搜索结果：Signal Processing

High-fidelity simulation of nanopore sequencing signals is critical for rigorous benchmarking and validation of the nanopore signal processing pipeline. However, existing signal simulators often fail to capture the non-linear dynamics of nanopore current signals, relying on static pore models or lacking optimization objectives tied to basecalling, resulting in synthetic signals with low basecalling accuracy and fidelity. We introduce NanoSimFormer, an end-to-end Transformer-based signal simulator that integrates basecaller guidance during training to generate high-fidelity nanopore signals. NanoSimFormer achieves a median basecalling accuracy exceeding 99% and Q-scores above 22.8 for Oxford Nanopore Technologies' latest DNA R10.4.1 and direct RNA sequencing, closely mirroring real experimental baselines. It faithfully recapitulates experimental variant calling performance across the five human samples, achieving F1-scores of 0.9953-0.9973 and 0.7862-0.8612 for single-nucleotide polymorphisms and small indels detections, respectively. Compared with previous simulators, NanoSimFormer also substantially reduces false positives in homopolymer and short tandem repeat regions. NanoSimFormer-derived reads enable high-quality de novo bacterial assembly with consensus error rates below one mismatch per 100 kbp and maintain high correlations with experimental abundance in metagenomic and transcriptomic datasets. NanoSimFormer is freely available on GitHub at: https://github.com/BioinfoSZU/NanoSimFormer. Supplementary data are available at Bioinformatics online.

Long COVID is increasingly associated with disruption in brain homeostasis, manifesting as severe neurological dysfunction, brain fog and cognitive impairment. This present study investigated localised cognitive deficits in long COVID patients by examining brain blood oxygenation-level-dependent (BOLD) signal activity using ultra-high-field 7&#x202f;T (7T) task-based functional magnetic resonance imaging (fMRI). Whole-brain BOLD signal differences were assessed across 19 long COVID patients, and 27 healthy controls (HC) including 12 COVID-recovered (Cov-RHC) and 15 COVID-19-na&#xef;ve HC (nHC). 225 fMRI volumes were acquired during the Stroop colour-word task. Functional and anatomical images were processed using SPM12 to extract the BOLD signal intensity time course from whole-brain voxels for inferences between cohorts during task-fMRI. Significantly low BOLD activation in long COVID patients was observed compared to Cov-RHC in the anterior cingulate cortex (p&#x202f;=&#x202f;0.002, cluster size=650, Z-value&#x202f;=&#x202f;4.67), and the precuneus (p&#x202f;=&#x202f;<0.001, cluster size&#x202f;=&#x202f;1893, Z-value&#x202f;=&#x202f;4.67). Furthermore, BOLD intensities in precuneus showed a negative association with self-reported pain scores (p&#x202f;=&#x202f;0.040) and the duration of illness (p&#x202f;=&#x202f;0.03) in long COVID patients, suggesting significant correlation between BOLD signal and an increase in duration of illness and pain levels. No statistically significant BOLD differences were observed for inter-group comparisons between nHC vs. long COVID, and nHC vs. Cov-RHC. Response times to incongruent (p&#x202f;=&#x202f;0.002) and congruent task stimuli (p&#x202f;=&#x202f;0.001) significantly varied between nHC and long COVID cohorts, demonstrating overall faster information processing by nHC. Reduced BOLD signals to 'core' brain regions in long COVID imply reduced cognitive control by intrinsic networks that mediate information processing, cognitive and executive functions due to perturbations linked to cerebral blood flow, oxygenation status, and ongoing neuroinflammation.

House mice (Mus musculus), like other rodents, communicate using sonic and ultrasonic vocalizations (USVs), but their functions are still poorly understood. One of the main challenges for studying any acoustic communication is processing and analyzing audio files. Our aims here are to provide a critical and comprehensive review of the new bioacoustic tools available for processing and analyzing recordings of mouse vocalizations. We consider each method as used in a serial data processing pipeline and how to minimize errors at each step to prevent error propagation (or error cascades). First, we review methods for processing audio files of recordings of mice. We compare conventional approaches for visualizing vocalizations (time-frequency representations) with an alternative method adapted to the mouse auditory system. We compare machine learning (ML) and signal processing methods for automating USV detection and emphasize the need for better methods for denoising audio files and reliable frequency contour (ridge) tracking and feature extraction. Second, we review methods for analyzing detected USVs, focusing on classification and sequencing approaches. Classifying USVs is a challenging task because, while some calls are discrete, others show graded variation within and between call classes. We compare supervised classifications and unsupervised labeling, and we emphasize the importance of reliable manual (researcher-based) methods as a gold standard for automated ML approaches. We review classifications of mouse vocalizations in the literature, and we propose a new hierarchical framework for the classification of USVs. We examine methods for sequencing USVs and consider their relative advantages. Finally, we address the unresolved technological challenges for these methods to study rodent vocalizations and propose potential solutions for the future.

