF) were enrolled in this study. We developed a 3D strain extraction framework on cardiac CT images, containing a 2.5D GN-U-Net network for LA segmentation, axis-oriented 3D view extraction, and LA strain measure. The segmentation accuracy was evaluated using Dice similarity coefficient (DSC). The model-extracted LA volumes and emptying fraction (EF) were compared with ground-truth measurements using intraclass correlation coefficient (ICC), correlation coefficient (r), and Bland-Altman plot (B-A). The automatically extracted LA strains were evaluated against the LA strains measured from 2D echocardiograms. We utilized this framework to gauge the effect of AF burden on LA strain, employing the atrial high rate episode (AHRE) burden as the measurement parameter. ResultsThe GN-U-Net LA segmentation network achieved a DSC score of 0.9603 on the test set. The framework-extracted LA estimates demonstrated excellent ICCs of 0.949 (95 % CI: 0.93–0.97) for minimal LA volume, 0.904 (95 % CI: 0.86–0.93) for maximal LA volume, and 0.902 (95 % CI: 0.86–0.93) for EF, compared with expert measurements. The framework-extracted LA strains demonstrated moderate agreement with the LA strains based on 2D echocardiography (ICCs >0.703). Patients with AHRE > 6 min had significantly lower global strain and LAEF, as extracted by the framework than those with AHRE ≤ 6 min. ConclusionThe promising results highlighted the feasibility and clinical usefulness of automatically extracting 3D LA strain from CT images using a DL-based framework. This tool could provide a 3D-based alternative to echocardiography for assessing LA function.

A longer time to relapse is associated with a larger increase in differences between paired primary and recurrent IDH wild-type glioblastomas at both the transcriptomic and genomic levels

Glioblastoma (GBM) is the most common malignant brain tumor in adults, which remains incurable and often recurs rapidly after initial therapy. While large efforts have been dedicated to uncover genomic/transcriptomic alternations associated with the recurrence of GBMs, the evolutionary trajectories of matched pairs of primary and recurrent (P-R) GBMs remain largely elusive. It remains challenging to identify genes associated with time to relapse (TTR) and construct a stable and effective prognostic model for predicting TTR of primary GBM patients. By integrating RNA-sequencing and genomic data from multiple datasets of patient-matched longitudinal GBMs of isocitrate dehydrogenase wild-type (IDH-wt), here we examined the associations of TTR with heterogeneities between paired P-R GBMs in gene expression profiles, tumor mutation burden (TMB), and microenvironment. Our results revealed a positive correlation between TTR and transcriptomic/genomic differences between paired P-R GBMs, higher percentages of non-mesenchymal-to-mesenchymal transition and mesenchymal subtype for patients with a short TTR than for those with a long TTR, a high correlation between paired P-R GBMs in gene expression profiles and TMB, and a negative correlation between the fitting level of such a paired P-R GBM correlation and TTR. According to these observations, we identified 55 TTR-associated genes and thereby constructed a seven-gene (ZSCAN10, SIGLEC14, GHRHR, TBX15, TAS2R1, CDKL1, and CD101) prognostic model for predicting TTR of primary IDH-wt GBM patients using univariate/multivariate Cox regression analyses. The risk scores estimated by the model were significantly negatively correlated with TTR in the training set and two independent testing sets. The model also segregated IDH-wt GBM patients into two groups with significantly divergent progression-free survival outcomes and showed promising performance for predicting 1-, 2-, and 3-year progression-free survival rates in all training and testing sets. Our findings provide new insights into the molecular understanding of GBM progression at recurrence and potential targets for therapeutic treatments.

