Detection Algorithm Research Articles

A Review of Deep Learning Based Object Detection Algorithms

With the rapid development of artificial intelligence technology, the traditional manual feature-based object detection method has been gradually replaced by deep learning-based object detection technology. Object detection is the basis and prerequisite for many computer vision tasks aimed at recognizing targets in an image and determining their class and location. The purpose of object detection algorithm research based on deep learning is to improve the detection accuracy and detection speed of the detection algorithm, so that the object detection algorithm can be more safe and convenient applied to People's Daily production and life. This article reviews the evolution of object detection algorithms, focusing on two-stage, one-stage and object detection algorithms based on the transformer architecture. In addition, the performance, advantages and disadvantages of different algorithms are compared, and the development trend of object detection algorithms based on deep learning is summarized in the end.

MAR-YOLOv9: A multi-dataset object detection method for agricultural fields based on YOLOv9

With the development of deep learning technology, object detection has been widely applied in various fields. However, in cross-dataset object detection, conventional deep learning models often face performance degradation issues. This is particularly true in the agricultural field, where there is a multitude of crop types and a complex and variable environment. Existing technologies still face performance bottlenecks when dealing with diverse scenarios. To address these issues, this study proposes a lightweight, cross-dataset enhanced object detection method for the agricultural domain based on YOLOv9, named Multi-Adapt Recognition-YOLOv9 (MAR-YOLOv9). The traditional 32x downsampling Backbone network has been optimized, and a 16x downsampling Backbone network has been innovatively designed. A more streamlined and lightweight Main Neck structure has been introduced, along with innovative methods for feature extraction, up-sampling, and Concat connection. The hybrid connection strategy allows the model to flexibly utilize features from different levels. This solves the issues of increased training time and redundant weights caused by the detection neck and auxiliary branch structures in traditional YOLOv9, enabling MAR-YOLOv9 to maintain high performance while reducing the model’s computational complexity and improving detection speed, making it more suitable for real-time detection tasks. In comparative experiments on four plant datasets, MAR-YOLOv9 improved the mAP@0.5 accuracy by 39.18% compared to seven mainstream object detection algorithms, and by 1.28% compared to the YOLOv9 model. At the same time, the model size was reduced by 9.3%, and the number of model layers was decreased, reducing computational costs and storage requirements. Additionally, MAR-YOLOv9 demonstrated significant advantages in detecting complex agricultural images, providing an efficient, lightweight, and adaptable solution for object detection tasks in the agricultural field. The curated data and code can be accessed at the following link: https://github.com/YangxuWangamI/MAR-YOLOv9.

AIoT-Based Visual Anomaly Detection in Photovoltaic Sequence Data via Sequence Learning

Anomaly detection is a common analytical task aimed at identifying rare cases that differ from the majority of typical cases in a dataset. In the management of photovoltaic (PV) power generation systems, it is essential for electric power companies to effectively detect anomalies in PV sequence data, as this helps operators and experts understand and interpret anomalies within PV arrays when making response decisions. However, traditional methods that rely on manual labor and regular data collection are difficult to monitor in real time, resulting in delays in fault detection and localization. Traditional machine learning algorithms are slow and cumbersome in processing data, which affects the operational safety of PV plants. In this paper, we propose a visual analytic approach for detecting and exploring anomalous sequences in a PV sequence dataset via sequence learning. We first compare the sequences with their reconstructions through an unsupervised anomaly detection algorithm (Long Short-Term Memory) based on AutoEncoders to identify anomalies. To further enhance the accuracy of anomaly detection, we integrate the artificial intelligence of things (AIoT) technology with a strict time synchronization data collection and real-time processing algorithm. This integration ensures that data from multiple sensors are synchronized and processed in real time. Then, we analyze the characteristics of the anomalies based on the visual comparison of different PV sequences and explore the potential correlation factors to analyze the possible causes of the anomalies. Case studies based on authentic enterprise datasets demonstrate the effectiveness of our method in the anomaly detection and exploration of PV sequence data.

