Label-free serum SERS combined with RFE-GBDT algorithm for non-invasive screening of liver cancer.

UAV-based deep learning for biodiversity monitoring: Advances, applications, and future directions

This scoping review assesses machine learning (ML)-based prediction models for autism spectrum disorder (ASD) in early childhood, with the aim of providing a technical and conceptual foundation for improving early ASD detection. Relevant studies on ML-driven ASD prediction models were systematically retrieved from eight databases: PubMed, Embase, Web of Science Core Collection, Cochrane Library, China National Knowledge Infrastructure (CNKI), China Biomedical Database (CBM), Wanfang Data Knowledge Service Platform (WF), and VIP Chinese Science and Technology Journal Database. The scoping review methodology was strictly followed for data extraction and analysis. A total of 16 studies focusing on the application of diverse machine learning algorithms for ASD identification and prediction were included. Among these, 4 studies (25%) employed multiple algorithms for predictive modeling. The most frequently utilized algorithms were tree-based methods (7 studies, 44%), neural networks (NNs) (7 studies, 44%), support vector machines (SVMs) (5 studies, 31%), and regularized logistic regression (3 studies, 19%). Twelve studies (75%) reported Area Under the Curve (AUC) values, all exceeding the 0.7 threshold. Notably, 7 studies (44%) achieved excellent predictive performance with AUC values surpassing 0.9. ML-based models hold substantial promise for the early identification of ASD, which is critical for improving patient outcomes. Future research should focus on standardizing ML model frameworks, refining theoretical underpinnings to enhance practical applicability, and promoting clinical implementation following rigorous validation. These efforts will further enhance the accuracy and utility of such predictive models.

The current approach of Therapeutic Drug Monitoring (TDM) relies on blood analysis to closely monitor drugs with a narrow therapeutic window. This method is uncomfortable for the patient and time-consuming and therefore challenging for frequent monitoring. Electrochemical analysis in sweat is a promising alternative, as sweat sensors are non-invasive and can continuously measure drug concentrations. This study explores novel techniques to improve the analytical performance of voltammetric sensors for TDM in a sweat matrix. Methotrexate (MTX) is selected as the model analyte as it is a widely used therapeutic drug for treatment of cancer, rheumatoid arthritis, among other disorders. Changes in pH and interference from amino acids originating from sweat have been shown to impact the measurement of target drugs such as MTX. Herein, an algorithm is developed to compensate for potential pH fluctuations in sweat by using the relation between the pH level and the peak potential of the electro-oxidized analyte to estimate the pH and calculate the concentration of the analyte. Additionally, an algorithm was developed to separate peaks of distinct amino acids with a similar oxidation potential as MTX. The algorithm uses Gaussian fitting for subtracting and linear discriminant analysis (LDA) to identify the peak related to the analyte. The results demonstrate that the algorithms are effective for the detection of MTX and present an approach to compensating for sweat matrix-related interferences in wearable sweat sensors, driving development for low-cost continuous therapeutic drug monitoring.

The Global Precipitation Measurement (GPM) core satellite’s Dual-frequency Precipitation Radar (DPR) provides new insights into detecting solid-to-liquid phase transition heights of precipitation particles. However, significant anomalies can be found in the “binMixedPhaseTop” parameter within its official Level 2 data products during shallow convective precipitation events. Statistical analysis of the Yangtze and Huai River Basins from 2014 to 2023 reveals that the anomaly rate for shallow convective precipitation is 64.94%, far exceeding those for convective (0.61%) and stratiform (0.48%) precipitation. Further investigation demonstrates that the warm-rain process characteristic of shallow convective systems lacks a distinct solid-to-liquid phase transition layer, leading to algorithmic misinterpretations. An improved identification algorithm for “binMixedPhaseTop” is proposed to address this problem. It incorporates additional checks to ensure that the phase-transition height is only identified when a melting layer is present, thereby preventing misjudgments and redundant computations. The enhanced version is validated against 10-year observations from South China (2014–2023), demonstrating 91% anomaly suppression through extreme deviation truncation. This study highlights the need for algorithm refinement to accurately detect phase-change heights in different precipitation types, thereby enhancing the reliability of the “binMixedPhaseTop” product for convective precipitation detection. 星载双频降水测量雷达GPM DPR能够探测降水相变层顶高度.然而, 其Level 2数据 (V07A) 中的“binMixedPhaseTop”物理量在浅对流降水中表现异常.在2014年—2023年夏季GPM DPR探测江淮流域降水数据中, "binMixedPhaseTop"在浅对流降水中的异常率高达64.94%, 显著高于其在对流降水 (0.61%) 和层状降水 (0.48%) 中的异常率.本文认为浅对流降水的暖雨过程缺乏固态到液态的清晰相变层, 导致GPM DPR现有算法出现误判.为此, 本文提出一种改进的“binMixedPhaseTop”识别算法, 即通过增加相变层存在性验证, 有效避免了无相变层时的误判及冗余计算.验证结果表明改进算法有效降低了GPM DPR探测浅对流降水相变高度的识别异常率.

