Red blood cell (RBC) aggregation is an important hemorheological property that influences blood viscosity, microcirculation, and stored-blood quality. However, conventional measurements commonly require repeated flow cessation or flow-rate modulation, limiting continuous monitoring and providing little information on spatial heterogeneity. This study presents a microfluidic platform for the spatiotemporal mapping of RBC aggregation during continuous blood flow. Multiple high-resistance side chambers connected to a main channel create low-shear-rate regions for aggregation while maintaining high shear in the main channel for RBC disaggregation. Flow rates and shear rates are evaluated using a hydraulic circuit model, numerical simulation, and micro-PIV measurements. An aggregation index (AI) map is introduced to quantify spatial and temporal changes in the side chambers. AI remains high and stable at flow rates of 0.5~1 mL/h. RBC aggregation increases significantly at concentrations of 15 mg/mL or with more dextran solution (Cdex). At the selected flow rate of 1 mL/h and Cdex = 40 mg/mL, the 95% confidence interval of AI is 0.477~0.600 for the proposed method, compared with 0.364~0.460 for the previous method. The proposed AI value is approximately 30% higher than that obtained using the previous method. Moreover, AI decreases progressively during four weeks of RBC storage. These findings demonstrate continuous and multiple-location detection of RBC aggregation and support the use of this platform for assessing hemorheological alterations and storage-induced RBC deterioration.
A decline in olfactory ability represents one of the earliest signs of Alzheimer’s disease (AD) and can be valuable information for early diagnosis at the stage of mild cognitive impairment (MCI). Nevertheless, the underlying neurophysiological mechanisms of olfactory impairment have not been systematically studied and, so far, have not been applied to make objective diagnoses using electroencephalography (EEG). To fill this gap, this proof-of-concept study investigates the possibility of using olfactory-evoked EEG complexity to discriminate between healthy subjects (HSs) and MCI patients. To this purpose, a publicly available olfactory oddball EEG-recording dataset was considered. First, a strategy for cleaning the EEG signals was implemented and applied, including exclusion of participants, channels, and epochs affected by substantial artifacts and noise. Then, a dedicated preprocessing pipeline was implemented: in particular, the cleaned signals were partitioned into three temporal intervals according to the stimulus onsets (i.e., pre-stimulus, early post-stimulus, and late post-stimulus periods). For each window of interest, a novel metric—namely, the Multivariate Multiscale Multi-Frequency Entropy (M3FrEn)—was computed across 10 temporal scales. The obtained results showed significant main effects of group and stimulus, as well as a significant group-by-stimulus interaction across all scales, as assessed by linear mixed-effects models. Post hoc analysis revealed a significantly reduced stimulus-related entropy modulation in MCI subjects compared to healthy controls across several early post-stimulus scales, with the strongest effect at scale 6 (adjusted p=0.0189, Hedges’g=−2.88). An exploratory, fully nested subject-level classification analysis, in which feature selection was performed independently within each cross-validation fold, achieved an accuracy of approximately 92% with all classifiers, consistently relying on the early post-stimulus feature. These preliminary findings suggest that M3FrEn captures olfactory-related EEG alterations in MCI, providing proof-of-concept evidence for its potential as a candidate biomarker.
In this study, we used the SERS method for the first time to measure the spectra of four polysaccharide markers of fungal infections: linear β-(1→3)- and β-(1→6)-linked D-glucans, branched mannan of Candida albicans and galactomannan of Aspergillus fumigatus. Aqueous solutions of the polysaccharides were studied in concentrations from 10 pg/mL to 100 μg/mL. The spectra were analyzed using machine learning methods: principal component analysis for data visualization and partial least squares with a ridge regularizer, which were used to construct metrics reflecting the accuracy of substance recognition relative to each other. The spectral changes with varying analyte concentration were observed and stable calibration has been achieved. Subsequent measurements of fungal polysaccharides in the presence of a physiological concentration of human serum albumin (45 mg/mL), used to model blood serum, enabled accurate analyte detection in a clinically relevant concentration range of 10 pg/mL to 100 ng/mL. In this case, the calibration dependence was calculated using the partial least squares method with the L1-regularizer. Blind testing was evaluated using a train-derived applicability-domain criterion based on the disagreement between the model prediction and an independent concentration estimate.