Rail Detection Research Status of Multi -Sensing Technology

Traditional rail inspection methods, primarily relying on manual inspection and basic equipment, face significant challenges including low efficiency, poor accuracy, and susceptibility to human error. This paper presents a comprehensive review of the current research status and future trends in multi-sensor rail transit inspection vehicles. The focus is on the integrated application of various sensing technologies, particularly inertial systems and visual sensing, to achieve high-precision and efficient measurement of rail geometrical parameters. An in-depth analysis of key technologies and challenges in railway track inspection is provided, exploring how advanced sensing methods can overcome the limitations of traditional approaches. The paper also examines the market prospects for rail transit inspection, highlighting the growing demand for more sophisticated and reliable inspection systems. Furthermore, potential future research directions in this field are outlined, including the development of AI-powered data analysis, real-time defect detection algorithms, and the integration of emerging technologies such as LiDAR and thermal imaging. This review aims to provide valuable insights for researchers, engineers, and industry professionals working on advancing rail inspection technologies.

Fusing convolutional learning and attention-based Bi-LSTM networks for early Alzheimer’s diagnosis from EEG signals towards IoMT

The Internet of Medical Things (IoMT) is poised to play a pivotal role in future medical support systems, enabling pervasive health monitoring in smart cities. Alzheimer’s disease (AD) afflicts millions globally, and this paper explores the potential of electroencephalogram (EEG) data in addressing this challenge. We propose the Convolutional Learning Attention-Bidirectional Time-Aware Long-Short-Term Memory (CL-ATBiLSTM) model, a deep learning approach designed to classify different AD phases through EEG data analysis. The model utilizes Discrete Wavelet Transform (DWT) to decompose EEG data into distinct frequency bands, allowing for targeted analysis of AD-related brain activity patterns. Additionally, the data is segmented into smaller windows to handle the dynamic nature of EEG signals, and these segments are transformed into spectrogram images, visually depicting brain activity distribution over time and frequency. The CL-ATBiLSTM model incorporates convolutional layers to capture spatial features, attention mechanisms to emphasize crucial data, and BiLSTM networks to explore temporal relationships within the sequences. To optimize the model’s performance, Bayesian optimization is employed to fine-tune the hyperparameters of the ATBiLSTM network, enhancing its ability to generalize and accurately classify AD stages. Incorporating Bayesian learning ensures the most effective model configuration, improving sensitivity and specificity for identifying AD-related patterns. Our model extracts discriminative features from EEG data to differentiate between AD, Mild Cognitive Impairment (MCI), and healthy controls (CO), offering a more comprehensive approach than existing two-class detection algorithms. By including the MCI category, our method facilitates earlier identification and potentially more impactful therapy interventions. Achieving a 96.52% accuracy on Figshare datasets containing AD, MCI, and CO groups, our approach demonstrates strong potential for practical use, accelerating AD identification, enhancing patient care, and contributing to the development of targeted treatments for this debilitating condition.

Pattern detection in the vehicular activity of bus rapid transit systems

In this paper, we explore different methods to detect patterns in the activity of bus rapid transit (BRT) systems focusing on two aspects of transit: infrastructure and the movement of vehicles. To this end, we analyze records of velocity and position of each active vehicle in nine BRT systems located in the Americas. We detect collective patterns that characterize each BRT system obtained from the statistical analysis of velocities in the entire system (global scale) and at specific zones (local scale). We analyze the velocity records at the local scale applying the Kullback-Leibler divergence to compare the vehicular activity between zones. This information is organized in a similarity matrix that can be represented as a network of zones. The resulting structure for each system is analyzed using network science methods. In particular, by implementing community detection algorithms on networks, we obtain different groups of zones characterized by similarities in the movement of vehicles. Our findings show that the representation of the dataset with information of vehicles as a network is a useful tool to characterize at different scales the activity of BRT systems when geolocalized records of vehicular movement are available. This general approach can be implemented in the analysis of other public transportation systems.