Elderly patients are highly susceptible to drug-drug interaction (DDI)-induced liver injury, yet comprehensive real-world evidence remains scarce. This study aimed to leverage a natural language processing (NLP) model to efficiently identify hepatotoxicity-associated DDI signals in this population. This retrospective study analyzed electronic health records from 109,263 elderly inpatients. An integrated approach combining laboratory thresholds and NLP was utilized to precisely identify liver injury cases. High-throughput case-control analyses were conducted using additive and multiplicative interaction models to screen for DDI signals. To minimize false positives, time-dependent causality inference and sensitivity analyses were performed, followed by validation using in vivo animal experiments. Among 3,227 drug combinations evaluated, 111 signals were identified, 58 of which demonstrated consistent time-dependent risk trends. Notably, among the top 20 DDI signals with the strongest associations, a marked risk elevation was observed for cardiovascular medications such as aspirin (additive: 1.16; multiplicative: 2.49), clopidogrel (additive: 1.12; multiplicative: 2.12), and atorvastatin (additive: 0.52; multiplicative: 1.88) when co-administered with the antimicrobial agent piperacillin/tazobactam. Animal experiments and sensitivity analyses further corroborated these findings. We established a robust NLP-based framework to efficiently evaluate DDI-induced hepatotoxicity in elderly inpatients. By identifying novel high-risk combinations, this study provides critical insights for optimizing pharmacotherapy and highlights the potential of artificial intelligence technologies in real-world pharmacovigilance study.

Detecting concealed information is a critical challenge in forensic investigations, security screening, and cognitive neuroscience. Conventional approaches using the concealed information test primarily rely on binary classification, distinguishing between recognized (concealed) and unrecognized (neutral) stimuli. This limits interpretability and fails to reveal the nature of the concealed knowledge. In this pilot study, we present a novel multimodal framework that moves beyond binary detection by decoding the category of concealed information into object, person, or place based on neurophysiological signals recorded during a concealed information test. High density electroencephalography and physiological signals, including skin temperature, galvanic skin response, and plethysmogram, were simultaneously recorded from ten subjects as they viewed visual stimuli representing these three categories. Temporal and spectral features were extracted from both modalities, followed by machine learning based multimodal fusion for classification. The proposed framework achieved an overall accuracy of 94.2%, significantly outperforming unimodal EEG (73%) and physiological (54.2%) baselines. Further analysis showed that similar decoding performance can be achieved using as few as eight strategically selected EEG electrodes, supporting the feasibility of lightweight, wearable implementations. The most informative electrodes were located over prefrontal and frontocentral regions, aligning with cognitive processes related to attention, recognition, and deception. These findings demonstrate that neurophysiological signals enables not only the detection of concealed knowledge but also the identification of the type of hidden information. The integration of EEG and physiological signals enhances both sensitivity and interpretability by capturing complementary aspects of cognitive and affective processing during recognition. By enhancing the CIT paradigm from binary recognition toward semantic decoding, this pilot study advances the development of interpretable deception detection systems and bridges laboratory neuroscience with real world forensic applications.

Nanopore sensing holds transformative potential for revolutionizing protein and glycan sequencing. However, translating this potential into practical, high-fidelity identification is severely bottlenecked by the challenge of processing massive amounts of highly similar nanopore ionic-current data, spurring an urgent need for robust, AI-driven solutions. Prevailing deep learning methods suffer from two limitations: they often fail to capture the fine-grained temporal dynamics essential for distinguishing structurally similar analytes, and their generic training strategies inadequately extract weak discriminative features, thus limiting classification precision. Here, we present SEDA-Former (Signal Enhancement and Dynamic Attention Transformer), a deep temporal learning framework designed for high-resolution nanopore single-molecule identification. SEDA-Former incorporates a multi-window sliding standard-deviation method for feature enhancement, a multi-channel temporal convolutional network to mine weak features in temporal dynamics, and a progressive adaptive attention training strategy that dynamically reweights sample losses based on learning difficulty. Across a diverse set of challenging benchmark datasets, including nanopore signals of 15 glycosides, 24 ginsenosides, 8 DNA molecules, and 17 cholic acid conjugates, spanning varying levels of signal complexity, SEDA-Former consistently achieves substantially higher classification accuracy than state-of-the-art methods and demonstrates robust cross-dataset transferability. SEDA-Former provides a versatile and scalable solution to facilitate single-molecule identification in nanopore sensing.

