False Positive Alarms Research Articles

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

Network anomaly detection in cars: A case for time-sensitive stream filtering and policing

Connected vehicles are threatened by cyber-attacks as in-vehicle networks technologically approach (mobile) LANs with several wireless interconnects to the outside world. Malware that infiltrates a car today faces potential victims of constrained, barely shielded Electronic Control Units (ECUs). Many ECUs perform critical driving functions, which stresses the need for hardening security and resilience of in-vehicle networks in a multifaceted way. Future vehicles will comprise Ethernet backbones that differentiate services via Time-Sensitive Networking (TSN). The well-known vehicular control flows will follow predefined schedules and TSN traffic classifications. In this paper, we exploit this traffic classification to build a network anomaly detection system. We show how filters and policies of TSN can identify misbehaving traffic and thereby serve as distributed guards on the data link layer. On this lowest possible layer, our approach derives a highly efficient network protection directly from TSN. We classify link layer anomalies and micro-benchmark the detection accuracy in each class. Based on a topology derived from a real-world car and its traffic definitions we evaluate the detection system in realistic macro-benchmarks based on recorded attack traces. Our results show that the detection accuracy depends on how exact the specifications of in-vehicle communication are configured. Most notably for a fully specified communication matrix, our anomaly detection remains free of false-positive alarms, which is a significant benefit for implementing automated countermeasures in future vehicles.

Pursuing drug laboratories: Analysis of drug precursors with High Kinetic Energy Ion Mobility Spectrometry

High Kinetic Energy Ion Mobility Spectrometry (HiKE-IMS) is a technique for rapid and reliable detection of trace compounds down to ppbV-levels within one second. Compared to classical IMS operating at ambient pressure and providing the ion mobility at low electric fields, HiKE-IMS can also provide the analyte-specific field dependence of the ion mobility and a fragmentation pattern at high reduced electric field strengths. The additional information about the analyte obtained by varying the reduced electric field strength can contribute to reliable detection. Furthermore, the reduced number of ion-molecule reactions at the low operating pressure of 10 – 40 mbar and the shorter reaction times reduce the impact of competing ion-molecule reactions that can cause false negatives. In this work, we employ HiKE-IMS for the analysis of phenyl-2-propanone (P2P) and other precursor chemicals used for synthesis of methamphetamine and amphetamine. The results show that the precursor chemicals exhibit different behavior in HiKE-IMS. Some precursors form a single significant ion species, while others readily form a fragmentation pattern. Nevertheless, all drug precursors can be distinguished from each other, from the reactant ions and from interfering compounds. In particular, the field-dependent ion mobility as an additional separation dimension aids identification, potentially reducing the number of false positive alarms in field applications. Furthermore, the analysis of a seized illicit P2P sample shows that even low levels of P2P can be detected despite the complex background present in the headspace of real samples.

Improving sepsis prediction in intensive care with SepsisAI: A clinical decision support system with a focus on minimizing false alarms.