Optimizing federated learning approaches with hybrid Convolutional Neural Networks ‐ Bidirectional Encoder Representations from Transformers for precise estimation of average localization errors in wireless sensor networks

SummaryWireless sensor networks (WSNs) require precise node location in order to function properly, and they are essential in many different applications. In this research, we propose a unique method to maximize the accurate estimate of average localization errors (ALEs) in WSNs by utilizing federated learning (FL) approaches. The suggested approach combines the extraction of spatial features with collection of contextual data through integration of hybrid convolutional neural network–bidirectional encoder representations from transformers (CNN‐BERT) model. Effectively applying min–max normalization to input features minimizes data from flowing into test, validation, and training sets. The technique is centered around a collaborative learning architecture, wherein the weights of the model are iteratively assessed and modified on centralized server. The suggested methodology's effectiveness in an array of settings is illustrated by mean absolute error (MAE), mean squared error (MSE), and root mean square error (RMSE) values for node localization in WSN. Network B demonstrated accuracy with an MAE of 0.05, MSE of 0.03, and RMSE of 0.10; Network C demonstrated strong accuracy with an MAE of 0.09, MSE of 0.08, and RMSE of 0.12; and Network A generated an MAE of 0.04, MSE of 0.06, and RMSE of 0.08. Furthermore, the centralized server, which is crucial for collaborative learning, obtained exceptional MAE of 0.01, MSE of 0.02, and RMSE of 0.99 demonstrating the superiority of the FL‐optimized method in improving localization accuracy. The outcomes demonstrate the FL‐optimized hybrid model's superiority over conventional methods in providing accurate node localization in a variety of WSN settings.

Using Machine Learning to Evaluate the Value of Genetic Liabilities in the Classification of Hypertension within the UK Biobank

Background and Objective: Hypertension increases the risk of cardiovascular diseases (CVD) such as stroke, heart attack, heart failure, and kidney disease, contributing to global disease burden and premature mortality. Previous studies have utilized statistical and machine learning techniques to develop hypertension prediction models. Only a few have included genetic liabilities and evaluated their predictive values. This study aimed to develop an effective hypertension classification model and investigate the potential influence of genetic liability for multiple risk factors linked to CVD on hypertension risk using the random forest and the neural network. Materials and Methods: The study involved 244,718 European participants, who were divided into training and testing sets. Genetic liabilities were constructed using genetic variants associated with CVD risk factors obtained from genome-wide association studies (GWAS). Various combinations of machine learning models before and after feature selection were tested to develop the best classification model. The models were evaluated using area under the curve (AUC), calibration, and net reclassification improvement in the testing set. Results: The models without genetic liabilities achieved AUCs of 0.70 and 0.72 using the random forest and the neural network methods, respectively. Adding genetic liabilities improved the AUC for the random forest but not for the neural network. The best classification model was achieved when feature selection and classification were performed using random forest (AUC = 0.71, Spiegelhalter z score = 0.10, p-value = 0.92, calibration slope = 0.99). This model included genetic liabilities for total cholesterol and low-density lipoprotein (LDL). Conclusions: The study highlighted that incorporating genetic liabilities for lipids in a machine learning model may provide incremental value for hypertension classification beyond baseline characteristics.

Abstract B003: Preliminary machine learning integration of DNA methylation-based tumor immune microenvironment deconvolution with histopathological slides for bladder cancer prognostication