Digital twin-enhanced robotic system for remote diesel engine assembly defect inspection

Purpose This study aims to streamline and enhance the assembly defect inspection process in diesel engine production. Traditional manual inspection methods are labor-intensive and time-consuming because of the complex structures of the engines and the noisy workshop environment. This study’s robotic system aims to alleviate these challenges by automating the inspection process and enabling easy remote inspection, thereby freeing workers from heavy fieldwork. Design/methodology/approach This study’s system uses a robotic arm to traverse and capture images of key components of the engine. This study uses anomaly detection algorithms to automatically identify defects in the captured images. Additionally, this system is enhanced by digital twin technology, which provides inspectors with various tools to designate components of interest in the engine and assist in defect checking and annotation. This integration facilitates smooth transitions from manual to automatic inspection within a short period. Findings Through evaluations and user studies conducted over a relatively long period, the authors found that the system accelerates and improves the accuracy of engine inspections. The results indicate that the system significantly enhances the efficiency of production processes for manufacturers. Originality/value The system represents a novel approach to engine inspection, leveraging robotic technology and digital twin enhancements to address the limitations of traditional manual inspection methods. By automating and enhancing the inspection process, the system offers manufacturers the opportunity to improve production efficiency and ensure the quality of diesel engines.

Automated cardiac arrest DETECTion incorporated into a wristband: validation of a photoplethysmography algorithm in patients with spontaneous and induced cardiac arrest

Abstract Background Although survival from out-of-hospital cardiac arrest (OHCA) has improved, this does not apply to unwitnessed cases, as the arrest is often recognized too late. Automated cardiac arrest detection and alarming using a wristband is a promising technology to facilitate early assistance after cardiac arrest in a home setting. In the DETECT-1 study, a photoplethysmography(PPG) algorithm for circulatory arrest detection has been developed in patients with induced circulatory arrest. Purpose To assess the performance of the developed PPG algorithm for circulatory arrest detection in: 1) patients with spontaneous cardiac arrest; 2) subjects with induced circulatory arrest using blood pressure cuff inflation. Methods We selected patients with episodes of spontaneous cardiac arrest who participated in the DETECT-1 study and patients/healthy subjects with short-lasting induced circulatory arrest using blood pressure cuff inflations from the DETECT-1 and DETECT-2 studies. The DETECT-1 study included patients who underwent transcatheter aortic valve implantation (TAVI), subcutaneous ICD implantation, or ventricular tachycardia ablation. Included patients wore the PPG-wristband during the procedure. Continuous ECG and invasive blood pressure data were collected as reference standard. The DETECT-2 study included healthy volunteers who wore the PPG-wristband during physical exercise. Sensitivity for spontaneous and induced circulatory arrest and false positive alarms were assessed using the previously developed algorithm for circulatory arrest detection. All episodes were analyzed as individual events. Results A total of 39 individuals were included; nine patients with spontaneous cardiac arrests during TAVI and 20 subjects with 116 induced circulatory arrests. Of the patients with spontaneous cardiac arrest, two had ventricular fibrillation, five had pulseless electrical activity (figure), and two had asystole. Eight out of nine cardiac arrest events were detected by the circulatory arrest detection algorithm (sensitivity 89% [95% CI 51-99%]). Of the subjects with induced circulatory arrests, the PPG algorithm detected 114 induced circulatory arrests, yielding a sensitivity of 98% (95% CI 93-100%). In the entire group, there were no false positive alarms during 60 recording hours. Conclusions Automated cardiac arrest detection using a PPG-wristband is feasible with excellent sensitivity for both spontaneous and induced circulatory arrest. False positive alarms were absent in this controlled setting. Further study is warranted to assess algorithm performance in a larger sample of spontaneous cardiac arrests, and false positive alarms during daily life use. If available, automated cardiac arrest detection and alarming has the potential to markedly increase survival from OHCA in a home setting.Episode of spontaneous cardiac arrest

Automated cardiac arrest detection using a photoplethysmography wristband: algorithm performance in relation to sex, skin color type and arm hair density