Monitoring bacterial contamination on surfaces is critical for hygiene and safety assurance in food, healthcare, and pharmaceutical settings, although routine methods still remain slow and laboratory dependent. Here, we report a portable respirometric platform based on sealed sensor sachets incorporating optical oxygen sensors to rapidly detect and quantify total aerobic viable counts (TVC) from swabbed surfaces. Following standardized surface swabbing, samples were incubated in the sachets and oxygen depletion kinetics were recorded with a handheld reader and microbial activity was inferred from oxygen consumption. Reference quantification was obtained by serial dilution and aerobic plate counting, enabling direct benchmarking of the respirometric readout against an industry-accepted culture method such as ISO 4833:2013. The platform demonstrated strong agreement with plate counts (R2 > 0.94), achieving a median detection limit of 2.69 log10 CFU/cm2 and a dynamic range of 0–6 log10 CFU/cm2 for surface-associated microbial loads. The sensor system was also applied in a semi-industrial setting in a meat processing plant. Across replicate measurements, the assay provided consistent kinetic signatures and quantitative outputs, suitable for rapid and practical decision-making. Compared with traditional culture-based approaches (requiring up to 72 h), the respirometric sachets delivered actionable results within 10 h using a portable, low-infrastructure workflow, supporting rapid on-site hygiene verification and sanitation control.
Sensitive detection of treatment-associated Raman spectral alterations in tumor tissues remains challenging, particularly when such changes are not readily apparent from conventional morphological evaluation. Here, we developed a label-free Raman biosensing strategy combined with explainable machine learning to characterise treatment-associated spectral signatures in melanoma tumours. A B16-F10 melanoma-bearing mouse model was used to compare untreated and PBS-treated controls with cohorts receiving immune checkpoint blockade, anti-angiogenic intervention, or combination therapy. Raman spectra were acquired from multiple spatial regions of melanoma tissues and analyzed using nonlinear dimensionality reduction, supervised classification, and SHAP-based feature interpretation. Although cohort-averaged spectra showed substantial overlap, multivariate analysis revealed treatment-dependent spectral organization, with the combination-treatment cohort showing the most compact and distinguishable spectral profile. Supervised models, including convolutional neural networks, support vector machines, and k-nearest neighbors, further supported the reproducibility of treatment-associated Raman signatures when evaluated using mouse-level validation strategies. SHAP analysis identified discriminative Raman features mainly located within lipid, phospholipid, ester, protein, and collagen-associated vibrational domains, suggesting potential contributions from metabolic- and extracellular-matrix-related biochemical components to treatment-associated spectral discrimination. These findings indicate that Raman spectroscopy integrated with explainable machine learning provides a sensitive, label-free method for distinguishing treatment-associated spectral differences among melanoma tissues. The proposed approach may serve as a complementary spectroscopic tool alongside conventional histological and molecular analyses for investigating treatment-associated tissue-state alterations.
Precise quantification of respiratory pathogen transmission is urgently needed to underpin disease surveillance and support diagnostic decision-making in indoor healthcare settings. Although numerical simulations are widely used to study macro-scale transmission, the key physical parameters governing droplet dynamics remain insufficiently understood. This study develops a multi-scale transmission model to evaluate the effects of droplet evaporation, sedimentation, and ventilation on viral transmission. Based on the Wells evaporation–sedimentation theory, a time-varying model is formulated incorporating droplet size distribution, environmental humidity, and ventilation conditions, with analytical expressions derived for concentration distributions across different respiratory activities (breathing, speaking, coughing, and sneezing). Results show that droplet size is decisive for transmission distance and lower relative humidity significantly extends sedimentation range. Small coughing droplets can travel up to 2.5 m, while the bimodal sneezing distribution (1.5 µm and 74 µm) generates high-concentration zones up to 2 m. To validate the model, a molecular communication testbed is developed as a biosensing-oriented platform integrating transmitters, receivers, and configurable ventilation modules with real-time sensing capabilities, enabling systematic verification under varied breathing modes, humidity, and ventilation conditions. Experimental results show strong agreement with theoretical predictions. This framework provides a quantitative basis for biosensing-enabled environmental monitoring and diagnostic-oriented risk assessment, informing ventilation optimization and infection control measures in indoor environments.