Prediction of sepsis using machine-learning approaches has recently gained traction. However, the lack of translation of these algorithms into clinical routine remains a major issue. Existing early sepsis detection methods are either based on the older definition of sepsis or do not accurately detect sepsis leading to the high frequency of false-positive alarms. This results in a well-known issue of clinicians' "alarm fatigue", leading to decreased responsiveness and identification, ultimately resulting in delayed clinical intervention. Hence, there is a fundamental, unmet need for a clinical decision system capable of accurate and timely sepsis diagnosis, running at the point of need. In this work, SepsisAI-a deep-learning algorithm based on long short-term memory (LSTM) networks was developed to predict the early onset of hospital-acquired sepsis in real-time for patients admitted to the ICU. The models are trained and validated with data from the PhysioNet Challenge, consisting of 40,336 patient data files from two healthcare systems: Beth Israel Deaconess Medical Center and Emory University Hospital. In the short term, the algorithm tracks frequently measured vital signs, sparsely available lab parameters, demographic features, and certain derived features for making predictions. A real-time alert system, which monitors the trajectory of the predictions, is developed on top of the deep-learning framework to minimize false alarms. On a balanced test dataset, the model achieves an AUROC, AUPRC, sensitivity, and specificity of 0.95, 0.96, 88.19%, and 96.75%, respectively at the patient level. In terms of lookahead time, the model issues a warning at a median of 6 hours (IQR 6 to 20 hours) and raises an alert at a median of 4 hours (IQR 2 to 5 hours) ahead of sepsis onset. Most importantly, the model achieves a false-alarm ratio of 3.18% for alerts, which is significantly less than other sepsis alarm systems. Additionally, on a disease prevalence-based test set, the algorithm reported similar outcomes with AUROC and AUPRC of 0.94 and 0.87, respectively, with sensitivity, and specificity of 97.05%, and 96.75%, respectively. The proposed algorithm might serve as a clinical decision support system to assist clinicians in the accurate and timely diagnosis of sepsis. With exceptionally high specificity and low false-alarm rate, this algorithm also helps mitigate the well-known issue of clinician alert fatigue arising from currently proposed sepsis alarm systems. Consequently, the algorithm partially addresses the challenges of successfully integrating machine-learning algorithms into routine clinical care.

Reliable Detection of Chemical Warfare Agents Using High Kinetic Energy Ion Mobility Spectrometry.

High Kinetic Energy Ion Mobility Spectrometers (HiKE-IMS) ionize and separate ions at reduced pressures of 10-40 mbar and over a wide range of reduced electric field strengths E/N of up to 120 Td. Their reduced operating pressure is distinct from that of conventional drift tube ion mobility spectrometers that operate at ambient pressure for trace compound detection. High E/N can lead to a field-induced fragmentation pattern that provides more specific structural information about the analytes. In addition, operation at high E/N values adds the field dependence of ion mobility as an additional separation dimension to low-field ion mobility, making interfering compounds less likely to cause a false positive alarm. In this work, we study the chemical warfare agents tabun (GA), sarin (GB), soman (GD), cyclosarin (GF) and sulfur mustard (HD) in a HiKE-IMS at variable E/N in both the reaction and the drift region. The results show that varying E/N can lead to specific fragmentation patterns at high E/N values combined with molecular signals at low E/N. Compared to the operation at a single E/N value in the drift region, the variation of E/N in the drift region also provides the analyte-specific field dependence of ion mobility as additional information. The accumulated data establish a unique fingerprint for each analyte that allows for reliable detection of chemical warfare agents even in the presence of interfering compounds with similar low-field ion mobilities, thus reducing false positives.

A novel intrusion detection system based on a hybrid quantum support vector machine and improved Grey Wolf optimizer

The Internet of Things (IoT) has grown significantly in recent years, allowing devices with sensors to share data via the internet. Despite the growing popularity of IoT devices, they remain vulnerable to cyber-attacks. To address this issue, researchers have proposed the Hybrid Intrusion Detection System (HIDS) as a way to enhance the security of IoT. This paper presents a novel intrusion detection model, namely QSVM-IGWO, for improving the detection capabilities and reducing false positive alarms of HIDS. This model aims to improve the performance of the Quantum Support Vector Machine (QSVM) by incorporating parameters from the Improved Grey Wolf Optimizer (IGWO) algorithm. IGWO is introduced under the hypothesis that the social hierarchy observed in grey wolves enhances the searching procedure and overcomes the limitations of GWO. In addition, the QSVM model is employed for binary classification by selecting the kernel function to obtain an optimal solution. Experimental results show promising performance of QSVM-IGWO in terms of accuracy, Recall, Precision, F1 score, and ROC curve, when compared with recent detection models.

Trust-Based Optimized Reporting for Detection and Prevention of Black Hole Attacks in Low-Power and Lossy Green IoT Networks.