Porcine reproductive and respiratory syndrome virus (PRRSV) is the most economically important pathogen of swine, yet the molecular mechanisms governing its entry into host cells remain incompletely understood. The minor envelope glycoprotein GP2, together with GP3 and GP4, forms an essential complex that engages the entry receptor CD163; however, the specific GP2 regions required for receptor association have not been fully defined experimentally. Here, we investigated GP2 processing, intracellular trafficking and interactions with viral glycoproteins and CD163. We demonstrate that the GP2 signal peptide (SP) is cleaved in both transfected and infected cells and is necessary and sufficient for ER localization. Removal of the SP disrupted GP2 maturation, impaired interactions with GP3, GP4 and GP5, and significantly reduced viral infectivity in infectious clone assays. Deletion of the SP also abolished GP2-CD163 association, indicating that proper SP-dependent processing is required for receptor engagement. Using co-immunoprecipitation and colocalization analyses, we identified two highly conserved regions within the GP2 ectodomain that associate with CD163. Deletion of either region completely eliminated PRRSV infection. Together, these findings define the structural determinants within GP2 required for association with CD163 and advance our understanding of the early steps of PRRSV entry.

Electrocardiogram (ECG)-based diagnostics are pivotal in early cardiac disorder detection, yet existing models often fail to integrate temporal, spectral, and spatial dynamics inherent in complex arrhythmic patterns. Most traditional approaches are unimodal, relying either on time-domain signal processing or spatially limited CNN models, thereby overlooking cross-domain dependencies and subtle morphological cues. Addressing this gap, this research proposes FusionHeartNet, a unified deep learning framework that fuses signal- and image-based representations using a dual-spectrum feature embedding (DSFE) strategy. DSFE synergistically extracts morphological descriptors and spectral signatures via Fourier and wavelet transforms, while spatial morphology is preserved through GAF and CWT scalograms. These dual-domain features are refined by a multi-focus attention module (MFAM) and classified through the heart fusion classifier (HFC), which is optimized using Bayesian optimization with adaptive learning rate scheduling (BO-ALRS). Experimental validation on the MIT-BIH Arrhythmia Database demonstrates an accuracy of 98.47%, F1-score of 91.67%, and kappa of 0.9311, significantly outperforming baseline models. FusionHeartNet sets a new benchmark for robust, multi-dimensional ECG analysis, offering clinically viable precision in early heart disease detection.

Adaptive behaviors shaped by prior experience are essential for increasing animal survival. Aversive experiences play a pivotal role in memory formation and in updating subsequent learning rules. While the negative value of aversive signals, which are both necessary and sufficient to drive a conditioned response, is considered to be innately specified, it can also be subject to experience-dependent scaling. Previous reports demonstrated synaptic potentiation in nociceptive pathways following robust aversive learning. However, the neuronal basis of experience-dependent value updating remains largely unknown. Recently, we demonstrated that long-term potentiation (LTP) in the parabrachial-central amygdala (PB-CeA) pathway, an important circuit involved in pain processing and aversive learning, enhances the negative value and thereby updates future learning rules. Here, we present a protocol that combines behavioral analysis using pathway-specific optogenetic induction of in vivo LTP with mathematical modeling to examine value modification using Bayesian inference of the unconditioned stimulus value using the Rescorla-Wagner model. This protocol enables investigation of the mechanisms underlying experience-dependent value modulation and learning-rule changes in mice. Potentially, this protocol may provide a framework for understanding learning rules across a wide range of species and for the development of treatments for stress-related disorders. Key features &#x2022; This protocol enables pathway-specific in vivo LTP induction to investigate the causal relationship between synaptic plasticity and behavioral outputs. &#x2022; The protocol combines behavioral manipulation with computational modeling to interpret value plasticity of the instructive signal in terms of learning-rule updating via circuit-level plasticity.