In order to improve the safety and energy utilization of vehicles, a combination of vehicle stability criterion models and traffic flow models is proposed to plan vehicle paths from the level of path planning to avoid extreme and inefficient working conditions, enabling fast and safe driving under slippery road conditions. The traffic flow model is used to predict the future changes in the traffic environment that the vehicle will face, and then the stability criterion model is used to assess the safety of future traffic in order to plan the fastest and safest path for the hybrid vehicle. Specifically, the generalized Aw–Rascle–Zhang (GARZ) macroscopic traffic flow model is solved using the flux vector splitting format in order to predict the future changes in speed and traffic density that the hybrid vehicle will face. In addition, the front-wheel steering angle responses given by the same driver at different speeds and different distances relative to the vehicle in front were collected using the driver-in-the-loop simulation platform Prescan. Simulink models based on front-wheel drive (FWD) front-wheel steering (FWS) vehicles and all-wheel steering (AWS) distributed drive vehicles (DDVs) give the force saturation factor [Formula: see text] response corresponding to different front-wheel steering angles. The stability criterion model of the vehicle was established by using artificial neural network (ANN) to train [Formula: see text] corresponding to different speeds and traffic densities. The parameters predicted by the traffic flow model (vehicle speed and traffic density) were evaluated for stability using the newly established stability criterion model. The vehicle traveling paths were optimized based on the above methods to ensure the safety of vehicle traveling on slippery road surfaces. Finally, real US-101 traffic flow data were used to verify the predictions of the traffic flow model.

This study presents an integrated machine learning framework for separating narrowband components and extracting Lloyd's mirror interference patterns from ship-radiated noise(SRN) spectrograms. The proposed methodology employs Independent Vector Analysis to separate narrowband spectral components from multi-hydrophone acoustic signals, subsequently applying DBSCAN and RANSAC algorithms for robust identification of parabolic Lloyd's mirror patterns in residual spectrograms. Experimental validation utilizing SRN data acquired during the SAVEX-15 sea trials demonstrates effective narrowband component separation, as verified through DEMON and LOFAR analyses, alongside accurate pattern extraction capabilities. The unsupervised framework exhibits enhanced reliability under adverse noise conditions and enables precise closest point of approach(CPA) estimation. The developed methodology offers an automated and robust solution for SRN analysis, significantly improving acoustic signal interpretation and target identification capabilities in maritime defense applications.

A bstract The ATLAS experiment has performed a measurement of coherent exclusive J/ψ → μ + μ − production in ultraperipheral Pb+Pb collisions at $$ \sqrt{s_{\textrm{NN}}}=5.36 $$ s NN = 5.36 TeV. The data was recorded at the Large Hadron Collider (LHC) during 2023, and corresponds to an integrated luminosity of 79 μ b − 1 . Exclusive J/ψ candidates were selected with a dedicated track-sensitive trigger based on the ATLAS transition radiation tracker. The analysis involves reconstruction of the dimuon invariant mass based on muon tracks from the inner detector, as the muon transverse momentum range of interest precludes the use of the standard muon reconstruction and identification algorithms. Differential cross sections are measured as a function of J/ψ rapidity and are compared with theoretical predictions. After extrapolation to $$ \sqrt{s_{\textrm{NN}}}=5.02 $$ s NN = 5.02 TeV, they are also compared with previous measurements performed by other experiments using data from LHC Run 2. While the results agree reasonably well with theoretical predictions, they are in tension with previous Run-2 results for the central rapidity region.

Adaptive algorithm for pigment identification from unmixing spectral data: Case study with two versions of a XVI century painting.

Protein phosphorylation is a central post-translational modification regulating cellular signaling, frequently dysregulated in cancer. Mass spectrometry (MS)-based phosphoproteomics has emerged as a powerful approach to systematically profile phosphorylation events, thereby revealing aberrant kinase activity and therapeutic vulnerabilities that are not captured by genomic or transcriptomic analyses. Recent advances across the workflow-including optimized sample preparation and phosphopeptide enrichment, isotope- or label-free quantitative strategies, high-resolution mass spectrometry platforms, specialized algorithms for site identification and quantification, and integrative informatics analyses-have enabled the detection of tens of thousands of phosphorylation sites even from small clinical specimens. These developments have facilitated the characterization of signaling pathways across diverse cancer types, leading to the identification of targetable kinases and informing therapeutic strategies. In this review, we highlight studies that employed phosphoproteomic analyses of clinical specimens or patient-derived cancer cells to delineate signaling characteristics and to propose and validate therapeutic targets. Collectively, MS-based phosphoproteomics is poised to become a cornerstone of precision oncology. By enabling comprehensive and quantitative mapping of phosphorylation events, this technology allows mechanistic dissection of cancer signaling pathways and uncovers therapeutic vulnerabilities that may be exploited with targeted agents.

