Performance Of Classifier Research Articles

ECG signal generation using feature disentanglement auto-encoder.

The demand for ECG datasets, particularly those containing rare classes, poses a significant challenge as deep learning becomes increasingly prevalent in ECG signal research. While Generative Adversarial Networks (GANs) and Variational Autoencoders (VAEs) are widely adopted, they encounter difficulties in effectively generating samples for classes with limited instances. To address this issue, we propose a novel Feature Disentanglement Auto-Encoder (FDAE) designed to dissect various generative factors under a contrastive learning framework within ECG data to facilitate the generation of new ECG samples. The FDAE enhances and extends the AE structure with novel methodologies, which involve: (1) partitioning the latent space into three distinct representations to capture various generative factors; (2) utilizing a contrastive loss function to improve feature disentanglement capabilities; and (3) incorporating additional classifiers to enhance representation learning, alongside a discriminator aimed at boosting the realism of synthesized signals. Furthermore, our FDAE generates new signals by swapping latent codes of existing signals and combining freely or substituting patient-independent representations with those randomly generated by a VAE. To validate our approach, we conduct heartbeat classification experiments on the publicly available MIT-BIH arrhythmia database, using FAKE-train/FAKE-test partitions and data augmentation. The results highlight the FDAE's ability to improve ECG classifier performance and excel in synthesizing ECG signals. Furthermore, we apply the model to the Icentia11K dataset and conducted classification enhancement experiments. The results further highlight the model's strong generalization ability in ECG synthesis. This work has the potential to improve the robustness and generalization of deep learning models for ECG analysis, particularly in medical applications where rare cardiac events are often underrepresented in available datasets.

Automated system for diagnosing pulmonary fibrosis using crackle analysis in recorded lung sounds based on iterative envelope mean fractal dimension filter.

Patients with pulmonary fibrosis (PF) often experience long waits before getting a correct diagnosis, and this delay in reaching specialized care is associated with increased mortality, regardless of the severity of the disease. Early diagnosis and timely treatment of PF can potentially extend life expectancy and maintain a better quality of life. Crackles present in the recorded lung sounds may be crucial for the early diagnosis of PF. This paper describes an automated system for differentiating lung sounds related to PF from other pathological lung conditions using the average number of crackles per breath cycle (NOC/BC). The system is divided into four main parts: (1) pre-processing, (2) separation of crackles from normal breath sounds using the iterative envelope mean fractal dimension (IEM-FD) filter, (3) crackle verification and counting, and (4) estimating NOC/BC. The system was tested on a dataset consisting of 48 (24 fibrotic and 24 non-fibrotic) subjects and the results were compared with an assessment by two expert respiratory physicians. The set of HRCT images, reviewed by two expert radiologists for the presence or absence of pulmonary fibrosis, was used as the ground truth for evaluating the PF and non-PF classification performance of the system. The overall performance of the automatic classifier based on receiver operating curve-derived cut-off value for average NOC/BC of 18.65 (AUC=0.845, 95 % CI 0.739-0.952, p<0.001; sensitivity=91.7 %; specificity=59.3 %) compares favorably with the averaged performance of the physicians (sensitivity=83.3 %; specificity=56.25 %). Although radiological assessment should remain the gold standard for diagnosis of fibrotic interstitial lung disease, the automatic classification system has strong potential for diagnostic support, especially in assisting general practitioners in the auscultatory assessment of lung sounds to prompt further diagnostic work up of patients with suspect of interstitial lung disease.

Improved deep canonical correlation fusion approach for detection of early mild cognitive impairment.

Detection of early mild cognitive impairment (EMCI) is clinically challenging as it involves subtle alterations in multiple brain sub-anatomic regions. Among different brain regions, the corpus callosum and lateral ventricles are primarily affected due to EMCI. In this study, an improved deep canonical correlation analysis (CCA) based framework is proposed to fuse magnetic resonance (MR) image features from lateral ventricular and corpus callosal structures for the detection of EMCI condition. For this, obtained structural MR images of healthy controls and EMCI subjects are preprocessed. Lateral ventricles and corpus callosum structures are segmented from these images and features are extracted. Extracted features from different brain structures are fused using non-linear orthogonal iteration-based deep CCA. Fused features are employed to differentiate healthy controls and EMCI condition using extreme learning machine classifier. Results indicate that fused callosal and ventricular features are able to detect EMCI. Improved deep CCA algorithm with tuned hyperparameters achieves the highest classifier performance with an F-score of 82.15%. The proposed framework is compared with state-of-the-art CCA approaches, and the results demonstrate its improved performance in EMCI detection. This highlights the potential of the proposed framework in the automated diagnosis of preclinical MCI conditions.

