Engineering

Abstract: Diabetes mellitus affects over 537 million adults worldwide, yet the impact of hyperglycemia on absorbable suture degradation remains poorly understood. This study evaluated the in vitro degradation of polyglycolic acid (PGA) and PGLA 910 copolymer sutures under glucose concentrations of 0, 300, and 600 mg/dL over 28 days. A total of 120 sutures (60 per material) were incubated at 37°C in phosphate-buffered saline with different glucose concentrations. Mechanical properties were assessed at baseline and at 7, 14, 21, and 28 days via tensile testing. Medium pH was monitored weekly, and surface morphology was characterized by scanning electron microscopy. PGLA 910 demonstrated higher baseline strength than PGA (14.56 ± 1.11 N vs. 13.37 ± 1.00 N; p < 0.001). Both materials maintained strength during the first 7 days (p = 0.080), but progressive loss occurred from day 14 onward. At day 28 under normoglycemic conditions (0 mg/dL glucose) , PGA retained only 1.12 N (8.4% of initial strength), while PGLA 910 retained 3.81 N (26.2%; p < 0.001). Glucose concentration did not significantly affect degradation (p = 0.153). Medium pH correlated strongly with residual strength (ρ = 0.86–0.89; p < 0.001). SEM revealed more severe surface degradation in PGA than in PGLA 910. We conclude that PGLA 910 offers superior resistance to hydrolytic degradation, making it a more suitable choice for diabetic patients when extended mechanical integrity is required. However, both materials degrade extensively within 4 weeks and are unsuitable for applications requiring support beyond 3 weeks. The strong pH–strength correlation confirms the autocatalytic nature of the degradation mechanism.

Abstract: Heart disease remains one of the leading causes of mortality worldwide, highlighting the need for accessible tools for early risk assessment. This study investigates the use of established machine learning (ML) techniques to predict heart disease risk from self-reported information that can be collected through telephone or online health questionnaires. The analysis is based on the publicly available Heart-2020 dataset, derived from the U.S. Centers for Disease Control and Prevention (CDC) Behavioral Risk Factor Surveillance System (BRFSS). Four widely used ML models were trained and evaluated: Multi-Layer Perceptron (MLP), Support Vector Machine (SVM), Bagged Trees (BT), and Extreme Gradient Boosting (XGBoost). Model performance was assessed using precision, recall, F1-score, and the Area Under the Receiver Operating Characteristic Curve (AUC). To further improve predictive performance and robustness, an ensemble architecture based on a second-stage MLP was implemented to combine the outputs of the individual classifiers. In addition, several feature engineering techniques, including Pearson correlation analysis, Analysis of Variance (ANOVA), and Principal Component Analysis (PCA), were investigated. The proposed approaches achieved F1-scores exceeding 0.80, demonstrating strong predictive capability using only non-clinical, self-reported information. These findings suggest that simple questionnaire-based data can support automated early-warning systems capable of identifying individuals who may benefit from further medical evaluation. Beyond individual risk assessment, the proposed methodology could also facilitate large-scale population health monitoring, contributing to preventive healthcare strategies and informed public health policy development.

Abstract: Background: Breast cancer is the most common cancer among women, and early detection through mammography is essential for reducing mortality. Artificial intelligence can support radiologists by improving diagnostic accuracy. Aim: To develop and evaluate a two-stage ensemble machine learning pipeline for breast cancer diagnosis from digital mammograms. Methods: The proposed framework combines image preprocessing, multiple convolutional neural networks trained under different conditions, and a second-stage classifier that integrates the CNN outputs. Several machine learning models and feature selection techniques were evaluated using publicly available mammography datasets. Results: The ensemble approach consistently outperformed the individual CNN models. The MLP classifier achieved the best overall balance between precision and recall, while the heuristic fusion method provided the highest sensitivity. Feature selection reduced model complexity while maintaining comparable performance, and cross-validation confirmed the robustness of the proposed methodology. Discussion: Combining complementary information from multiple CNNs with classical machine learning improves diagnostic performance and provides a robust framework for computer-aided breast cancer diagnosis. Conclusions: The proposed two-stage ensemble offers an effective and interpretable approach for mammographic breast cancer classification. A demonstration application incorporating Grad-CAM explainability further supports its potential use as a clinical decision-support tool.