Abstract Background While survival from witnessed out-of-hospital cardiac arrest (OHCA) has improved, this is not the case for unwitnessed cardiac arrest as the event goes unnoticed. Automated cardiac arrest detection and alerting through a wristband is considered a promising technique also facilitating early assistance for this large group. In the DETECT-1 study, a photoplethysmography(PPG) algorithm for circulatory arrest detection was developed in patients with induced circulatory arrest. However, patient characteristics, such as skin color and arm hair density, can influence the accuracy of the PPG signal. Thus, we performed further analyses to study the effect of patient characteristics on the algorithm’s performance. Purpose To assess the performance of the developed PPG algorithm for cardiac arrest detection in relation to sex, skin color type and arm hair density. Methods Patients who underwent short-lasting circulatory arrest induction during transcatheter aortic valve implantation, ventricular tachycardia ablation or subcutaneous ICD implantation wore a PPG wristband during the procedure. Continuous ECG and invasive blood pressure recordings served as reference standards. A PPG algorithm for circulatory arrest detection was developed. The algorithm’s performance was assessed in subgroups based on sex, skin color type (according to the Fitzpatrick scale), and arm hair density. The sensitivity for the detection of circulatory arrest and false positive alarms were compared between the different subgroups. Results In total, 151 patients were included in the current analysis. Analyzed data consisted of 159 hours of PPG data including 176 induced circulatory arrest events. The population consisted of 57 (38%) females and 94 (62%) males. The sensitivity for circulatory arrest detection was 100% in the female group vs. 98% in the male group (p=ns); false positive alarms occurred in 4 females and 9 males, resulting in a positive predictive value (PPV) of 94% and 92% (p=ns), respectively. Within the skin color subgroup analysis, 114 (75%) patients had skin color type I or II, and 37 (25%) patients had a skin color type III, IV, V or VI. There were no differences in sensitivity (p=ns) and PPV (p=ns) in the skin color subgroup (table 1). The population consisted of 106 (71%) patients with nil or sparse arm hair density, 43 (29%) patients with moderate or dense arm hair density, and 2 not reported. There were no differences in sensitivity (p=ns) and PPV (p=ns) within the arm hair density subgroup (table 1). Conclusions Automated cardiac arrest detection using a smartwatch is feasible and its accuracy is not affected by sex, skin color type or arm hair density in this study. Further study is needed to assess performance of the algorithm in a larger sample size and a more diverse study population. If proven to be effective, automated cardiac arrest detection has the potential to improve survival from OHCA.Sensitivity/PPV for arrest detection

Eyes on nature: Embedded vision cameras for terrestrial biodiversity monitoring

Abstract We need comprehensive information to manage and protect biodiversity in the face of global environmental challenges, and artificial intelligence is required to generate that information from vast amounts of biodiversity data. Currently, vision‐based monitoring methods are heterogenous; they poorly cover spatial and temporal dimensions, overly depend on humans, and are not reactive enough for adaptive management. To mitigate these issues, we present a portable, modular, affordable and low‐power device with embedded vision for biodiversity monitoring of a wide range of terrestrial taxa. Our camera uses interchangeable lenses to resolve barely visible and remote targets, as well as customisable algorithms for blob detection, region‐of‐interest classification and object detection to automatically identify them. We showcase our system in six use cases from ethology, landscape ecology, agronomy, pollination ecology, conservation biology and phenology disciplines. Using the same devices with different setups, we discovered bats feeding on durian tree flowers, monitored flying bats and their insect prey, identified nocturnal insect pests in paddy fields, detected bees visiting rapeseed crop flowers, triggered real‐time alerts for waterfowl and tracked flower phenology over months. We measured classification accuracies (i.e. F1‐scores) between 55% and 95% in our field surveys and used them to standardise observations over highly resolved time scales. Our cameras are amenable to situations where automated vision‐based monitoring is required off the grid, in natural and agricultural ecosystems, and in particular for quantifying species interactions. Embedded vision devices such as this will help addressing global biodiversity challenges and facilitate a technology‐aided agricultural systems transformation.