Automated system for diagnosing pulmonary fibrosis using crackle analysis in recorded lung sounds based on iterative envelope mean fractal dimension filter.

Patients with pulmonary fibrosis (PF) often experience long waits before getting a correct diagnosis, and this delay in reaching specialized care is associated with increased mortality, regardless of the severity of the disease. Early diagnosis and timely treatment of PF can potentially extend life expectancy and maintain a better quality of life. Crackles present in the recorded lung sounds may be crucial for the early diagnosis of PF. This paper describes an automated system for differentiating lung sounds related to PF from other pathological lung conditions using the average number of crackles per breath cycle (NOC/BC). The system is divided into four main parts: (1) pre-processing, (2) separation of crackles from normal breath sounds using the iterative envelope mean fractal dimension (IEM-FD) filter, (3) crackle verification and counting, and (4) estimating NOC/BC. The system was tested on a dataset consisting of 48 (24 fibrotic and 24 non-fibrotic) subjects and the results were compared with an assessment by two expert respiratory physicians. The set of HRCT images, reviewed by two expert radiologists for the presence or absence of pulmonary fibrosis, was used as the ground truth for evaluating the PF and non-PF classification performance of the system. The overall performance of the automatic classifier based on receiver operating curve-derived cut-off value for average NOC/BC of 18.65 (AUC=0.845, 95 % CI 0.739-0.952, p<0.001; sensitivity=91.7 %; specificity=59.3 %) compares favorably with the averaged performance of the physicians (sensitivity=83.3 %; specificity=56.25 %). Although radiological assessment should remain the gold standard for diagnosis of fibrotic interstitial lung disease, the automatic classification system has strong potential for diagnostic support, especially in assisting general practitioners in the auscultatory assessment of lung sounds to prompt further diagnostic work up of patients with suspect of interstitial lung disease.

UniEmbed: A Novel Approach to Detect XSS and SQL Injection Attacks Leveraging Multiple Feature Fusion with Machine Learning Techniques

Abstract Web applications are essential in the digital age, but their security vulnerabilities expose sensitive data and organizational integrity to sophisticated attacks. Among the most prevalent and damaging vulnerabilities in web applications are cross-site scripting (XSS) and SQL injection attacks. In this paper, we introduce UniEmbed, a unified approach for detecting XSS and SQL injection attacks using machine learning classifiers. This novel approach leverages natural language processing techniques, combining features from Word2Vec, the Universal Sentence Encoder (USE), and FastText to extract meaningful data from web applications. Extensive experiments were conducted using various machine learning classifiers on three benchmark datasets to evaluate the performance of the unified detection approach, demonstrating exceptional results. Experimental results demonstrate the superior performance of the MLP classifier. For the XSS attack dataset, the MLP classifier achieved an accuracy of 0.9982 and an F1-score of 0.9983, with minimal false positives and false negatives. Similarly, the hard voting classifier yielded the same outstanding results. For SQL injection attacks, the MLP classifier maintained exceptional performance, achieving an F1-score of 0.9980 and accuracy rates exceeding 0.9980 across two datasets. The classifier effectively minimized false positives and false negatives. The ROC curves further corroborate the effectiveness of the proposed method, indicating high true positive rates and low false positive rates. Furthermore, comparative analysis showed that the UniEmbed method consistently outperformed individual feature extraction methods across all classifiers. These findings indicate that the proposed UniEmbed method, particularly when combined with the MLP classifier, is highly effective in detecting both XSS and SQL injection attacks, making it a promising approach for enhancing web application security.