Abstract: This paper outlines a methodology for using Ultraviolet-Visible (UV-Vis) spectroscopy to measure the concentration of the primary nutrients (N, P, K) and calcium (Ca) for managing hydroponics fertigation solutions and is an extension of previous work presented by the team at GreenSys 2025. This project focuses on supplanting electrical conductivity (EC) as the go-to for inline control of the nutrient mixtures. EC only gives a picture of the total nutrient content of the solution and cannot give insight into the amount of each element present. Others have demonstrated UV-Vis spectroscopy in this measurement role using machine learning for the primary nutrients. The aim here was to improve upon that with a calibration method that is quick, easily reproducible, and usable for an inline controller for near real-time control in hydroponics while also adding support for Ca measurement.

Abstract: Post-COVID balance and postural-control alterations have been reported after SARS-CoV-2 infection, but individual-level identification remains challenging because post-acute sequelae are heterogeneous and often lack clear clinical ground truth. We analyzed motion-capture balance data from 140 participants, including 114 post-COVID participants and 26 Controls, who completed sensory organization/postural balance tasks under standard and Stroop-augmented conditions. Relative joint-angle trajectories were transformed into 0–5 Hz frequency-domain spectra, from which 13 balance-derived task contrasts were constructed to capture responses to cognitive distraction, visual input removal, and task-condition changes. Joint-wise spectral contrast features were encoded using a vector-quantized variational autoencoder. Abnormal-like subgroup discovery was then formulated as a control-anchored, graph-based QUBO clustering problem and solved using a quantum annealing workflow, with Gaussian mixture modeling used as an unconstrained classical comparator. Control-anchored QUBO clustering identified a selective subset of post-COVID participants with abnormal-like latent balance profiles while largely preserving the Control reference structure. Across all task contrasts, QUBO assigned 1 of 307 Control contrast-level observations (0.3%) and 118 of 1366 post-COVID observations (8.6%) to the abnormal-like group, compared with 63 of 307 Control observations (20.5%) and 266 of 1366 post-COVID observations (19.5%) under GMM. Frequency-band profiles further suggested that QUBO-derived abnormal-like assignments corresponded to interpretable postural-control patterns, particularly in low-frequency postural-sway-dominant and normal postural-control activity bands. These findings support the feasibility of objective-driven, QUBO-based quantum-assisted clustering for exploratory subgroup discovery in heterogeneous, label-limited biomedical datasets. The proposed framework does not establish diagnostic criteria for Long COVID, but provides a hypothesis-generating approach for interpretable post-COVID neuromotor phenotyping and related medical data analysis applications.

Abstract: Biomechanics is a research area in which we explore the human body with tools from mathematics and physics. We create models of the human body to study its function, limitations, and performance, and to develop technologies that replace or support its components. The human is inherently intertwined with our research focus. But the human is not just our topic of research, not just something captured in our models and data. We, as researchers, are equally entangled with what we study. We follow our intuition, our experiences, and our knowledge to steer research in particular directions. We choose which topics to pursue, which people to recruit into our teams, which participants to include or which specimens to select, and how to interpret and report what that research produces. In this article, we look at the biomechanics field through four perspectives on where humans are involved: the participants and specimens, the data and models, the researchers, and the research topics. These four perspectives form an interconnected cycle, bias in any one of them propagates through the others. Throughout, our focus is on sex and gender differences within these four perspectives. We therefore begin with a brief exploration of what sex and gender mean, and why the distinction matters for biomechanics research.