Data enhanced YOLOv8s algorithm for X-ray weld defect detection

Data enhanced YOLOv8s algorithm for X-ray weld defect detection

Artificial intelligence-enabled electrocardiogram for left ventricular diastolic dysfunction and long-term risk of new-onset atrial fibrillation

Abstract Background Heart failure (HF) and atrial fibrillation (AF) are intertwined, sharing common risk factors and influencing each other in a reciprocal manner. While left ventricular diastolic dysfunction (LVDD) stands as a hallmark of HF's hemodynamic profile, its direct correlation with heightened AF risk remains uncertain. Our recent validation of an artificial intelligence-enabled electrocardiogram (AI-ECG) algorithm for LVDD detection has opened new avenues. Yet, whether AI-ECG-detected LVDD correlates with long-term AF risk remains unexplored. Methods We performed a retrospective study among all patients with a comprehensive LVDD assessment from the test population of the AI-ECG LVDD study between September 2001 and June 2022. Patients with prior documented AF according to health records or previous ECG were excluded. Patients were classified by AI-ECG as normal diastolic function (DF), grade 1 (G1), grade 2 (G2), or grade 3 (G3) LVDD. We assessed the risk of AF across AI-ECG LVDD grade, with normal DF as reference, with competing risk models. Results Of 97,023 patients free of AF at baseline, 63,533 (65.4%), 10,776 (11.1%), 18,940 (19.5%), and 3,774 (3.9%) were classified as normal DF, G1-LVDD, G2-LVDD, and G3-LVDD, respectively. Worse AI-ECG LVDD was associated with older age, lower left ventricular ejection fraction (LVEF), and higher rates of HF, diabetes, hypertension, chronic pulmonary disease, renal failure, and cardiovascular disease (p-of-trend<0.001 for all). Over a median follow-up of 4.8 years (interquartile range 1.8-9.2 years), new-onset AF occurred in 6,795 (7.0%) patients. In multivariable survival analysis, adjusted to multiple risk factors including LVEF and clinical HF, G2 and G3 LVDD, representing patients with high filling pressures, were independently associated with a higher AF-risk [G2-LVDD: adjusted hazard ratio (aHR) 1.64 (95% confidence interval (95%CI) 1.52-1.73), G3-LVDD: aHR 2.24 (95%CI 2.02-2.48)]. Conclusion The AI-ECG LVDD algorithm independently predicts long-term AF risk, regardless of LVEF and clinical HF status, highlighting its potential for identifying high-risk populations and guiding proactive monitoring and management.

Performance of apple watch for detecting atrial fibrillation among a known atrial fibrillation population

Abstract Background Photoplethysmography (PPG) based consumer accessible devices with algorithms for irregular rhythm detection are gaining popularity as an opportunistic screening tool for atrial arrhythmias, and are FDA cleared for use in patients without a history of atrial fibrillation (AF). However, these devices are increasingly used for the management of AF, a purpose for which some algorithms are not intended. Purpose To compare the performance of the Apple Watch Irregular Rhythm Notification (IRN) with an implantable cardiac monitor (ICM) for the detection of atrial fibrillation (AF) in patients with known AF from the DEFINE AFib study. Methods Patients with known AF, a LINQ or LINQ II ICM, and an Apple Watch with PPG-based IRN activated were included in the analysis. ICM detected AF episodes served as the standard to which Apple Watch IRNs were compared. All AF episodes detected by the ICM device were manually adjudicated by two trained, independent reviewers with a third independent reviewer in situations of disagreement. Only AF episodes ≥60 minutes during a 75-minute window in which the Apple Watch was monitoring were included in accordance with the known Apple Watch algorithm detection parameters. To eliminate oversampling of ICM detected episodes compared to Apple Watch IRNs, 1,000 simulations of one randomly selected episode per subject was done to avoid single-subject bias. Results In total, 97 subjects were analyzed of which 42 had an AF episode ≥60 minutes. Applying the 75-minute window of Apple Watch monitoring to episodes ≥60 minutes, 20 subjects had qualifying, adjudicated ICM detected AF. These 20 subjects accounted for 481 qualifying AF episodes. Of the ICM detected episodes, 191 (40%) occurred while an Apple Watch was not being worn. In total, 11/20 (55%) subjects had an IRN generated by the Apple Watch. Adjudication of the ICM episodes resulted in 290 true positive AF episodes ≥60 minutes during a 75-minute window while the watch was actively monitoring, with 74 (26%) of those episodes triggering an IRN resulting in a per episode sensitivity of 24 ± 7% and PPV of 91 ± 4% (Figure 1A). The efficacy of Apple Watch to detect AF episodes of differing lengths was assessed for episodes between 1-6 hours, 6-12 hours, 12-24 hours, and > 24 hours with sensitivities of 19 ± 19%, 38 ± 8%, 32 ± 10%, and 31 ± 8%, respectively (Figure 1B). The Apple Watch generated an annualized false positive rate of 2.25 IRNs for every patient year of monitoring. At the subject level, Apple Watch had a sensitivity of 48%, specificity of 100%, PPV of 100%, and NPV of 71% (Figure 1A). Conclusions In patients with a history of AF, the Apple Watch IRN was most sensitive at the individual subject level while watch performance was lower by episode. The current IRN algorithm favors longer episode durations. The limited sensitivity, dependence on user wear-time, and reliance on long-durations of AF limit the use of Apple Watch for AF management.