An Enhanced Approach Using AGS Network for Skin Cancer Classification

Skin cancer accounts for over 40% of all cancer diagnoses worldwide. However, accurately diagnosing skin cancer remains challenging for dermatologists, as multiple types of skin cancer often appear visually similar. The diagnostic accuracy of dermatologists ranges between 62% and 80%. Although AI models have shown promise in assisting with skin cancer classification in various studies, obtaining the large-scale medical image datasets required for AI model training is not straightforward. To address this limitation, this study proposes the AGS network, designed to overcome the challenges of small datasets and enhance the performance of skin cancer classifiers. The AGS network integrates three key modules: Augmentation (A), GAN (G), and Segmentation (S). It was evaluated using eight deep learning classifiers—GoogLeNet, DenseNet201, ResNet50, MobileNet V3, EfficientNet B0, ViT, EfficientNet V2, and Swin Transformers—on the HAM10000 dataset. Five model configurations were also tested to assess the contribution of each module. The results showed that all eight classifiers demonstrated consistent performance improvements with the AGS network. In particular, EfficientNet V2 + AGS achieved the most significant performance gains over the baseline model, with an increase of +0.1808 in Accuracy and +0.1674 in F1-Score. Among all configurations, ResNet50+AGS achieved the best overall performance, with an Accuracy of 95.87% and an F1-Score of 95.73%. While most previous studies focused on single augmentation methods, this study demonstrates the effectiveness of combining multiple augmentation techniques within an integrated framework. The AGS network demonstrates how integrating diverse methods can improve the performance of skin cancer classification models.

Evaluating the Exposome Score for Schizophrenia in a Transdiagnostic Psychosis Cohort: Associations With Psychosis Risk, Symptom Severity, and Personality Traits.

Investigations of causal pathways for psychosis can be guided by the identification of environmental risk factors. A recently developed composite risk tool, the exposome score for schizophrenia (ES-SCZ), which controls for intercorrelations between risk factors, has shown fair to good performance. We tested the transdiagnostic psychosis classifier performance of the ES-SCZ with the Bipolar-Schizophrenia Network for Intermedial Phenotypes data and examined its relationship with clinical-level outcomes. We computed the case-control classifier performance for the ES-SCZ from cross-sectional data on 1055 volunteers with psychotic diagnoses (schizophrenia, schizoaffective, bipolar psychosis) and 510 controls. Multivariate regression models were used to control for the correlations between outcomes and to correct for the effects of age, sex, and family socioeconomic status across outcomes. We estimated association for the ES-SCZ with psychosis and mood symptom severity, the 5-factor model of personality, and function across biologically defined biotypes, traditional diagnostic categories, and controls. ES-SCZ classifier performance for psychosis was fair to good. ES-SCZ associations with personality factor scores were qualitatively similar between psychosis groups and controls with decreased conscientiousness and agreeableness and increased neuroticism. The patterns of associations between ES-SCZ and symptoms differed across biotypes and diagnoses. Biotype 3 and bipolar disorder had consistent within-group associations where greater exposome score predicted more severe symptoms and worse function. ES-SCZ performance was consistent with previous reports in this transdiagnostic psychosis sample (adjusted odds ratio: 3.331 [2.834, 3.915], P < .001; area under the curve: 0.762 [0.735, 0.789]). Individual differences in ES-SCZ magnitude may be useful for investigating causal pathways between developmentally relevant exposures and symptomatic expression of psychosis.

A comprehensive study on the interplay between dataset characteristics and oversampling methods

Addressing class imbalance in oversampling domain using machine learning methods requires careful selection of techniques and classifiers for optimal outcomes. While the importance of technique choice is well recognized, research on how dataset characteristics affect classification results remained limited. This study fills this gap by analyzing 16 datasets, categorized by financial relevance, temporal relevance, minority rate, minority sample count, and feature count. The effectiveness of various oversampling techniques is systematically evaluated and ranked using F1 and AUC, providing a structured framework for assessing the suitability of these techniques across diverse datasets. The evaluation involved 15 classifiers, resulting in 75 models combining four oversampling techniques and a baseline classifier. A ranking mechanism identified five top-performing models, emphasizing that classifier performance is influenced by the choice of the oversampling method, depending on dataset type. Notably, the traditional Synthetic Minority Oversampling Technique (SMOTE) outperformed the approaches based on Generative Adversarial Network (GAN) across different classifiers and datasets. Among classifiers, random forest proved to be the most robust across all dataset types, surpassing boosting-based classifiers. Overall, this study provides valuable insights into selecting the optimal oversampling methods and classifiers for specific dataset characteristics, offering a framework for addressing class imbalance in various contexts.