Abstract: Early and accurate classification of melanoma is essential for improving patient outcomes and supporting clinical decision-making. Although numerous predictive models have been proposed, comparisons between classical statistical approaches and modern machine learning algorithms are often limited by heterogeneous analytical workflows and inconsistent validation strategies. This study aimed to compare the predictive performance of classical statistical and machine learning models for melanoma classification using a fully reproducible analytical framework. A retrospective observational study was conducted using the publicly available BCN20000 dermoscopic dataset from the ISIC Archive [1,2]. After standardized data preprocessing, four routinely available clinical variables (age, sex, anatomical site and melanocytic status) were used to develop Logistic Regression, Generalized Additive Models, Random Forest and Extreme Gradient Boosting (XGBoost) classifiers. All models were trained and evaluated using the same stratified training/testing split and their performance was assessed through discrimination, calibration and SHAP explainability analysis. Machine learning models, particularly XGBoost and Random Forest, achieved superior predictive performance compared with conventional statistical approaches, while patient age emerged as the most influential predictor of malignancy. The proposed framework provides a transparent and reproducible approach for objectively comparing predictive models and supports the development of accurate, interpretable and reproducible clinical decision-support systems for melanoma classification.

Abstract: Organic Electrochemical Transistors (OECTs) have emerged as continuous monitoring sensors for in-vitro models. Source and Measure Units (SMUs) are typical instruments used to simultaneously control input and output signals on OECTs. Conventional SMUs are unsuitable for in-vitro application due to their limited channels count, large footprint and high cost. This work presents the design and the development of a multichannel, portable and cost-effective SMU platform for the continuous and long-term characterization of OECT biosensors. The hardware is based on an electronic board interfacing via customized connection to a six-OECT set-up, integrated in a standard culture plate and operated within a cell incubator. System functionality is managed by Arduino Nano microcontroller through ad-hoc software, enabling automated characterization of the OECTs. The platform characterization performance is evaluated by comparing the I – V characteristics and transient responses of the tested OECTs with those obtained by the commercial Keysight B2912A SMU. Long-term and continuous operation is verified by a 30 minutes-sampling, five days measurement. The assessed features of the platform and the achieved results suggest that the proposed prototype is more appropriate than conventional SMUs for OECT-based in-vitro biosensing. The presented solution has the potential to support biological research by offering an operator-independent and in-vitro compatible apparatus.

Abstract: Adipose-derived stem cells (ADSCs) have garnered substantial attention within regenerative medicine due to their ready availability, ease of isolation, and potent immunomodulatory and tissue-repair capabilities. Growing evidence indicates that ADSC therapeutic efficacy stems predominantly from their secreted extracellular vesicles (EVs), rather than from direct cellular engraftment. However, clinical application of native ADSC-derived EVs (ADSC-EVs) is impeded by insufficient yield, inadequate targeting precision, and rapid systemic clearance. Recent advances in engineering technologies—including genetic modification, membrane functionalization, cargo encapsulation, and biomimetic integration—have provided innovative solutions to improve the production, stability, homing efficiency, and therapeutic potency of ADSC-EVs. In this review, we provide a systematic overview of the various engineered approach platforms for ADSC-EVs with a focus on parental cell reprogramming, cargo engineering, membrane protein engineering and hybrid or EV-mimetic platforms. We also report recent advances in engineered ADSC-EVs for multiple biomedical applications, including tissue regeneration, immunomodulation, anti-fibrotic therapies, cancer treatment and radiation-induced injury repair. We review key translational foibles surrounding the lack of standardization, reproducible production, quality assurance and safety. Lastly, we will provide perspectives on future directions to accelerate ADSC-EV clinical translation. Artificial smart cargo optimization, next-generation biomimetic targeting platforms and standardized GMP-compliant production systems offer the most promising prospects to overcome current translational bottlenecks.