A reinforcement learning–based active interception algorithm for wireless networks topology identification

Crowdsensing-based automated operational modal analysis for indirect bridge structural health monitoring

Abstract The installation quality of gas-insulated switchgear (GIS) directly impacts the long-term operational reliability of power grids, where dust contamination in installation environments poses a critical risk factor for insulation degradation and partial discharge. To address the challenge of traditional dust removal methods failing to dynamically respond to dust dispersion under complex airflow conditions, this study proposes an indoor dust identification and collaborative scheduling optimization method for GIS based on IWOA. This approach provides an efficient and intelligent decision support tool to enhance cleanliness control in GIS installation environments.

Abstract The Very High Frequency Omnidirectional Range (VOR) is a radio navigation system used in civil aviation for direction finding. Operating in the open VHF band, it is susceptible to radio interference such as AM and FM, leading to degraded navigation service performance. Leveraging the known spectral characteristics of VOR signals, this paper proposes an interference recognition algorithm combining cascaded notch filtering and spectral features. A VOR channel monitoring and interference recognition system based on the ZYNQ platform is designed and implemented. This system employs a hardware-software co-design approach, significantly optimizing resource utilization. When the interference ratio ranges from 0 dB to 20 dB, the interference recognition rate consistently remains above 95%, with both false alarm and false rejection rates below 8%. The system demonstrates outstanding real-time performance and reliability, providing an effective solution for engineering applications.

To address node heterogeneity, complex functional dependences, and the limited ability of traditional metrics to capture system-level failure impacts in combat networks, this paper proposes a node-importance evaluation and critical-node identification method that couples failure propagation modeling with flow-blockage theory. We first construct a directed heterogeneous network with five functional node types and explicitly define their resource interfaces and dependency paths. An improved threshold-based propagation mechanism and a composite influence function integrating propagation probability, neighbor overlap, and the KHC topological index are then introduced, and a propagation-efficiency coupled identification algorithm is developed using the max-flow/min-cut principle to quantify traffic degradation under failures. Simulations across multiple failure scenarios and network topologies show that the proposed method significantly outperforms conventional centrality measures in identifying system-level high-loss nodes, yielding more actionable, task-chain–focused results with strong adaptability and robustness. These findings provide theoretical and algorithmic support for combat-network vulnerability assessment, resilient command-system design, and suppression-path planning.

Survey-french national health data system (SNDS) linkage: A win-win methodology for longitudinal studies, algorithm validation, and real-world evidence.

The availability of a mathematical model of an object plays a key role in the synthesis of control systems, as it enables a deeper analysis of the system dynamic properties, and allows for the assessment of its stability, controllability, and observability. However, in real-world conditions, the object parameters are often unknown, which brings methods of parametric identification to the forefront. These methods facilitate the construction of a model based on experimental data. Although modern object identification methods are capable of obtaining mathematical models, they face a number of fundamental challenges, such as selecting the model structure, achieving the required accuracy, managing computational complexity, and mitigating sensitivity to noise and data quality. This work places primary emphasis on what is deemed more critical: namely, the accuracy, noise immunity of the algorithm, and the quality of the input data. An algorithm for the parametric identification of a linear dynamic object with a single input and a single output, described by an n-th order differential equation, under conditions of incomplete a prior information is considered. Incomplete a prior information here implies the absence of information on the derivatives of the identified object input and output signals. An approach is proposed for refining the parameter estimates obtained via the least squares method through the application of the instrumental variable. The mutual correlation functions between the instrumental variable and the output signal of a third-order object, as well as between the instrumental variable and the generalized noise, are presented. The results of identifying a third-order object, obtained by the classical least squares method and the instrumental variable method, are provided. The comparative results demonstrate the effectiveness of the proposed method for refining the least squares estimates using the instrumental variable approach.