Arc faults in pantograph catenary systems pose a significant threat to the reliability, safety, and efficiency of the electric railways. However, Conventional approaches relying on image processing and deep learning are hindered in real-time applications due to computational delays exceeding the arc time constant. Therefore, this study proposes a novel image-free arc detection method that directly analyzes measured traction current signals using DWT-ANN (Discrete Wavelet Transform-Artificial Neural Networks). The significance of this work lies in its ability to extract transient arc features directly from the traction current waveform, providing a computationally efficient solution for real-time monitoring without the need for additional sensing modalities. Various mother wavelet families are evaluated, and specific detail levels are identified as the most informative for capturing arc transients. Among these, Daubechies db9 and Symlet sym8 showed the highest discriminative performance and were used as inputs to a compact ANN structure. The network, trained with normalized feature data, achieved a high regression coefficient, indicating excellent classification accuracy. Additionally, algorithm robustness was validated by training the neural network on db4-extracted features and testing it across all accepted wavelets, with consistent detection performance. The results confirm the robustness and practicality of the proposed DWT-ANN framework for real-time arc detection in railway systems.

Glucosylceramide (GlcCer)-based liposomes are glycosphingolipid-rich liposomal systems whose intracellular behavior requires further characterization. In this study, we investigated the intracellular localization and time-dependent behavior of rice bran-derived GlcCer-based liposomes in TIG-103 human dermal fibroblasts. A small amount of BODIPY-labeled sphingolipid was incorporated into the liposomal membrane, and Cascade Blue-labeled dextran was encapsulated in the aqueous lumen to construct a F&#xf6;rster resonance energy transfer (FRET) system. Co-staining with LysoTracker and ER-Tracker was performed to examine intracellular localization. Liposomes encapsulating &#x3b1;-4-methylumbelliferyl glucopyranoside (&#x3b1;-4MUG) or &#x3b2;-4-methylumbelliferyl glucopyranoside (&#x3b2;-4MUG) were also used to assess intracellular substrate processing. After uptake into TIG-103&#x202f;cells, GlcCer liposomes showed punctate cytoplasmic distribution at early time points together with detectable FRET-related signals. With prolonged incubation, the spatial relationship between the BODIPY signal and the FRET signal changed, and a relatively strong perinuclear BODIPY-positive region became evident. Image-based analyses indicated a higher correspondence of the BODIPY signal with LysoTracker than with ER-tracker/Cascade Blue-related signals. Intracellular 4-methylumbelliferone signals were detected in cells treated with liposomes encapsulating either &#x3b1;-4MUG or &#x3b2;-4MUG. An exploratory investigation of GlcCer liposomes in TIG-103&#x202f;cells under nutrient-deprived conditions was also examined without fluorescence-based trafficking experiments.

Alteration of blood perfusion leads to some of the most common cardiovascular pathologies. Current methods for measuring perfusion use fluorescent polystyrene microspheres (MS) that are systemically injected prior to processing to obtain the absolute number of MS trapped inside the tissue. The current standard method is cost-intensive and carries a high risk of MS loss, leading to underestimation of regional perfusion. This study aimed to develop an improved, cost-efficient protocol for measuring regional perfusion through the processing and direct imaging of fluorescent MS embedded ex vivo. Porcine and control samples treated with MS were chemically digested, filtered through either a polycarbonate (PCTE) or cellulose filter, and fluorescence was measured either through the standard fluorometric method or through the proposed direct imaging method. In the standard fluorometric method, interactions were found between the PCTE filter and porcine samples, leading to dampened signal and the subsequent underestimation of regional perfusion in practice. The proposed direct imaging method with cellulose filters showed improved sensitivity even within low MS levels (limit of detection improved significantly), amplification of sample fluorescence (11-13&#xd7; when compared to PCTE filters), parity between porcine and control samples, and a reduction in cost providing a significant improvement over the industry standard for fluorescent MS perfusion measurement (28-51 % reduction compared to standard method). The proposed method also removed the need for 2-ethoxy ethyl acetate, a teratogen and plastic softener, and reduced complexity in the workflow.

Cellular signals are essential for sensing the microenvironment and coordinating physiological functions. Their multimodal nature, encompassing electrophysiological, chemical, mechanical, and optical components, helps define cellular functional states and fate. High-resolution spatiotemporal analysis of these weak, dynamic, and heterogeneous signals is critical for elucidating fundamental life processes, uncovering disease mechanisms, and advancing precision medicine. Recent advances in integrated circuit (IC) technology, particularly the co-integration of complementary metal-oxide-semiconductor (CMOS) and micro-electro-mechanical systems (MEMS), have enabled unprecedented capabilities for cellular signal analysis, driving a transition from conventional instruments and microfluidic platforms to chip-level cellular analysis. This review summarizes the key technological foundations, multimodal sensing mechanisms, and emerging applications of IC technology for cellular signal analysis. It explores the core principles of high-sensitivity, long-term stable signal acquisition, focusing on bioelectronic interfaces, biocompatible packaging, and low-noise signal processing. It also reviews breakthroughs in microelectrode arrays, field-effect transistors, CMOS image sensors, MEMS sensors, and multi-parameter chemical sensing chips for single-cell and population-level detection. The review highlights applications in drug screening, clinical diagnostics, single-cell analysis, and brain-computer interfaces. Finally, it addresses challenges such as biocompatibility, crosstalk suppression, and energy efficiency, while outlining future directions in material innovation, three-dimensional integration, and brain-inspired computing.