Abstract: A microcirculation was developed to study live endothelium under peristaltic flow. Vascular channel molds were fabricated with photolithography. Channels were formed with polydimethylsiloxane and seeded with human endothelial cells. A peristaltic pump circulated growth media over cells for 24 hours with varying shear stress and volumetric flow. After 24 hours, cells were live-imaged with differential interference contrast microscopy or fixed and stained with fluorophore-conjugated phalloidin and 4’,6-diamidino-2’-phenylindole,dihydrochloride. Endothelial cell alignment was measured with purpose-written MATLAB code. Endothelial cell length and nucleus size were measured with semi-automated Zen and Image J analysis. Compared to a static no flow condition (0 Pa shear stress), significant endothelial cell alignment was observed during flow of 0.23 mL/min (2 Pa, p=0.0011) and flow of 0.46 mL/min (4 Pa, p=0.0004). Cell length (0 Pa, 50±14 µm; 2 Pa, 99±26 µm; 4 Pa, 110±30 µm, p< 0.0001) and nucleus area (0 Pa, 0.10±0.02 pixels2; 2 Pa, 11±0.01 pixels2; 4 Pa, 13±0.02 pixels2, p< 0.0001) increased significantly whereas nucleus circularity (0 Pa, 0.86±0.06; 2 Pa, 0.84±0.04; 4 Pa, 0.85±0.03, p=0.07) trended toward decrease. An ex vivo microcirculation was developed for laboratory experimentation. This relatively simple and cost-effective system may have translational utility to investigate human endothelial cell biology during flow.

Abstract: (1) Introduction: Stroke remains a leading cause of long-term disability, with 80–90% of survivors experiencing gait disturbances. Functional electrical stimulation (FES) is widely used to improve motor recovery and gait; however, its effectiveness remains uncertain because outcomes vary across studies. (2) Objectives: To evaluate the effectiveness of FES, alone or combined with rehabilitation strategies, for gait recovery in individuals with chronic stroke. (3) Review Summary: A systematic search of the PubMed/MEDLINE, ClinicalTrials.gov, Cochrane Central Register of Controlled Trials (Cochrane CENTRAL), and PEDro (Physiotherapy Evidence Database) identified randomized controlled trials published between 2016 and 2026. Adults with chronic stroke receiving lower-limb FES, alone or combined with conventional rehabilitation, were included. Outcomes included gait performance, balance, motor function, and activities of daily living. Methodological quality was assessed using the Cochrane Risk of Bias 2 tool. Five studies involving 109 participants met the inclusion criteria. FES improved gait speed, balance, lower-limb motor function, and functional independence, particularly when combined with task-specific rehabilitation. Improvements in coordination, neuromuscular activation, and corticospinal excitability also supported its role in motor recovery. (4) Conclusions: FES is a relevant adjunctive intervention for post-stroke rehabilitation. Although methodological heterogeneity remains, current evidence supports its potential to improve gait and functional outcomes.

Abstract: Rapid detection of organophosphorus pesticides (OPs) is urgently needed. A recombinant organophosphorus hydrolase (OPH) was expressed in E. coli and immobilized onto polyaniline nanofibers (PANF) and magnetic particles (PAMP) via adsorption–crosslinking (EAC) or adsorption–precipitation–crosslinking (EAPC). The immobilized OPH was characterized and integrated into an electrochemical biosensor for methyl-paraoxon detection. The EAPC method on PAMP gave the highest activity recovery (80.4%). Immobilized OPH showed optimal activity at pH 11 and 55 °C, with improved thermal stability (50% activity retained after 2 h at 60 °C) and storage stability (>80% after 30 days). The biosensor exhibited a linear response to methyl-paraoxon from 1 to 5000 μmol·L⁻¹ (R² > 0.99), with optimal detection at pH 12 and a scan rate of 150 mV·s⁻¹. The EAPC-PAMP immobilization strategy effectively stabilizes OPH, and the resulting biosensor offers a fast, sensitive, and stable platform for OP residue analysis.