Immunocompromised individuals experience an impaired immune function due to conditions that might be either congenital or acquired over the course of their lives. Epidemiological studies often rely on clinical definitions which, in some cases, benefit from being translated into machine-readable algorithms for application to electronic health records (EHRs) databases. The transient nature of certain immunocompromised states and the variability of phenotypes, definitions, coding practices, and data availability entangle this operation. To address these challenges, we conducted a scoping review of existing immunocompromised status definitions in MEDLINE, focusing on epidemiologic and pharmacoepidemiologic studies involving immunocompromised populations. Data extraction was guided by clinical experts, categorizing conditions and medications into seven categories: genetic/hereditary conditions, infectious diseases, malignancies and chemotherapy, organ and stem-cell transplantations, severe systemic conditions, immunosuppressive drugs, and autoimmune conditions associated with immunosuppressant use. Out of 137 citations, 56 studies were included. Most of the studies focused on a particular disease or therapeutic area. Frequently cited diagnoses included HIV/AIDS (17.9%) and organ transplantation (14.2%). Methotrexate, corticosteroids, TNF-alpha inhibitors, and calcineurin inhibitors were the most common drugs used to define immunocompromised status. Building on this review and expert opinion, we developed a phenotype algorithm that combines diagnostic, therapeutic, and procedural data in a modular way to identify immunocompromised populations in EHR data sources. The proposed phenotype algorithm can be applied across diverse data sources, settings and research questions. Future research should test its applicability across heterogeneous EHR data sources.

The marine information industry is a core component of the digital marine economy, characterized by its wide coverage and strong economic driving force. To accurately reflect its development status, this study first clarifies the definition of the marine information industry (taking marine information as the core resource, covering equipment manufacturing, software development, and information services related to marine information production, collection, transmission, processing, and application) and constructs a “perception-transmission-processing-application” industrial chain system. Based on the National Economic Industry Classification (GB/T 4754-2017), the industry is divided into a core layer (4 categories: marine information collection, communication and transmission, storage and analysis, application and sharing services) and a support layer (1 category: marine information supporting industry), totaling 11 sub-industries. Subsequently, a marine information enterprise identification method based on big data mining is established, integrating enterprise names, core businesses, patent information, and other multi-source data, and using text feature matching and machine learning algorithms for accurate identification. Finally, three value-added calculation methods (industry stripping method, value-added rate method, and input-output method) are proposed for different types of sub-industries. The trial calculation of T city’s 2022 marine information industry reveals that marine information communication and transmission account for 49% of the total value-added, storage and analysis for 24%, application and sharing services for 13%, and supporting industries for 12%. This method is suitable for value-added calculation when basic data is detailed, providing a standardized calculation framework for marine information industry statistics and filling the gap in current research on marine-derived industry accounting. It also offers a reference for the accounting of other marine-derived industries. Beyond its contribution to marine industrial statistics, this framework is directly relevant to aquaculture and fisheries management by enabling standardized accounting of marine information infrastructures that support stock assessment, environmental monitoring, biosecurity surveillance, and evidence-based decision-making for sustainable aquatic food systems.

PurposeThe purpose of this study is to estimate the prevalence and incidence of treatment-range adolescent idiopathic scoliosis (≥20°) over 10-years in Ontario youth 10–17 years of age, by validating a population-based health administrative data algorithm for case ascertainment.Patients and MethodsAlgorithms were developed using a combination of health administrative data: diagnostic, fee and/or specialty codes from physician billing data over various look-back periods. Algorithms’ ability to distinguish between youth with scoliosis, confirmed by a tertiary-care spine specialist (AIS+; n = 2732), and a provincially derived comparator group without (AIS-; n = 49,049) were evaluated using sensitivity, specificity, positive and negative predictive values with their 95% confidence intervals. The top performing algorithm was used to estimate sex- and age-standardized prevalence and incidence between 2012 and 21. Annual rate ratios were calculated using a negative binomial regression model, adjusted for age, sex, and age–sex interaction. Significance was accepted at p < 0.05.ResultsThe AIS+ cohort had a median curve magnitude of 35° (interquartile range: 25.5–45.5). Of the 93 algorithms tested, the top was “2 physician billing codes for scoliosis in 2 years” with sensitivity: 83.1% (95% CI, 81.6–84.5%), specificity: 99.3% (95% CI, 99.2–99.3%), positive predictive value: 86.3% (95% CI, 85.0–87.6%), negative predictive value: 99.1% (95% CI, 99.0–99.1%). Annual prevalence estimates averaged 513.3/100 000 and incidence 128.2/100 000. There was a modest annual increase in the adjusted rate ratios: 1% for prevalence, 2% for incidence. Rates were highest for females at 13-years and males at 15-years of age, with rates 65% lower for males.ConclusionThe selected health administrative data algorithm demonstrated excellent diagnostic accuracy in identifying radiographically confirmed, treatment-range adolescent idiopathic scoliosis in 10–17-year-old youth. This is an efficient and scalable method for clinically meaningful population-level cohort creation that will facilitate surveillance of scoliosis diagnostic and treatment trends and longitudinal outcome research.