The Internet of Things (IoT) is empowering various sectors and aspects of daily life. Green IoT systems typically involve Low-Power and Lossy Networks (LLNs) with resource-constrained nodes. Lightweight routing protocols, such as the Routing Protocol for Low-Power and Lossy Networks (RPL), are increasingly being applied for efficient communication in LLNs. However, RPL is susceptible to various attacks, such as the black hole attack, which compromises network security. The existing black hole attack detection methods in Green IoT rely on static thresholds and unreliable metrics to compute trust scores. This results in increasing false positive rates, especially in resource-constrained IoT environments. To overcome these limitations, we propose a delta-threshold-based trust model called the Optimized Reporting Module (ORM) to mitigate black hole attacks in Green IoT systems. The proposed scheme comprises both direct trust and indirect trust and utilizes a forgetting curve. Direct trust is derived from performance metrics, including honesty, dishonesty, energy, and unselfishness. Indirect trust requires the use of similarity. The forgetting curve provides a mechanism to consider the most significant and recent feedback from direct and indirect trust. To assess the efficacy of the proposed scheme, we compare it with the well-known trust-based attack detection scheme. Simulation results demonstrate that the proposed scheme has a higher detection rate and low false positive alarms compared to the existing scheme, confirming the applicability of the proposed scheme in green IoT systems.

Automated cardiac arrest detection using a photoplethysmography wristband: algorithm development and validation in patients with induced circulatory arrest in the DETECT-1 study

Unwitnessed out-of-hospital cardiac arrest is associated with low survival chances because of the delayed activation of the emergency medical system in most cases. Automated cardiac arrest detection and alarming using biosensor technology would offer a potential solution to provide early help. We developed and validated an algorithm for automated circulatory arrest detection using wrist-derived photoplethysmography from patients with induced circulatory arrests. In this prospective multicentre study in three university medical centres in the Netherlands, adult patients (aged 18 years or older) in whom short-lasting circulatory arrest was induced as part of routine practice (transcatheter aortic valve implantation, defibrillation testing, or ventricular tachycardia induction) were eligible for inclusion. Exclusion criteria were a known bilateral significant subclavian artery stenosis or medical issues interfering with the wearing of the wristband. After providing informed consent, patients were equipped with a photoplethysmography wristband during the procedure. Invasive arterial blood pressure and electrocardiography were continuously monitored as the reference standard. Development of the photoplethysmography algorithm was based on three consecutive training cohorts. For each cohort, patients were consecutively enrolled. When a total of 50 patients with at least one event of circulatory arrest were enrolled, that cohort was closed. Validation was performed on the fourth set of included patients. The primary outcome was sensitivity for the detection of circulatory arrest. Of 306 patients enrolled between March 14, 2022, and April 21, 2023, 291 patients were included in the data analysis. In the development phase (n=205), the first training set yielded a sensitivity for circulatory arrest detection of 100% (95% CI 94-100) and four false positive alarms; the second training set yielded a sensitivity of 100% (94-100), with six false positive alarms; and the third training set yielded a sensitivity of 100% (94-100), with two false positive alarms. In the validation phase (n=86), the sensitivity for circulatory arrest detection was 98% (92-100) and 11 false positive circulatory arrest alarms. The positive predictive value was 90% (95% CI 82-94). The automated detection of induced circulatory arrests using wrist-derived photoplethysmography is feasible with good sensitivity and low false positives. These promising findings warrant further development of this wearable technology to enable automated cardiac arrest detection and alarming in a home setting. Dutch Heart Foundation (Hartstichting).

Performance comparison between two computer-aided detection colonoscopy models by trainees using different false positive thresholds: a cross-sectional study in Thailand.