Abstract: The growing deterioration of water resources and the energy requirements of conven-tional wastewater treatment technologies have increased interest in systems that combine organic matter removal with bioelectrochemical conversion. This study evaluated a la-boratory-scale dual-chamber microbial fuel cell (MFC) operated with real wastewater using Scenedesmus acutus (S. acutus) and a native microbial consortium as anodic bio-catalysts at 25 and 30 °C. The system was assessed through continuous monitoring of voltage, pH, temperature, and CH₄, H₂, and CO₂ signals in the anodic headspace, together with physicochemical characterization and COD removal.COD removal efficiencies of 33.7 and 30.3% were obtained for S. acutus at 25 and 30 °C, respectively, whereas the native microbial consortium achieved 29.8 and 43.4% removal under the same conditions. The consortium at 30 °C showed the most favorable combi-nation of COD removal and electrical response, whereas S. acutus at 30 °C reached the highest maximum voltage and stored energy, although with greater signal variability. The CH₄, H₂, and CO₂ signals differed among conditions and were consistent with the possible participation of fermentative and methanogenic processes alongside electrogenic activity, although gas production rates and the contribution of individual pathways were not quantified. Overall, the results demonstrate the operational feasibility of the pro-posed MFC for coupling wastewater treatment with a measurable electrical response and support further evaluation of native microbial consortia as anodic biocatalysts.

Abstract: Background and Objectives: The COVID-19 pandemic highlighted persistent limitations of personal protective equipment (PPE), particularly the difficulty of maintaining rapid and repeated hand hygiene during uninterrupted patient care. To address this challenge, we developed and performed a preliminary benchtop validation of a lightweight proof-of-concept face shield integrating an on-demand disinfectant reservoir for immediate point-of-care hand sanitisation. Materials and Methods: A multidisciplinary clinical–engineering team designed a dual-function wearable system consisting of a polyethylene terephthalate glycol-modified (PETG) visor (23 × 23 cm; 0.20 mm) and a modular disinfectant reservoir with capacities of 25, 50, 75, and 100 mL. The prototype was developed using computer-aided design, fused deposition modelling three-dimensional printing, and PETG thermoforming. A touch-activated self-sealing valve delivered approximately 3 mL of disinfectant per activation. No human participants were involved. Benchtop testing evaluated dispensing consistency, valve resealing, leak-tightness, attachment stability, chemical compatibility with alcohol-based disinfectants, and mechanical durability under repeated fill–drain cycles. Ergonomic assessment was limited to biomechanical modelling of mass distribution and estimated cervical torque. Results: The visor weighed 13.4 g, while the complete system weighed 103.4 g when empty and 203.4 g when loaded with 100 mL of disinfectant. Estimated cervical torque increased from 0.084 N·m to 0.282 N·m under maximum-load conditions. The prototype demonstrated reproducible disinfectant dispensing within predefined acceptance criteria, reliable valve resealing, absence of continuous leakage, secure reservoir attachment, and preserved PETG stability following exposure to alcohol-based disinfectants. Minor early dispensing variability and superficial contact wear were observed but did not impair device functionality. The estimated prototype material cost was approximately €10–20 (US$11–22). Conclusions: This proof-of-concept study demonstrates the technical feasibility of integrating facial protection with immediate on-demand hand sanitisation within a single reusable wearable device. The prototype achieved predefined engineering and performance endpoints during benchtop testing while maintaining low estimated material costs and a modular design. Further studies involving human participants, microbiological evaluation, ergonomic validation, and clinical usability testing are required before clinical implementation can be considered.