Exploration of designing an automatic classifier for questions containing code snippets-A case study of Oracle SQL certification exam questions.

This study uses the Oracle SQL certification exam questions to explore the design of automatic classifiers for exam questions containing code snippets. SQL's question classification assigns a class label in the exam topics to a question. With this classification, questions can be selected from the test bank according to the testing scope to assemble a more suitable test paper. Classifying questions containing code snippets is more challenging than classifying questions with general text descriptions. In this study, we use factorial experiments to identify the effects of the factors of the feature representation scheme and the machine learning method on the performance of the question classifiers. Our experiment results showed the classifier with the TF-IDF scheme and Logistics Regression model performed best in the weighted macro-average AUC and F1 performance indices. The classifier with TF-IDF and Support Vector Machine performed best in weighted macro-average Precision. Moreover, the feature representation scheme was the main factor affecting the classifier's performance, followed by the machine learning method, over all the performance indices.

Enhancing the Optimization of BI-LSTM Classifier with Ensemble Methods, Regularization, and Cross-Validation Techniques for Email Spam Detection

Email spam, a persistent and escalating issue, continues to disrupt the digital communication landscape, causing inconvenience and time loss for users worldwide. With technological advancements, spammers continually adapt and refine their tactics to infiltrate email inboxes. Staying current with state-of-the-art anti-spam techniques is imperative to secure emails and eliminate unwanted messages. Our research work embarks on an exploration of supercharging email spam detection through the augmentation of a Bidirectional Long Short-Term Memory (BI-LSTM) classifier. Our approach integrates ensemble methods, regularization techniques, and cross-validation into the fabric of the BI-LSTM model, creating a formidable spam detection system. Our paper delves into the intricate technical aspects of these methodologies, elucidating their synergy in fortifying the classifier's performance

Artificial intelligence-enabled safety monitoring in Alzheimer's disease clinical trials.

Investigators conducting clinical trials have an ethical, scientific, and regulatory obligation to protect the safety of trial participants. Traditionally, safety monitoring includes manual review and coding of adverse event data by expert clinicians. Our study explores the use of natural language processing (NLP) and artificial intelligence (AI) methods to streamline and standardize clinician coding of adverse event data in Alzheimer's disease (AD) clinical trials. Our quantitative retrospective study aimed to develop a gold standard AD adverse event data set, evaluate the predictive performance of NLP-based models to classify adverse events, and determine whether automated coding is more efficient, accurate, reliable, and consistent than clinician coding. Our study was conducted at the University of Southern California's Alzheimer's Therapeutic Research Institute (ATRI). ATRI serves as the clinical and data coordinating center for the Alzheimer's Clinical Trial Consortium (ACTC). We collected demographic and adverse event data from eight completed clinical trials in participants (n=1920) with symptomatic AD conducted between 2005 and 2020. Original expert clinician-confirmed codes were used for all model performance comparisons. F1 score was used as the primary model selection metric. Final classifier performance was evaluated using predictive accuracy. Clinician effort was measured in time to code, review, and confirm coded adverse events. In a sample of 1000 adverse events, AI-based AE coding achieved higher accuracy (∼20% increase in accuracy) and was more cost-effective (∼80% cost reduction) than traditional clinician coding. Our study results demonstrate how approaches that effectively combine AI and human expertise can improve the efficiency and quality of adverse event coding and clinical trial safety monitoring.

Real-Time Freezing of Gait Prediction and Detection in Parkinson's Disease.

Freezing of gait (FOG) is a walking disturbance that can lead to postural instability, falling, and decreased mobility in people with Parkinson's disease. This research used machine learning to predict and detect FOG episodes from plantar-pressure data and compared the performance of decision tree ensemble classifiers when trained on three different datasets. Dataset 1 (n = 11) was collected in a previous study. Dataset 2 (n = 10) included six new participants and four participants from Dataset 1 who were re-tested (approximately 2 years later), and Dataset 3 (n = 21) combined Datasets 1 and 2. The prediction model trained on Dataset 3 had a 2.28% higher sensitivity and 3.09% lower specificity compared to the models trained on Dataset 1. The model trained on Dataset 3 identified 86.84% of the total FOG episodes compared to 74.31% from the model trained on Dataset 1. Also, the model using Dataset 3 identified the FOG episodes 0.3 s earlier than the model developed with Dataset 1. The model trained using Dataset 3 showed improved performance in sensitivity, identification time, and FOG identification. The improvements using the expanded dataset (Dataset 3) in this study compared to the previous model reinforce the validity and generalizability of the original model. The model was able to predict and detect FOG well and is, therefore, ready to be implemented in a FOG prevention device.