This study aims to compare polyp detection performance of "Deep-GI," a newly developed artificial intelligence (AI) model, to a previously validated AI model computer-aided polyp detection (CADe) using various false positive (FP) thresholds and determining the best threshold for each model. Colonoscopy videos were collected prospectively and reviewed by three expert endoscopists (gold standard), trainees, CADe (CAD EYE; Fujifilm Corp.), and Deep-GI. Polyp detection sensitivity (PDS), polyp miss rates (PMR), and false-positive alarm rates (FPR) were compared among the three groups using different FP thresholds for the duration of bounding boxes appearing on the screen. In total, 170 colonoscopy videos were used in this study. Deep-GI showed the highest PDS (99.4% vs. 85.4% vs. 66.7%, p<0.01) and the lowest PMR (0.6% vs. 14.6% vs. 33.3%, p<0.01) when compared to CADe and trainees, respectively. Compared to CADe, Deep-GI demonstrated lower FPR at FP thresholds of ≥0.5 (12.1 vs. 22.4) and ≥1 second (4.4 vs. 6.8) (both p<0.05). However, when the threshold was raised to ≥1.5 seconds, the FPR became comparable (2 vs. 2.4, p=0.3), while the PMR increased from 2% to 10%. Compared to CADe, Deep-GI demonstrated a higher PDS with significantly lower FPR at ≥0.5- and ≥1-second thresholds. At the ≥1.5-second threshold, both systems showed comparable FPR with increased PMR.

Executive Summary: Society of Critical Care Medicine Guidelines on Recognizing and Responding to Clinical Deterioration Outside the ICU.

Clinical deterioration of patients hospitalized outside the ICU is a source of potentially reversible morbidity and mortality. To address this, some acute care facilities have implemented systems aimed at detecting and responding to such patients. To provide evidence-based recommendations for hospital clinicians and administrators to optimize recognition and response to clinical deterioration in non-ICU patients. The 25-member panel included representatives from medicine, nursing, respiratory therapy, pharmacy, patient/family partners, and clinician-methodologists with expertise in developing evidence-based clinical practice guidelines. We generated actionable questions using the Population, Intervention, Control, and Outcomes format and performed a systematic review of the literature to identify and synthesize the best available evidence. We used the Grading of Recommendations Assessment, Development, and Evaluation approach to determine certainty in the evidence and to formulate recommendations and good practice statements (GPSs). The panel issued 10 statements on recognizing and responding to non-ICU patients with critical illness. Healthcare personnel and institutions should ensure that all vital sign acquisition is timely and accurate (GPS). We make no recommendation on the use of continuous vital sign monitoring among "unselected" patients due to the absence of data regarding the benefit and the potential harms of false positive alarms, the risk of alarm fatigue, and cost. We suggest focused education for bedside clinicians in signs of clinical deterioration, and we also suggest that patient/family/care partners' concerns be included in decisions to obtain additional opinions and help (both conditional recommendations). We recommend hospital-wide deployment of a rapid response team or medical emergency team (RRT/MET) with explicit activation criteria (strong recommendation). We make no recommendation about RRT/MET professional composition or inclusion of palliative care members on the responding team but suggest that the skill set of responders should include eliciting patients' goals of care (conditional recommendation). Finally, quality improvement processes should be part of a rapid response system (GPS). The panel provided guidance to inform clinicians and administrators on effective processes to improve the care of patients at-risk for developing critical illness outside the ICU.

Advanced Cybercrime Detection: A Comprehensive Study on Supervised and Unsupervised Machine Learning Approaches Using Real-world Datasets

In the ever-evolving field of cybersecurity, sophisticated methods—which combine supervised and unsupervised approaches—are used to tackle cybercrime. Strong supervised tools include Support Vector Machines (SVM) and K-Nearest Neighbors (KNN), while well-known unsupervised methods include the K-means clustering model. These techniques are used on the publicly available StatLine dataset from CBS, which is a large dataset that includes the individual attributes of one thousand crime victims. Performance analysis shows the remarkable 91% accuracy of SVM in supervised classification by examining the differences between training and testing data. K-Nearest Neighbors (KNN) models are quite good in the unsupervised arena; their accuracy in detecting criminal activity is impressive, at 79.56%. Strong assessment metrics, such as False Positive (FP), True Negative (TN), False Negative (FN), False Positive (TP), and False Alarm Rate (FAR), Detection Rate (DR), Accuracy (ACC), Recall, Precision, Specificity, Sensitivity, and Fowlkes–Mallow's scores, provide a comprehensive assessment.