Abstract: Wearable glucose monitoring demands ultra-low-power local processing, but conventional neural networks rely on energy-intensive multiply–accumulate (MAC) operations that limit battery life. This study shows that a Spiking Neural Network (SNN), built on a regression adapted Leaky Integrate-and-Fire (LIF) neuron, can estimate blood glucose from multi-frequency bioimpedance and auxiliary biosignals with clinical-grade accuracy at a fraction of the computational cost. Using data from 98 patients (717 measurements, eGluco3 device, Azambuja Hospital, Brusque, Brazil) evaluated by 5-fold walk-forward cross-validation under ISO 15197:2013, three main findings emerge. First, a new calibration method—the Patient Fingerprint, built from each patient’s first K sensor readings—outperforms conventional one-hot patient encoding (14.2 ± 2.6 mg/dL vs. 15.4 ± 3.3 mg/dL mean absolute error) while working for patients never seen during training, a capability one-hot encoding lacks entirely. Second, this fingerprint model reaches 100% of samples within Consensus Error Grid Zones A+B across all validation folds, meeting the clinical-safety threshold, and does so without requiring any demographic or clinical metadata—sensor history alone renders such records redundant. Third, replacing the analog input encoding with a multiplication-free rate-coding scheme removes all first-layer MAC operations at a cost of only 2.7 mg/dL additional error, defining a concrete, quantified accuracy–hardware trade-off for neuromorphic deployment. Together, these results demonstrate that SNNs offer a clinically validated, calibration-free, and computationally efficient path to continuous glucose estimation on embedded wearable devices.

Abstract: The bystander effect describes the induction of responses in non-targeted cells through cell signaling by directly stimulated cells. While this phenomenon has been extensively studied in the context of ionizing radiation, its occurrence following electrical stimulation (ES) remains poorly understood. Conditioned medium from N2a neuroblastoma cells exposed to voltage-controlled biphasic pulses at 500 mV/mm and 100 Hz induced neuronal differentiation in non-stimulated cells through an ES bystander effect. Bystander medium promoted morphological changes associated with neuronal differentiation, including increased neurite outgrowth and a reduction in the proliferation marker KI-67, indicating that the effects of ES extend to neighboring non-targeted cells. Molecular analysis revealed increased expression and secretion of interleukin-6 (IL-6) following ES, while neutralization of IL-6 receptor inhibited the effects of ES, highlighting the role of IL-6 as a key mediator of this effect. We provide the first evidence that ES promotes a differentiation-associated bystander effect mediated by IL-6.

Abstract: Polysomnography (PSG) remains the clinical gold standard for sleep staging, but its cost, obtrusiveness, and dependence on laboratory settings limit its scalability and ecological validity. Wearable devices that acquire photoplethysmography (PPG), accelerometry (ACC), skin temperature, and derived cardiovascular features offer an unobtrusive alternative, and a rapidly growing body of work has applied machine learning and deep learning to translate these signals into sleep stage predictions. The resulting literature is, however, highly heterogeneous in its input modalities, model architectures, datasets, and evaluation protocols, which complicates direct comparison and obscures the field’s true state of progress. This systematic review analyzes 34 studies published between 2018 and 2026, identified through a structured search of five scientific databases and screened according to PRISMA-consistent criteria. We organize the reviewed work into a taxonomy of three modeling paradigms— traditional machine learning with handcrafted features, end-to-end deep learning (CNN-based, temporal sequence, and hybrid CNN–RNN architectures), and transfer learning — and we characterize the benchmark datasets and recording devices on which they are evaluated. We further consolidate reported performance by classification granularity, explicitly separating evaluation protocols, and we analyze cross-dataset generalization, recurring error patterns, and modality effects. Across the corpus, binary sleep–wake accuracy exceeds 90% with richer signals, three-class agreement reaches Cohen’s κ ≈ 0.66–0.73 with multimodal input, and the best cross-subject four-class PPG models plateau at κ ≈ 0.74–0.78 (accuracy ≈ 84%) on MESA, while five-class staging remains the weakest, limited chiefly by N1 detection. Deep sleep, REM, and N1 are consistently the hardest stages, intra-subject evaluation systematically inflates reported agreement, and zero-shot transfer to structurally different consumer wearables remains the principal unsolved problem. We conclude that robust domain generalization, standardized wearable-specific benchmarks, and label-efficient learning are the key priorities for translating wearable sleep staging into reliable, clinically applicable monitoring.