Impact of Clinical Factors on Accuracy of Ovarian Cancer Detection via Platelet RNA Profiling

Ovarian cancer (OC) presents a diagnostic challenge, often resulting in poor patient outcomes. Platelet RNA sequencing, which reflects the host's response to disease, shows promise for earlier OC detection. This study examines the impact of sex, age, platelet count and the training on cancer types other than OC on classification accuracy achieved in the previous platelet-alone training dataset. A total of 339 samples from healthy donors and 1396 samples from cancer patients, spanning 18 cancer types (including 135 OC cases) were analyzed. Logistic Regression was applied to verify our classifiers' performance and interpretability. Models were tested at 100% specificity and 100% sensitivity levels. Incorporating patients' age as an additional feature along with gene expression increased sensitivity from 68.6% to 72.6%. Models trained on data from both sexes and on female-only data achieved a sensitivity of 68.6% and 74.5%, respectively. Training solely on OC data reduced late-stage sensitivity from 69.1% to 44.1 but increased early-stage sensitivity from 66.7% to 69.7%. This study highlights the potential of platelet RNA profiling for OC detection and the importance of clinical variables in refining classification accuracy. Incorporating age with gene expression data may enhance OC diagnostic accuracy. The inclusion of male samples deteriorates classifier performance. Data from diverse cancer types improves advanced cancer detection but negatively impacts early-stage diagnosis.

Artificial intelligence tools for engagement prediction in neuromotor disorder patients during rehabilitation

BackgroundRobot-Assisted Gait Rehabilitation (RAGR) is an established clinical practice to encourage neuroplasticity in patients with neuromotor disorders. Nevertheless, tasks repetition imposed by robots may induce boredom, affecting clinical outcomes. Thus, quantitative assessment of engagement towards rehabilitation using physiological data and subjective evaluations is increasingly becoming vital.This study aimed at methodologically exploring the performance of artificial intelligence (AI) algorithms applied to structured datasets made of heart rate variability (HRV) and electrodermal activity (EDA) features to predict the level of patient engagement during RAGR.MethodsThe study recruited 46 subjects (38 underage, 10.3 ± 4.0 years old; 8 adults, 43.0 ± 19.0 years old) with neuromotor impairments, who underwent 15 to 20 RAGR sessions with Lokomat. During 2 or 3 of these sessions, ad hoc questionnaires were administered to both patients and therapists to investigate their perception of a patient’s engagement state. Their outcomes were used to build two engagement classification targets: self-perceived and therapist-perceived, both composed of three levels: “Underchallenged”, “Minimally Challenged”, and “Challenged”. Patient’s HRV and EDA physiological signals were processed from raw data collected with the Empatica E4 wristband, and 33 features were extracted from the conditioned signals. Performance outcomes of five different AI classifiers were compared for both classification targets. Nested k-fold cross-validation was used to deal with model selection and optimization. Finally, the effects on classifiers performance of three dataset preparation techniques, such as unimodal or bimodal approach, feature reduction, and data augmentation, were also tested.ResultsThe study found that combining HRV and EDA features into a comprehensive dataset improved the synergistic representation of engagement compared to unimodal datasets. Additionally, feature reduction did not yield any advantages, while data augmentation consistently enhanced classifiers performance. Support Vector Machine and Extreme Gradient Boosting models were found to be the most effective architectures for predicting self-perceived engagement and therapist-perceived engagement, with a macro-averaged F1 score of 95.6% and 95.4%, respectively.ConclusionThe study displayed the effectiveness of psychophysiology-based AI models in predicting rehabilitation engagement, thus promoting their practical application for personalized care and improved clinical health outcomes.

Accurate Diagnosis of Lower Respiratory Infections Using Host Response and Respiratory Microbiome from a Single Metatranscriptome Test of Bronchoalveolar Lavage Fluid.