Improved algorithm for fracture-dissolution pore detection in resistivity imaging logging based on dung beetle optimization

Abstract Resistivity imaging logging has become a direct and precise method for visualizing the structural complexities of reservoir fractures and dissolution pores. The current use of Otsu's thresholding for segmentation results in poor segmentation quality and significant noise. Accurate segmentation of sub-images containing fracture and dissolution pore targets is essential for automated structure identification and subsequent parameter calculation. This study leverages the rapid convergence and robust global optimization capabilities of the Dung Beetle Optimizer to develop enhanced image segmentation approaches. Specifically, it introduces a refined K-Means algorithm for multi-category image segmentation and an Otsu algorithm for multi-threshold image segmentation, both optimized by the Dung Beetle Optimizer. Compared to conventional binary segmentation algorithms, this new algorithm effectively isolates noise and extracts multi-category information. Using the segmented sub-images, this paper integrates mathematical morphology techniques to compute parameters such as area, perimeter, tortuosity length, and pore shape factor for identified targets. Additionally, Principal Component Analysis is used to derive recognition factors for fractures and dissolution pores. Applications show that this factor can identify matrix, fracture, and dissolution pore targets in complex background images. By combining parameter information of the target area, the method effectively removes false information in resistivity imaging and segments sub-images of fractures and dissolution pores, calculating fracture area ratio, dissolution pore area ratio, and total area ratio.

Experiments for the Assessment of the Probabilistic Multistage Algorithm for Damage Detection in Flat and Curved CFRP Panels

The Probabilistic Multistage (PM) algorithm was developed by authors for identifying and localizing damages in isotropic plates. More specifically, PM algorithm consists of a fully automated probabilistic damage imaging methodology based on ultrasonic guided wave propagation and a 5-sensor array working in a pitch-catch approach. The algorithm processes the gathered data in three steps to identify key features associated with damage. The aim of this work is to assess the effectiveness of the PM algorithm on more complex structures. Specifically, the study investigates three cases of study, made of Carbon Fiber Reinforced Plastic (CFRP) composite, characterized by different geometries and layups. The algorithm is initially tested on panels with artificial damage in the form of a Teflon disk located in a specific location between the middle laminae of the panels, which is used to replicate the effect of delamination. In order to expand the experimental dataset without incurring additional costs or waste, new damage conditions are simulated by adding masses on the upper surface of the panels. Each plate is investigated considering three different damage sizes and 16 different damage locations. The proposed algorithm successfully detects damages both within and outside the sensor network. The PM algorithm produces a clear damage positioning map and a positioning (probabilistic field) range for the identified damage. This information can be used to assist operators in conducting inspections more efficiently by focusing on the highlighted areas, which may potentially lead to reduced maintenance and repair expenses.