Abstract: The reliability of unrelated-donor searches depends on high-resolution HLA typing, yet a large fraction of records in national stem-cell donor registries were generated at low or intermediate resolution and are therefore under-used in modern matching. Here we develop an Extreme Learning Machine (ELM) approach that upgrades low/mid- to high-resolution HLA data by learning the haplotype and diplotype structure of a national donor population and assigning the most probable high-resolution genotypes together with posterior probabilities. The model was trained on the Greek national registry (Hellenic Transplant Organization, established 2002; 117,345 donors, ~20% low-resolution) and validated on two independent Greek cohorts (ORAM, n = 20,100; GRPT, n = 4,353) using accuracy and call-rate metrics. The population-specific ELM achieved a per-locus accuracy of 70–94% (depending on the confidence threshold) with an overall call rate of 98.1%, recovering usable high-resolution information and increasing the proportion of registry donors usable in high-resolution matching. The method is fast, lightweight and population-tailored, complementing established expectation-maximisation imputation tools.

Abstract: Wrist-worn devices offer a practical means of monitoring gait, yet no validated end-to-end pipeline exists for deriving digital mobility outcomes (DMOs) including cadence, stride length (SL), and walking speed (WS) in people with multiple long-term conditions (MLTC, the coexistence of two or more long-term conditions). This study presents the first modular pipeline for wrist-worn devices for DMO estimation, validated in 45 older adults with MLTC (65–90 years), whose conditions spanned four multimorbidity clusters (cardiometabolic, painful conditions, pulmonary, and cancer), across laboratory tasks using stereophotogrammetry as the reference. Algorithms were selected independently for each block (gait sequence detection (GSD), initial contact detection (ICD), SL) using novel, fine-tuned, and adaptive versions of established methods developed on an independent cohort. Blocks were first validated independently before being integrated into a pipeline capturing cumulative error propagation. GSD achieved a recall of 0.90. ICD was robust across algorithms, with the best-performing algorithm achieving a recall of 0.77 and precision of 0.82. At the pipeline level, the best-performing pipeline achieved SL error of 0.15 m with near-zero bias, cadence absolute error of 8.58 steps/min, and WS absolute error of 0.14m/s. These findings support wrist-worn devices for objective gait assessment in multimorbid populations, establishing a validated open-source pipeline for real-world deployment.

Abstract: Extracellular vesicles (EVs), including exosomes and microvesicles, have emerged as promising therapeutic vectors and diagnostic biomarkers across various branches of biomedicine. However, the clinical translation of EV-based technologies remains constrained by persistent challenges in manufacturing: insufficient yield from primary cell sources, limited control over cargo composition, and the absence of scalable, standardized production platforms. Nanosecond pulsed electric fields (nsPEF) represent an emerging biophysical approach that can address several of these limitations. Unlike conventional electroporation, which targets the plasma membrane using microsecond-to-millisecond pulses, nsPEF delivers ultrashort (1-300 ns), high-amplitude (10-300 kV/cm) pulses that penetrate intracellularly to directly perturb endosomal membranes, the endoplasmic reticulum, and the multivesicular body (MVB) compartment, the very organelles where exosome biogenesis and cargo sorting happen. Through coordinated effects on intracellular calcium mobilization, cytoskeletal remodeling, SNARE-mediated membrane fusion, and phospholipid redistribution, nsPEF can stimulate rapid, non-lethal vesicle release, a process labeled as "electro-exocytosis." Emerging evidence suggests that nsPEF does not merely increase EV yield but actively modulates the proteomic, lipidomic, and nucleic acid composition of released vesicles, offering a potential route to cargo engineering. In addition, the same biophysical principles that drive electro-exocytosis can be exploited in reverse: nsPEF-mediated transient permeabilization of EVs membranes allows for post-isolation loading of exogenous therapeutic cargo, small molecules, nucleic acids, or proteins into pre-formed vesicles without destroying their structural integrity. This review discusses current knowledge on EVs biogenesis and release mechanisms, introduces the biophysical foundations of nsPEF-cell and nsPEF-membrane interactions, and, by evaluating the experimental evidence supporting nsPEF-driven EVs engineering, outlines a translational roadmap for the application and development of this technology toward clinical-grade EVs manufacturing.