Rapid identification of free thiols is essential in pharmaceutical quality control and food safety. Herein, we demonstrate the sensing concept of “two signals from one event” by exploiting the intrinsic dual functionality of selenium (Se) nanomaterials to construct a signal-on, dual-mode sensing platform for free thiols. The single event is thiol-triggered seeded growth: free thiols interact with L-cysteine-modified Se seeds, driving seed fusion and crystallization into chiral trigonal Se structures. This same event yields two readouts simultaneously: a colorimetric response arising from particle size enlargement (UV-Vis red-shift) and a turn-on circular dichroism (CD) signal from long-range chiral ordering. The method achieves a detection threshold of 10 μmol/L for various free thiols and can be completed within 60 min from seed preparation to signal readout without complex instrumentation. It serves as a general indicator of total free thiols rather than distinguishing individual thiol species. Above the threshold, concentration-dependent color deepening and spectral redshifts provide semi-quantitative estimation of the concentration range. Its practicality was validated by detecting D-penicillamine in commercial tablets. This strategy offers an approach for on-site free thiol screening and highlights the potential of chiral inorganic nanomaterials in multi-mode sensing via chiral transfer and amplification.
Saxitoxin (STX) is a highly potent marine biotoxin, and trace contamination in aquatic environments can pose serious risks to human health. Therefore, reliable detection of low-concentration STX is crucial for water safety monitoring. Here, we developed a robust electrochemical aptasensor based on a gold nanoparticle/chitosan (AuNP/CS) conductive network for STX detection in freshwater samples. The chitosan matrix provides a three-dimensional scaffold for aptamer immobilization, while interconnected AuNPs create efficient electron-transfer pathways across the sensing interface. This integrated architecture improves interfacial conductivity and supports stable target-induced aptamer recognition. The aptasensor exhibits a linear response from 1 to 1000 nM and a limit of detection of 0.74 nM, with an apparent dissociation constant Kd = 70.29 ± 29.2 nM, together with high batch-to-batch consistency and long-term stability, retaining 93% of its initial response after 25 days. In spiked freshwater samples, the aptasensor achieved recoveries ranging from 99.10% to 111.33%, demonstrating the practical potential of this platform for monitoring STX contamination in aquatic environments.
Background: Rapid screening of acetylcholinesterase (AChE) inhibition is essential for monitoring exposure to organophosphate compounds, particularly in field settings where access to laboratory infrastructure is limited. This study aimed to develop a portable paper-based colorimetric biosensor for the semi-quantitative detection of AChE activity in blood samples. Methods: The biosensor was based on a pH-dependent color change adapted from the modified Edson method. The platform was first optimized using experimental models and then evaluated in human capillary blood samples collected under real field conditions. Reference serum cholinesterase activity was determined by a certified clinical laboratory and used for comparison with the colorimetric response of the biosensor. RGB (red, green, and blue) image analysis was performed in a subset of samples to digitally characterize the chromatic response of the device. Results: Samples with preserved AChE activity showed a visible color shift associated with substrate hydrolysis, whereas samples with reduced or inhibited activity maintained darker blue-dominant tones. RGB analysis of human samples revealed significant associations between AChE activity and the R and G channels, supporting the ability of the platform to distinguish between chromatic patterns associated with different ranges of enzymatic activity. Conclusions: The proposed platform demonstrated the feasibility of translating a pH-based AChE assay into a portable paper-based format for semi-quantitative visual and RGB-assisted screening. Its evaluation in human samples demonstrated the feasibility of its use under field conditions as a proof-of-concept. Further studies are needed to optimize its analytical performance and validate its application in larger and more diverse populations.
Nanoneedle arrays provide a promising interface for intracellular delivery, yet scalable control of array geometry and membrane penetration mechanics remains insufficiently understood. Here, we developed a rapid and scalable strategy for fabricating geometry-tunable silicon nanoneedle arrays. One-step SF6/O2 etching produced ordered arrays with center-to-center spacing of 1–5 μm, whereas pseudo-Bosch etching produced high-aspect-ratio (HAR) nanoneedles. Using a microwell-assisted cell-on-probe atomic force microscopy platform, we quantified the first penetration force, penetration probability, and number of penetration events at the single-cell level. For one-step-etching arrays, increasing spacing from 1 to 5 μm reduced the first penetration force from 34.87 ± 2.90 to 4.05 ± 0.30 nN and increased the penetration probability from 0.21 ± 0.03 to 0.87 ± 0.04. A phenomenological inverse-square model captured the force–spacing relationship, supporting an array-level load-sharing mechanism. Under identical vibration-assisted microfluidic conditions, the FITC-dextran-positive fraction increased from 22.8% for 1 μm arrays to 77.1% for 5 μm arrays, whereas 1 μm HAR arrays achieved 28.8%. These results identify array spacing as a key factor governing single-cell penetration and delivery and provide a mechanistic basis for nanoneedle-based biosensor and cell-interface design.