Lower respiratory tract infections (LRTIs) diagnosis is challenging because noninfectious diseases mimic its clinical features. The altered host response and respiratory microbiome following LRTIs have the potential to differentiate LRTIs from noninfectious respiratory diseases (non-LRTIs). Patients suspected of having LRTIs are retrospectively enrolled and a clinical metatranscriptome test is performed on bronchoalveolar lavage fluid (BALF). Transcriptomic and metagenomic analysis profiled the host response and respiratory microbiome in patients with confirmed LRTI (n = 126) or non-LRTIs (n = 75). Patients with evidenced LRTIs exhibited enhanced pathways on chemokine and cytokine response, neutrophile recruitmentand activation, along with specific gene modules linked to LRTIs status and key blood markers. Moreover, LRTIs patients exhibited reduced diversity and evenness in the lower respiratory microbiome, likely driven by an increased abundance of bacterial pathogens. Host marker genes are selected, and classifiers are developed to distinguish patients with LRTIs, non-LRTIs, and indeterminate status, achieving an area under the receiver operating characteristic curve of 0.80 to 0.86 and validated in a subsequently enrolled cohort. Incorporatingrespiratory microbiome features further enhanced the classifier's performance. In summary, a single metatranscriptome test of BALF proved detailed profiles of host response and respiratory microbiome, enabling accurate LRTIs diagnosis.

A Hierarchical Machine Learning-Based Strategy for Mapping Grassland in Manitoba’s Diverse Ecoregions

Accurate and reliable knowledge about grassland distribution is essential for farmers, stakeholders, and government to effectively manage grassland resources from agro-economical and ecological perspectives. This study developed a novel pixel-based grassland classification approach using three supervised machine learning (ML) algorithms, which were assessed in the province of Manitoba, Canada. The grassland classification process involved three stages: (1) to distinguish between vegetation and non-vegetation covers, (2) to differentiate grassland from non-grassland landscapes, and (3) to identify three specific grassland classes (tame, native, and mixed grasses). Initially, this study investigated different satellite data, such as Sentinel-1 (S1), Sentinel-2 (S2), and Landsat 8 and 9, individually and combined, using the random forest (RF) method, with the best performance at the first two steps achieved using a combination of S1 and S2. The combination was then utilized to conduct the first two steps of classification using support vector machine (SVM) and gradient tree boosting (GTB). In step 3, after filtering out non-grassland pixels, the performance of RF, SVM, and GTB classifiers was evaluated with combined S1 and S2 data to distinguish different grassland types. Eighty-nine multitemporal raster-based variables, including spectral bands, SAR backscatters, and digital elevation models (DEM), were input for ML models. RF had the highest classification accuracy at 69.96% overall accuracy (OA) and a Kappa value of 0.55. After feature selection, the variables were reduced to 61, increasing OA to 72.62% with a Kappa value of 0.58. GTB ranked second, with its OA and Kappa values improving from 67.69% and 0.50 to 72.18% and 0.58 after feature selection. The impact of raster data quality on grassland classification accuracy was assessed through multisensor image fusion. Grassland classification using the Hue, Saturation, and Value (HSV) fused images showed higher OA (59.18%) and Kappa values (0.36) than the Brovey Transform (BT) and non-fused images. Finally, a web map was created to show grassland results within the Soil Landscapes of Canada (SLC) polygons, relating soil landscapes to grassland distribution and providing valuable information for decision-makers and researchers. Future work may include extending the current methodology by considering other influential variables, like meteorological parameters or soil properties, to create a comprehensive grassland inventory across the whole Prairie ecozone of Canada.

Primary angle-closed diseases recognition through artificial intelligence-based anterior segment-optical coherence tomography imaging.

In this study, artificial intelligence (AI) was used to deeply learn the classification of the anterior segment-Optical Coherence Tomography (AS-OCT) images. This AI systems automatically analyzed the angular structure of the AS-OCT images and automatically classified anterior chamber angle. It would improve the efficiency of AS-OCT image analysis. The subjects were from the glaucoma disease screening and prevention project for elderly people in Shanghai community. Each scan contained 72 cross-sectional AS-OCT frames. We developed a deep learning-based AS-OCT image automatic anterior chamber angle analysis software. Classifier performance was evaluated against glaucoma experts' grading of AS-OCT images as standard. Outcome evaluation included accuracy (ACC) and area under the receiver operator curve (AUC). 94895 AS-OCT images were collected from 687 participants, in which 69,243 images were annotated as open, 16,433 images were annotated as closed, and 9219 images were annotated as non-gradable. The class-balanced train data were formed from randomly extracting the same number of open angle images as the closed angle images, which contained 22,393 images (11127 open, 11256 closed). The best-performing classifier was developed by applying transfer learning to the ResNet-50 architecture. against experts' grading, this classifier achieved an AUC of 0.9635. Deep learning classifiers effectively detect angle closure based on automated analysis of AS-OCT images. This system could be used to automate clinical evaluations of the anterior chamber angle and improve efficiency of interpreting AS-OCT images. The results demonstrated the potential of the deep learning system for rapid recognition of high-risk populations of PACD.