Tightly coupled stereo vision-inertial odometry based on point and line feature

Abstract To improve the accuracy and robustness of visual simultaneous localization and mapping (SLAM) in low-texture environments, this paper proposes a robust and fast stereo vision inertial SLAM pose estimation method that combines point and line features with an inertial measurement unit (IMU). The method tightly couples visual point and line features with IMU constraints, forming a least-squares problem through the error of each constraint term for nonlinear optimization. To address the issues of over-segmentation and time consumption in traditional line segment detection (LSD) algorithms, an improved LSD algorithm is adopted to accelerate line feature extraction. This approach merges nearby line segments based on spatial geometric relationships and filters out invalid segments, improving the time efficiency of the algorithm. Finally, experiments conducted in low-texture environments demonstrate that our algorithm achieves high localization accuracy and robustness.

DaE2: Unmasking malicious URLs by leveraging diverse and efficient ensemble machine learning for online security

Over 5.44 billion people now use the Internet, making it a vital part of daily life, enabling communication, e-commerce, education, and more. However, this huge Internet connectivity also raises concerns about online privacy and security, particularly with the rise of malicious Uniform Resource Locators (URLs). Recently, conventional ensemble models have attracted attention due to their notable benefits of reducing the variance in models, enhancing predictive performance, improving prediction accuracy, and demonstrating high generalization potential. But, its application in addressing the challenge of malicious URLs is still an open problem. These URLs often hide behind static links in emails or web pages, posing a threat to individuals and organizations. Despite blacklisting services, many harmful sites evade detection due to inadequate scrutiny or recent creation. Hence, to improve URL detection, a Diverse and Efficient Ensemble (DaE2) machine learning algorithm was developed using four ensemble models, that is, AdaBoost, Bagging, Stacking, and Voting to classify URLs. After preprocessing, the experimental result shown that all models achieved over 80 % accuracy, with AdaBoost reaching 98.5 % and Stacking offering the fastest runtime. AdaBoost and Bagging also delivered strong performance, with F1 scores of 0.980 and 0.976, respectively.

Regional microbial content of fermented traditional and industrial East Mediterranean sausages from the islands of Cyprus and Mytilini

Objective: To characterize the microbial biodiversity of fermented sausages from the East Mediterranean islands of Cyprus and Mytilini, and to identify differences between the microbial diversity of traditionally and industrially produced Cypriot sausages. Method: The microbial diversity of thirty sausages from Cyprus and Mytilini (traditionally and industrially produced) was analyzed using high throughput sequencing (HTS) (amplicon sequencing) of 16S rRNA gene and ITS loci fragments. By applying microbial signature detection and machine learning algorithms, we identified key microbes that distinguish traditionally produced sausages from those produced industrially. Focusing on selected microbial taxa and using interaction network analysis, we identified associations among the sausages’ microbiota that may affect the shaping of the sausages’ microbial consortia. Results: Cypriot and Mytilini sausages indicated increased relative representation of Lactobacillus and Leuconostoc . Cypriot sausages were distinguished by the presence of the fungi Debaryomyces hansenii and Candida spp. , whereas Mytilini sausages by the bacteria Lactococcus and Streptococcus . Traditionally produced sausages from the Pitsilia region of Cyprus were differentiated by the presence of the species Lactobacillus helveticus , whereas industrially produced sausages were differentiated by the species Leuconostoc mesenteroides . Focusing mainly on Lactobacillus and Leuconostoc , negative associations with pathogenic bacteria, such as Salmonella , and spoilage fungi, such as Fusarium poae , were revealed. Conclusion: The present metataxonomic analysis provided insights into the microbial communities that characterize Cypriot and Mytilini sausages. The findings provide an indication that the microbial diversity might be applied as an additional marker of Cypriot sausages’ authenticity.