An adaptive distributed intrusion detection system in local network: Hybrid classification methods

In the realm of cybersecurity, the incessant evolution of network attacks necessitates advanced and robust intrusion detection systems (IDS). The major issues with these systems are numerous: false positivenegative alarms, delayed response and detection time, size of processed data, adaptability to future threats, scalability of the system, difficulty in detecting distributed attacks, and downtime (fault tolerance). We propose a system that introduces a distributed framework aimed at enhancing network security by effectively identifying subtle deviations from normal network behavior. This is achieved through transfer learning based on artificial neural networks, and support vector machine (SVM), capitalizing on their complementary strengths in recognizing complex patterns and addressing high-dimensional datasets. To validate the efficacy of the proposed approach, the NSL-KDD dataset is utilized within a distributed IDS architecture. It consists of several intrusion detection nodes representing subnetworks. A node consists of two agents that work collaboratively. A way is proposed to avoid interference between analysis agents: the network agents manager monitors the functioning of the nodes and displays the results of each vulnerability-detecting node in each subnet separately. Such communication between agents should reduce FPAS (false positive alarms) significantly. The Detection engine extracts relevant features of network attacks to solve the problem of SVM in processing huge sizes of data and detect adaptive future threats to detect famous distributed denial of services (DDOS) attacks in real-time. The system is highly scalable by increasing the number of intrusion detection system nodes if necessary. Central processing is avoided to circumvent a system failure situation, where processing and decision-making take place at the detection node level within each subnet.

The operational and financial impact of adding anaerobic screening of platelets.

We evaluated the operational and safety impact of implementing anaerobic culture screening of apheresis and pooled platelets at the American Red Cross on the already established use of the aerobic culture screening of each donation performed no sooner than 24 h following collection. Platelets were screened for bacterial contamination with the BACT/ALERT 3D® (bioMérieux, Durham, NC) microbial detection testing system. The addition of anaerobic culture to the already existing aerobic culture resulted in sampling an additional 8-10 mL from each donation. Implementation of anaerobic testing resulted in an approximate 3.5-fold increased rate of False Positive BACT/ALERT alarms. There was a modest increase in the rate of True Positive alarms of 1.4-fold with increased detection of Klebsiella and Propionibacterium species, including Cutibacterium acnes. In addition, there was an approximate 3.5-fold increase rate of False Positives and a 13.5-fold increase rate of Indeterminates, the majority (~57%) were due to Cutibacterium acnes. The combined costs and lost revenue associated with adding anaerobic screening increased by ~$1,000,000/year due to testing cost and product discards. The addition of anaerobic culture to aerobic culture to the original donation (without the introduction of sampling delay) resulted in a significant increase in the rate of alerts. The 40% increased rate of True Positive alarms may have modestly improved platelet safety. However, there was a disproportionate increase in the rate of False Positive and Indeterminate bacterial culture alarms, which added substantial cost and overall loss of platelet products.

Ketone-Based Alert System for Insulin Pump Failures.