Excessive tryptamine in food poses a significant risk to human health, emphasizing the demand for efficient, sensitive, and rapid detection technologies. Surface-enhanced Raman Spectroscopy (SERS) aptasensors are presently attracting a lot of attention due to their ability to detect targets at low concentrations. However, their inherent instability and poor anti-interference ability hinder their widespread use. Herein, a two-layer core-satellite magnetic SERS aptasensor was constructed to achieve the sensitive detection of tryptamine. The SERS aptasensor consisted of magnetic SERS recognition probes (magnetite nanoparticles coated with gold nanoparticles and aptamer: Fe3O4@Au-apt) and SERS signal probes (gold nanoparticles of two different sizes functionalized with the Raman reporter molecule 4-mercaptobenzonitrile and complimentary deoxyribonucleic acid strands). In the absence of tryptamine, the SERS signal probes attached to the aptamer on the SERS recognition probes to form a two-layer core-satellite structure with an intense SERS signal at 2226&#xa0;cm-&#x2009;1 in the "biological-silent" region due to the 4-Mercaptobenzonitrile. In the presence of tryptamine, tryptamine bonded to aptamer on the SERS recognition probe leading to detachment of the SERS signal probes, weakening the SERS signal at 2226&#xa0;cm-&#x2009;1. The aptasensor exhibited a favorable linear range from 0.001 to 100 mg L-&#x2009;1, with a detection limit of 0.39&#x2009;&#xd7;&#x2009;10-&#x2009;3 mg L-&#x2009;1 toward tryptamine. The fabricated sensor was practically applied to the detection of tryptamine in liquor, white wine and vinegar samples, with results highly consistent with high performance liquid chromatography data, demonstrating the broad prospects of the developed analytical method in food threat detection.

In resonant modes of atomic force microscopy (AFM), such as tapping mode and multifrequency mode, the cantilever typically oscillates at its fundamental or higher-order resonance, with resonance frequencies reaching several hundred kilohertz. Among various amplitude demodulation methods, the lock-in amplifier is widely used in AFM due to its superior signal-to-noise ratio and strong immunity to interference. However, conventional lock-in amplifiers require high-speed, high-resolution analog-to-digital converters (ADCs) to oversample the high-frequency vibration signal, thereby imposing huge demands on fast data acquisition and real-time processing hardware. However, in most AFM applications, the effective bandwidth contained in the cantilever amplitude variations is much lower than the carrier vibration frequency, typically on the order of several kilohertz. To address this characteristic, a sub-Nyquist sampling-based amplitude demodulation method for resonant-mode AFM is proposed in this work. By exploiting the sub-Nyquist sampling principle, accurate demodulation of the cantilever amplitude is achieved at a sampling rate significantly lower than the Nyquist rate of the vibration signal. The reduced sampling rate provides additional computational time within each sampling period, enabling real-time amplitude demodulation without the need for high-speed ADCs and processing hardware. The proposed method is validated through force-distance curve measurements and imaging experiments. The extracted force-curve characteristics and image quality show good agreement with those obtained using a conventional lock-in amplifier. These results demonstrate that the method provides a practical and cost-effective approach for reducing hardware complexity in resonant-mode AFM systems.