Novel 3D-ECG algorithms for reliable screening of cardiac amyloidosis

Abstract Background Cardiac amyloidosis is associated with high morbidity and mortality. Early detection is crucial to allow for specific therapy and improvement of patient’s prognosis. However, there is no established screening tool for cardiac amyloidosis, leading to a delay in diagnosis in the majority of patients. Aims We aimed to develop and validate a screening tool for cardiac amyloidosis based on structured evaluation of 3-dimensional electrocardiograms (ECGs). Methods We included patients from a vascular center with confirmed cardiac AL- or ATTR-amyloidosis as well as controls of patients with other cardiovascular diseases but without amyloidosis into two independent cohorts, a derivation and validation cohort. All patients received 3-dimensional ECGs and vector loops were categorized based on pre-defined patterns by two independent cardiologists. Consecutively, an AI algorithm was trained in the derivation cohort (n=66 amyloidosis cases, n=89 controls). This algorithm was then applied to the validation cohort (n=33 amyloidosis cases, n=67 controls). Overall accuracy of the algorithm for detection of amyloidosis was evaluated in both cohorts. Results Overall, 99 patients with amyloidosis and 156 controls were included (mean age: 75±24 years, 82% male). In the derivation cohort, the AI algorithm reached a sensitivity of 85%, a specificity of 89%, a positive predictive value of 91%, and a negative predictive value of 87%. Applying the algorithm on the independent validation cohort, a sensitivity of 79%, specificity of 82%, a positive predictive value of 61%, and a negative predictive value of 92% was reached. The overall accuracy of the algorithm for detection of cardiac amyloidosis was 85%. Conclusion We here describe a novel screening tool, which allows for reliable detection of cardiac amyloidosis.

Artificial intelligence for automated left ventricular systolic dysfunction detection using a wearable cardiac patch incorporated with synchronized phonocardiogram and electrocardiogram

Abstract Background Left ventricular systolic dysfunction (LVSD) is characterized by a reduced left ventricular ejection fraction (LVEF) and is associated with three times the risk of developing Heart failure (HF). We aimed to test an artificial intelligence (AI) algorithm applied to synchronized phonocardiography (PCG) and electrocardiography (ECG) signals, recorded with a wearable device, to validate its potential as a screening tool for LVSD. Methods Using ECG and PCG data collected from 1960 admitted patients in a hospital, we trained an AI algorithm to detect the LVSD. There are two ways of defining LVSD, one group is defined by LVEF <50% and the other is defined by LVEF 40% as measured by echocardiography. The AI algorithm followed a two-step detection structure. In the first phase, ECG signals were processed by wavelet denoising and baseline wander correction, and PCG signals were converted to the frequency domain via wavelet transformation. We developed a 1D CNN-based U-Net variant model to pinpoint electromechanical activation time (EMAT). Addressing the challenge of processing multimodal signals, we devised a contrastive learning approach that mandates the encoders for different modalities to generate embeddings within the same feature space. This trained encoder, along with the decoder in U-Net, is subsequently fine-tuned for EMAT detection. The AI-EMAT% was obtained by adjusting detected EMAT based on RR interval. For the second phase, random forest regression is used to estimate LVEF, incorporating features such as AI-EMAT% and clinical data, including age and gender. For LVSD detection, we evaluated the performance of methods by bifurcating AI-estimated LVEF% (AI-predict LVEF%) based on thresholds of 40% and 50%, and by directly estimating AI-EMAT% cutoff. The evaluation of LVSD detection algorithm, conducted via 2-fold cross-validation, involves dividing the dataset into two subsets, with each alternately serving as training and testing sets to ensure a fair assessment. Results We recruited 1960 patients (1430 [72.96%] male). 357 (18.21%) had an ejection fraction of 50% or lower, and 157 (8.01%) had an ejection fraction of 40% or lower. In detecting LVSD of LVEF<50%, AI-EMAT% yielded an AUROC of 0.86, sensitivity of 73.7%, and specificity of 80.7%. While AI-predict LVEF% approach resulted in an AUROC of 0.89, sensitivity of 81.8%, and specificity of 81.5%. In detecting LVSD of LVEF<40%, AI-EMAT% yielded an AUROC of 0.89, sensitivity of 85.4%, and specificity of 76.0%. While AI-predict LVEF% outputs resulted in an AUROC of 0.91, sensitivity of 92.4%, and specificity of 76.8%. Conclusions We developed an automated algorithm to detect LVSD using a wearable device that incorporated synchronized PCG and ECG signals, demonstrating robust performance across different subgroups. This approach represents a strategy for automated screening of LV systolic dysfunction, particularly beneficial in resource-limited settings.Baseline and AUROCAI algorithm