An increasing number of individuals with type 1 diabetes (T1D) manage glycemia with insulin pumps containing short-acting insulin. If insulin delivery is interrupted for even a few hours due to pump or infusion site malfunction, the resulting insulin deficiency can rapidly initiate ketogenesis and diabetic ketoacidosis (DKA). To detect an event of accidental cessation of insulin delivery, we propose the design of ketone-based alert system (K-AS). This system relies on an extended Kalman filter based on plasma 3-beta-hydroxybutyrate (BOHB) measurements to estimate the disturbance acting on the insulin infusion/injection input. The alert system is based on a novel physiological model capable of simulating the ketone body turnover in response to a change in plasma insulin levels. Simulated plasma BOHB levels were compared with plasma BOHB levels available in the literature. We evaluated the performance of the K-AS on 10 in silico subjects using the S2014 UVA/Padova simulator for two different scenarios. The K-AS achieves an average detection time of 84 and 55.5 minutes in fasting and postprandial conditions, respectively, which compares favorably and improves against a detection time of 193 and 120 minutes, respectively, based on the current guidelines. The K-AS leverages the rapid rate of increase of plasma BOHB to achieve short detection time in order to prevent BOHB levels from rising to dangerous levels, without any false-positive alarms. Moreover, the proposed novel insulin-BOHB model will allow us to understand the efficacy of treatment without compromising patient safety.

Sparse and semi-attention guided faults diagnosis approach for distributed online services

Despite the rapid advance of unsupervised reconstruction models in online service fault diagnosis, existing methods still lead to frequent false positive or false negative alarms. Tracing to the cause, popular reconstruction frameworks may fall into an “identical shortcut”, where both normal and anomalous cases can be well-recovered and hence faults are escaped from monitoring. Moreover, regularly “concept drift” deployed by online services always leads to sequences distribution dramatically change, which is easily wrongly-detected as anomalies. In this work, we present SS-Attention to avoid two contradictory false alarms simultaneously in a unified framework. Specifically, we first revisit the frameworks of MLP, CNN, and transformer, and confirm the important role of self-attention in preventing the model from learning shortcut, which we therefore select as our backbone. Second, we come up with a sparse attention layer to help model capture more long-term dependencies and distinguish local “concept drift” from anomalies. Third, we propose a semi-attention layer that urges model only learn from normal regions and further avoid well-recovering the faults. We evaluate our model on three public online service datasets and surpass the state-of-the-art alternatives by a sufficiently large margin, achieving an average 95.91% F1-Score with an approx. 34.48% economy of training time.

A novel data-driven optimization framework for unsupervised and multivariate early-warning threshold modification in risk assessment of deep excavations

Due to the great uncertainty and complexity in the underground environment, precious risk perception in underground construction is a highly challenging problem. Excessive false-positive alarms (FPA) and false-negative alarms (FNA) generated by the unreasonable setting of early-warning values are the main causes to affect the safety of engineering in underground construction. Aiming to reduce the FPA and FNA in risk assessment of an excavation procedure, this paper provides a novel data-driven optimization framework for early-warning threshold modification based on massive field monitoring data. The multi-objective expected value optimization (MEVO) approach is applied to achieve the goal of unsupervised and multivariate modification tasks. Firstly, the objective functions of the MEVO approach are established with the combination of grey relational analysis and Cosine similarity measure. Secondly, the expected values of objective functions are calculated based on the expert evaluation method and Monte Carlo simulation. Finally, the sample sequence generated from observed monitoring data satisfying the expected values is selected as the optimal solution. A case study is utilized to verify the effectiveness and applicability of the proposed approach. The result from the case study indicates that: (1) The proposed approach modifies the multivariate early-warning values to decrease the number of alarms with a reduction ratio of 12% in the training set; (2) The early-warning frequency under the modified values is lower than that under the original values during the whole excavation stages in the testing set; (3) The modified threshold values fall within the threshold intervals based on expert evaluation method, which are capable of improving risk assessment in current practice. In short, our proposed approach enables a more accurate and intelligent evaluation of the risk state by formulating more credible criteria. It contributes to serving as a reliable decision tool for ensuring construction safety during deep excavation.