Multitask classification: assessing data complexity and determining correlations with classifier performance

Multitask classification: assessing data complexity and determining correlations with classifier performance

The Effects of the Guide Cone on the Flow Field and Key Classification Performance of an Industrial-Scale Micron Air Classifier

In this study, the effects of the structural parameters (SPs) of the guide cone, such as the surface inclination and the material recirculation gap size, on the two-phase flow field and classification performance of a real-sized industrial-scale micron air classifier were investigated. This was achieved using the two-way coupling of a computational fluid dynamics–discrete phase model in ANSYS 2022 R2, with the assistance of a high-performance system (HPC). The objective of this study was to determine the optimal SPs of the guide cone so as to achieve the best classification efficiency and satisfy the required particle size distribution curve, named the know-how curve (KHC), for the particle size range (0 ÷ 400 μm) used in producing quartz-based artificial stone. The bottom diameter (d) of the guide cone (CHL) was altered while keeping the outer diameter of the feeding tube unchanged. As a consequence, the material recirculation gap size was changed, and the size, shape, position, and rotational direction of the vortices formed in the secondary classification space and classification chamber were also changed. These vortices significantly affected the classification performance. Specifically, the classifiers with different guide cone structures, named CHL1, CHL2, CHL3, and CHL4, yielded Newton efficiencies of 75.06%, 87.26%, 95.5%, and 94.02%, respectively. According to the simulation results, the best guide cone structure is recommended to satisfy objectives such as (i) the highest classification efficiency, the smallest cut size, and the smallest classification sharpness index and (ii) those in (i) under the constraint of the required KHC.

A systematic review of Automated prediction of sudden cardiac deathusing ECG signals.

Sudden Cardiac Death (SCD) stands as a life-threatening cardiac event capable&#xD;of swiftly claiming lives. Researchers have devised numerous models aimed at automatically predicting SCD&#xD;through a combination of diverse feature extraction techniques and classifiers.&#xD; We did a rigorous review of research publications ranging from 2011 to 2023, with&#xD;a specific focus on the automated prediction of SCD, a growing health concern on a global&#xD;scale. Over the past two decades, Machine Learning (ML) techniques have emerged and evolved for&#xD;this purpose. Notably, since 2021, Deep Learning (DL) technology has also been incorporated into automatically predicting SCD. &#xD; This literature review comprehensively analyzes ML and DL models&#xD;employed in predicting SCD. The analysis yielded valuable insights into the fundamental structure&#xD;of cardiac fatalities, extracting relevant characteristics from ECG and HRV signals, using&#xD;databases, and evaluating classifier performance. The review offers a succinct yet thorough&#xD;examination of automated SCD prediction methodologies, emphasizing current constraints and&#xD;underscoring the necessity for further advancements. It serves as a valuable resource, providing&#xD;valuable insights and outlining potential research directions for aspiring scholars in the domain&#xD;of SCD prediction.&#xD;These automated methodologies have demonstrated the potential to achieve remarkable prediction&#xD;accuracy, reaching levels as high as 97%, and can forecast SCD events with a lead time of 30-70&#xD;minutes. Despite these promising outcomes, the quest for even greater accuracy and reliability persists. &#xD;While ML and DL methodologies have shown great promise, their performance is&#xD;intrinsically linked to the volume of training data available. Most predictive models rely on small-scale databases, raising concerns about their applicability in real-world scenarios. Furthermore,&#xD;these models predominantly utilize electrocardiogram and heart rate variability signals, often&#xD;overlooking the potential contributions of other physiological signals. Developing real-time,&#xD;clinically applicable models also represents a critical avenue for further exploration in this field.