Hydrogen peroxide (H2O2) may be illegally added to milk to delay visible spoilage, producing concentration-dependent changes in its refractive index. This study numerically evaluates the optical response of a B-sil/Al/Al2O3/WS2 surface plasmon resonance (SPR) configuration to these refractive-index variations. The angular SPR response was calculated at λ = 633 nm using the transfer matrix method under TM-polarized illumination. The B-sil/Al/Al2O3/WS2 architecture was established through sequential evaluation of prism material, Al thickness, Al2O3 thickness, and 2D interfacial material, followed by assessment of its response to H2O2-associated refractive-index changes in milk. The final structure consisted of 70 nm Al, 25 nm Al2O3, and 0.80 nm WS2. Across the simulated H2O2 conditions, the resonance angle shifted from 85.65° to 86.11°, while the angular sensitivity ranged from 383.82 to 406.45°/RIU. The best balance among the evaluated metrics was obtained for H2O2-C2, with a sensitivity of 406.45°/RIU, QF of 4.65 RIU−1, FoM of 436.48 RIU−1, LoD of 1.23 × 10−5, and CSF of 436.56. The WS2-containing interface produced calculated angular shifts for small refractive-index variations in the milk sensing medium. Comparison with reported SPR systems for milk-related sensing showed a comparable angular sensitivity range, while QF and detection accuracy were limited by the broad resonance profile. These results characterize the theoretical optical response of the B-sil/Al/Al2O3/WS2 multilayer under H2O2-associated refractive-index changes but do not establish chemical selectivity toward H2O2. Experimental implementation would require an appropriate selective filtering or recognition strategy, together with validation of matrix effects and practical sensing performance.
Nanostructured carbon materials are low-dimensional systems relevant to oncology biosensing, with their utility arising from an electronic structure coupled to defect and edge states, a charge-transfer behavior and optical response that report molecular binding, an interfacial chemistry governing contact with the analyte, and, in selected cases, magnetic properties enabling manipulation and readout. In functional terms, these behaviors trace to specific quantum-relevant features—quantum confinement, edge and defect states, and the resulting size-dependent optical and charge-transfer responses—rather than to a generic quantum-material designation. This critical review treats nanocarbons as engineered biointerfaces whose performance is set by how the carbon surface behaves in biological fluid, how recognition chemistry is anchored, and how the binding event is transduced, with magnetic responsiveness, stability, reproducibility, and fabrication control as decisive constraints. The analysis separates three material classes—non-magnetic nanocarbon sensors, hybrid carbon–magnetic systems, and defect-associated or potentially metal-free magnetic carbons—while grading evidence as direct, adjacent, comparator-derived, or prospective. Directly, graphene, carbon nanotubes, carbon dots, graphene quantum dots, and magnetic carbon hybrids serve in electrochemical, optical and fluorescent, field-effect, and magnetic-assisted formats. Clinical translation, however, remains constrained by biofouling, protein-corona formation, matrix interference, unstable functionalization, batch variability, incomplete standardization, and scarce validation in real samples and patient cohorts. Metal-free or defect-associated magnetic carbons therefore warrant caution, remaining prospective platforms until the preservation of magnetic response, reproducible functionalization, matrix compatibility, safety, and measurable analytical advantage are directly demonstrated.
Behavioral investigation in zebrafish is essential for understanding adaptive responses, where learning represents a key process influenced by external stimuli. The applied stimulus plays a critical role in shaping such responses, therefore making physiologically relevant stimulation strategies important. Hydrodynamic stimuli represent one such modality, providing a natural and non-invasive means of activating mechanosensory responses in aquatic organisms, thereby enabling behavioral manipulation in microfluidic environments. To address this, a microfluidic assay was developed to generate controlled hydrodynamic environments by employing multiple S-shaped magnetic microactuators (SMMAs). Further, motions of these SMMAs were independently controlled to produce spatiotemporally varying vortical flow fields, enabling flow-induced transportation of zebrafish larvae. Flow dynamics were characterized by employing micro-particle image velocimetry (µPIV). Compared to the control condition, transportation time under microactuator-assisted guidance was significantly reduced, with a maximum improvement of 94.3% observed for a representative target zone. Building on this validated transport capability, training-dependent behavioral adaptation was quantified using latency under repeated hydrodynamic-training, where a reduction of 82.7% was achieved. Post-training assessment further demonstrated short-term retention of the acquired behavioral response followed by progressive extinction. These findings demonstrate that the proposed paradigm serves as a foundational behavioral assay leveraging hydrodynamic cues for studying adaptive responses in microfluidics.