Ensuring Specificity as a Strategy for Increasing Alarm Safety

Backround: Alarms are systems created to ensure patient safety. However, when its specificity is not ensured, false positive alarms occur, causing the crying wolf phenomenon and compromising patient safety.&#x0D; Purpose: This study aimed to decrease the number of false-positive alarms by ensuring the standardization and specificity of alarms.&#x0D; Methods: This prospective, quasi-experimental study with a pre/post intervention study was conducted in the adult intensive care unit of a training and research hospital through repeated measurements at the beginning and after the intervention. &#x0D; Results: The total alarm load decreased by 46% after the intervention, with the heart rate, saturation, and blood pressure alarms being reduced at a rate of 59%, 56%, and 23%, respectively. The hourly mean number of alarms decreased from 16.8 to 9. Differences between heart rate, oxygen saturation, and blood pressure alarms in terms of the parameter were statistically significant (p

Strengthening Cloud Security: An Innovative Multi-Factor Multi-Layer Authentication Framework for Cloud User Authentication

Cloud multi-factor authentication is a critical security measure that helps strengthen cloud security from unauthorized access and data breaches. Multi-factor authentication verifies that authentic cloud users are only authorized to access cloud apps, data, services, and resources, making it more secure for enterprises and less inconvenient for users. The number of authentication factors varies based on the security framework’s architecture and the required security level. Therefore, implementing a secured multi-factor authentication framework in a cloud platform is a challenging process. In this paper, we developed an adaptive multi-factor multi-layer authentication framework that embeds an access control and intrusion detection mechanisms with an automated selection of authentication methods. The core objective is to enhance a secured cloud platform with low false positive alarms that makes it more difficult for intruders to access the cloud system. To enhance the authentication mechanism and reduce false alarms, multiple authentication factors that include the length, validity, and value of the user factor is implemented with a user’s geolocation and user’s browser confirmation method that increase the identity verification of cloud users. An additional AES-based encryption component is applied to data, which are protected from being disclosed. The AES encryption mechanism is implemented to conceal the login information on the directory provider of the cloud. The proposed framework demonstrated excellent performance in identifying potentially malicious users and intruders, thereby effectively preventing any intentional attacks on the cloud services and data.

Deep Learning Approach for Negation and Speculation Detection for Automated Important Finding Flagging and Extraction in Radiology Report: Internal Validation and Technique Comparison Study

Negation and speculation unrelated to abnormal findings can lead to false-positive alarms for automatic radiology report highlighting or flagging by laboratory information systems. This internal validation study evaluated the performance of natural language processing methods (NegEx, NegBio, NegBERT, and transformers). We annotated all negative and speculative statements unrelated to abnormal findings in reports. In experiment 1, we fine-tuned several transformer models (ALBERT [A Lite Bidirectional Encoder Representations from Transformers], BERT [Bidirectional Encoder Representations from Transformers], DeBERTa [Decoding-Enhanced BERT With Disentangled Attention], DistilBERT [Distilled version of BERT], ELECTRA [Efficiently Learning an Encoder That Classifies Token Replacements Accurately], ERNIE [Enhanced Representation through Knowledge Integration], RoBERTa [Robustly Optimized BERT Pretraining Approach], SpanBERT, and XLNet) and compared their performance using precision, recall, accuracy, and F1-scores. In experiment 2, we compared the best model from experiment 1 with 3 established negation and speculation-detection algorithms (NegEx, NegBio, and NegBERT). Our study collected 6000 radiology reports from 3 branches of the Chi Mei Hospital, covering multiple imaging modalities and body parts. A total of 15.01% (105,755/704,512) of words and 39.45% (4529/11,480) of important diagnostic keywords occurred in negative or speculative statements unrelated to abnormal findings. In experiment 1, all models achieved an accuracy of >0.98 and F1-score of >0.90 on the test data set. ALBERT exhibited the best performance (accuracy=0.991; F1-score=0.958). In experiment 2, ALBERT outperformed the optimized NegEx, NegBio, and NegBERT methods in terms of overall performance (accuracy=0.996; F1-score=0.991), in the prediction of whether diagnostic keywords occur in speculative statements unrelated to abnormal findings, and in the improvement of the performance of keyword extraction (accuracy=0.996; F1-score=0.997). The ALBERT deep learning method showed the best performance. Our results represent a significant advancement in the clinical applications of computer-aided notification systems.