Abstract Background: Analyzing the cellular composition of the tumor microenvironment (TME) is crucial for bladder cancer (BCa) prognosis but is hampered by technical and financial limitations of immunofluorescence, flow cytometry and scRNASeq, primarily due to cell disaggregation challenges, antibody selection and cost/throughput constraints. A DNA methylation-based (DNAm) technique, HiTIMED, was introduced, enabling precise differentiation of 17 unique cell types within solid tumors– including those related to tumor growth, angiogenesis, and immune responses– enhancing our ability to investigate the complex interactions within TME tied to exposures, patient outcomes, and treatment efficacy. Automated morphological identification of cellular signatures on histopathological slides has been fraught with challenges. The potential to identify signatures at higher resolution, beyond immune cells, could improve disease management and application of Whole Slide Imaging (WSI) in clinical research. Our research explores direct estimation of HiTIMED-inferred cellular distributions from WSI and benefits of multimodal integration of WSI with HiTIMED proportions for BCa prognostication. Methods: BCa WSI and Illumina Methylation 450k DNAm data for BCa were obtained from TCGA (MIBC, n=456) and Dartmouth-Hitchcock (NMIBC, n=136), divided into training, validation, and testing sets. HiTIMED deconvolved DNAm into 17 cellular components. Survival, stage, and demographics were incorporated. Graph convolutional networks (GCNs) predicted HiTIMED-derived cellular proportions from WSI (WSI2HiTIMED). Logistic regression and Cox Proportional Hazards (CoxPH) were used to compare how standardized cellular proportions estimated by HiTIMED and WSI2HiTIMED each correlate with advanced T-stage (T&amp;gt;2) and survival with age and sex adjustment. We combined a GCN modeling WSI with a neural network using HiTIMED proportions to predict 10-year survival, adjusting for age and sex with CoxPH. This integrated approach was benchmarked against CoxPH models using only imaging or cellular data. Results: HiTIMED and WSI2HiTIMED demonstrated consistent associations with advanced T-stage, respectively: 1) endothelial cells (OR=2.70, p=0.0095; OR=2.24, p=0.033), 2) epithelial cells (OR=0.02, p=0.0005; OR=0.51, p=0.112), 3) lymphocytes (OR=2.04, p=0.0635; OR=2.13, p=0.041), 4) stromal cells (OR=4.30, p=0.001; OR=2.32, p=0.0235), and 5) tumor cells (OR=0.41, p=0.0165; OR=0.42, p=0.020). The prognostic value for survival based on endothelial cell estimates from HiTIMED and WSI2HiTIMED was comparable, with p-values of 0.006 and 0.009, respectively. The C-index for BCa survival prediction improved from 0.730 using WSI GCN-derived hazards to 0.766 with HiTIMED proportions. Conclusion: DNAm offers a cost-effective method for improving cell type estimation from WSI and provides complementary information for BCa analysis. Future efforts will focus on enhancing precision across additional cell types with larger cohorts and exploring synergy with pathologist annotations and spatial transcriptomics. Citation Format: Zarif L. Azher, Ze Zhang, Brock C. Christensen, Lucas A. Salas, Margaret R. Karagas, Louis J. Vaickus, Hideki Fuyura, Keluo Yao, Joshua J. Levy. Preliminary machine learning integration of DNA methylation-based tumor immune microenvironment deconvolution with histopathological slides for bladder cancer prognostication [abstract]. In: Proceedings of the AACR Special Conference on Bladder Cancer: Transforming the Field; 2024 May 17-20; Charlotte, NC. Philadelphia (PA): AACR; Clin Cancer Res 2024;30(10_Suppl):Abstract nr B003.

Luminal androgen receptor subtype and tumor-infiltrating lymphocytes groups based on triple-negative breast cancer molecular subclassification.

In our previous study, we developed a triple-negative breast cancer (TNBC) subtype classification that correlated with the TNBC molecular subclassification. In this study, we aimed to evaluate the predictor variables of this subtype classification on the whole slide and to validate the model's performance by using an external test set. We explored the characteristics of this subtype classification and investigated genomic alterations, including genomic scar signature scores. First, TNBC was classified into the luminal androgen receptor (LAR) and non-luminal androgen receptor (non-LAR) subtypes based on the AR Allred score (≥ 6 and < 6, respectively). Then, the non-LAR subtype was further classified into the lymphocyte-predominant (LP), lymphocyte-intermediate (LI), and lymphocyte-depleted (LD) groups based on stromal tumor-infiltrating lymphocytes (TILs) (< 20%, > 20% but < 60%, and ≥ 60%, respectively). This classification showed fair agreement with the molecular classification in the test set. The LAR subtype was characterized by a high rate of PIK3CA mutation, CD274 (encodes PD-L1) and PDCD1LG2 (encodes PD-L2) deletion, and a low homologous recombination deficiency (HRD) score. The non-LAR LD TIL group was characterized by a high frequency of NOTCH2 and MYC amplification and a high HRD score.