Irritable bowel syndrome (IBS) is one of the most common functional gut disorders affecting the global population, characterized by chronic abdominal pain and altered bowel habits in the absence of structural disease. The brain-gut-microbiota axis, a bidirectional network integrating central nervous system processing, enteric and autonomic function, immune signaling, and gut microbial ecology, provides a mechanistic framework that helps explain the substantial symptom heterogeneity and variable treatment response observed across patients. Artificial intelligence (AI) and machine learning (ML) approaches offer the ability to model complex, nonlinear relationships across high-dimensional biological datasets generated from this axis, including microbiome composition profiles, resting-state functional MRI connectivity matrices, multiomics data layers, and psychological and clinical feature sets. This narrative review evaluated primary human studies applying AI and ML to brain-gut axis data in IBS, identified through structured searches of PubMed/MEDLINE and Scopus supplemented by citation chaining, with literature included up to April 2026. Across microbiome profiling, neuroimaging, multiomics integration, and psychological feature modeling, ML approaches have demonstrated proof-of-concept performance for IBS classification and, in a smaller number of studies, for prediction of clinically meaningful outcomes, including cognitive behavioral therapy (CBT) response. A notable early signal is the integration of baseline microbiome and brain features to predict CBT response, with high reported discrimination, although these results are derived from small, single-center cohorts with only internal validation and should be regarded as hypothesis-generating. The current evidence base is limited by small single-center cohorts, reliance on internal validation, healthy-control comparators, limited external replication, and substantial overfitting and data-leakage risk in high-dimensional small-sample settings. AI and ML applications in IBS are promising but remain exploratory and are not yet suitable for routine clinical use. Clinical translation will require larger multicenter datasets, harmonized preprocessing pipelines, external validation, calibration reporting, and evaluation against clinically realistic comparators and decision points.

Congestive Heart Failure (CHF) is a chronic condition where the heart fails to pump enough blood, leading to fluid buildup in the lungs and respiratory distress. Monitoring thoracic fluid levels is essential for managing CHF and other fluid-related disorders. Although invasive methods provide accurate measurements, they pose risks and discomfort, highlighting the need for non-invasive alternatives. This review explores advanced methods for thoracic fluid measurement, emphasizing non-invasive techniques. It assesses their accuracy, clinical relevance, and technological advancements, with a focus on improving reliability and usability. A comprehensive review of literature was conducted to evaluate existing thoracic fluid measurement techniques. The study examines bioimpedance spectroscopy, imaging modalities, and computational approaches. Additionally, advancements in sensor technology, signal processing, and machine learning applications are discussed. Case studies and clinical trials are reviewed to determine real-world applicability. Non-invasive methods, particularly bioimpedance-based techniques, have shown significant improvements in accuracy and ease of use. Innovations in sensors and signal processing have enhanced measurement precision. Clinical studies suggest these methods offer a viable alternative to traditional invasive approaches. However, challenges such as patient variability and standardization still need to be addressed. Non-invasive thoracic fluid measurement techniques have great potential for improving patient care in CHF and fluid overload conditions. While advancements have strengthened their effectiveness, further research is necessary to refine these technologies and facilitate broader clinical adoption. This review provides insights to guide future developments in noninvasive monitoring systems.

Inspired by the auditory system's capacity to process spatiotemporal sound patterns, voiceprint recognition plays a vital role in identity authentication and security. However, current platforms often face challenges of speech frequency and amplitude variability, hindering accurate feature extraction in noisy environments. To address these issues, a large-scale hybrid metal-halide dynamic memristor (MHDM) featuring an engineered gradient-distributed architecture is developed for adaptive voiceprint recognition. The spontaneously graded metal-halide functional layer allows for precise modulation of Schottky barriers and redistribution of interface charges. This design achieves &#xb5;s-scale response, enhances noise tolerance (over 20% improvement in signal-to-noise ratio), and enables kHz-scale dynamic signal processing. Experimental results demonstrate that the MHDM achieves a voiceprint recognition accuracy of 99.3%, maintaining high performance at 93.2% even in realistic background noise. These findings demonstrate the system's potential for secure and efficient voiceprint recognition, combining scalability with robust performance in noisy environments.

We developed a scalable pipeline for extracellular miRNA (ex-miRNA) profiling that integrates automated exRNA extraction, small RNA sequencing, and bioinformatic analysis including data processing and normalization. Automated extraction protocols, including doubling input volume or lyophilization to increase RNA yield, were benchmarked against leading manual methods, with donor pregnancy status serving as the primary biological variable. Small RNA sequencing was performed across conditions, enabling systematic evaluation of data quality and comparison of normalization strategies. Across methods and biofluids, DESeq2 most effectively reduced technical variability while preserving biological signal. Comparing specimen types, plasma exhibited the highest reproducibility and retention of biological signal, followed by serum, while urine exhibited greater variability and less differentially expressed miRNAs. Pregnancy-associated ex-miRNA signatures, including C19MC miRNAs, were consistently detected in both plasma and serum. Together, this study establishes a robust framework for scalable exRNA extraction and profiling, supporting standardized assay development for biomarker discovery and clinical applications.