A horseradish peroxidase (HRP)-based electrochemical biosensor was developed for melatonin (MLT) detection using a graphene-modified screen-printed electrode (G/SPE). HRP was immobilized at different temperatures (4 °C and 20 °C) and enzyme amounts (10 and 20 µL of 5 mg/mL HRP solution) to optimize the performance of the biosensors. FTIR analysis confirmed successful enzyme immobilization, while differential pulse voltammetry (DPV) showed a clear oxidation peak of MLT at 1.1 V with enhanced current responses for HRP-modified electrodes. The optimal biosensor, prepared at 4 °C with 10 µL enzyme of 5 mg/mL HRP solution, exhibited the best analytical performance, with a linear range of 0.4–3.6 µM, and a detection limit of 1.02 µM. Kinetic studies confirmed strong enzyme-substrate affinity, while the biosensors demonstrated excellent repeatability and a good stability. The method was successfully validated on pharmaceutical products, proving to be a reliable and an accurate tool for MLT quantification in real samples.
Fetal magnetocardiography (fMCG) provides direct non-invasive assessment of fetal cardiac electrophysiology, enabling detailed evaluation of cardiac rhythm, conduction, and repolarization. However, the clinical adoption of conventional fMCG has been limited by its reliance on superconducting quantum interference device (SQUID) systems, which require cryogenic cooling and specialized infrastructure. Optically pumped magnetometers (OPMs) have emerged as a promising cryogen-free alternative with the potential to broaden access to fetal electrophysiological assessment. This review summarizes the technological evolution and early clinical evaluation of OPM-based fMCG through an analysis of original in vivo human studies published up to June 2026. Twelve eligible studies were identified and synthesized narratively. Advances in sensor design, magnetic shielding, acquisition strategies, and signal-processing algorithms have enabled SQUID-comparable signal quality and cardiac interval measurements while substantially reducing cryogenic and infrastructure requirements. OPM-fMCG has demonstrated the potential to assess fetal cardiac time intervals, heart rate variability, fetal movement, and clinically important arrhythmias, including congenital long QT syndrome, atrioventricular block, and supraventricular and ventricular tachyarrhythmias. However, the available evidence remains dominated by small, single-centre studies, with relatively few fetuses affected by clinically significant arrhythmias. Prospective multicenter clinical validation, protocol standardization, independent replication, and regulatory evaluation are therefore required before OPM-fMCG can be integrated into routine diagnostic pathways for pregnancies requiring advanced fetal electrophysiological assessment.
Early detection of colorectal cancer (CRC) is challenging in low-resource settings because access to colonoscopy and centralized laboratory testing is limited. Urine-based metabolite biomarkers offer a non-invasive alternative for CRC triage, but translating a laboratory assay into a point-of-care (PoC) system requires standardized fluid handling, operator-independent timing, field-compatible reagents, and quantitative calibration. Here, we present a low-cost, semi-automated PoC platform integrating a validated, sequential, three-metabolite CRC-biomarker assay with robotic fluid handling, a motorized chromatographic module, an optical reader, Bluetooth electronics, and tablet-guided control. The platform measures urinary diacetylspermine and hippuric acid, with creatinine serving as a normalization reference, and reports absolute quantitative concentrations. Automated dilution, tube positioning, timed incubation, controlled column elution, and software-guided transfer reduce operator-dependent variation. Using pooled urine samples spiked at clinically relevant concentrations, the creatinine assay showed a strong quadratic response (0–50 mM, R2 = 0.998), as did the diacetylspermine assay (0–4 μM, R2 = 0.979), while the hippuric acid assay showed a linear response of R2 = 0.989. The RGB sensor tracked the concentration-dependent trend observed with a laboratory microplate reader (R2 = 0.968–0.999). Reagents reformulated as lyophilized or pre-weighed, vacuum-sealed kits withstood accelerated heat and humidity (45 °C/70% RH) testing and international shipping to the pilot site very well. At approximately USD 490 for the complete instrument (including tablet) and roughly USD 7.65 in consumables per screening test, the platform offers a practical, quantitative, and portable route for decentralized CRC screening.
Continuous physiological monitoring can support pilot-state assessment, but routine cockpit use requires sensing approaches that are both unobtrusive and physiologically reliable. Wearable ECG (wECG) provides stable cardiac recordings through skin-contact electrodes, whereas cockpit-integrated invisible ECG (iECG) can reduce user burden by acquiring signals through instrumented controls. However, iECG depends on intermittent hand contact and may show incomplete ECG morphology even when cardiac timing information is still preserved. This study proposes a window-based framework to assess the quality and task-specific usability of simultaneous wECG and iECG acquired during simulated flight. ECG data were collected in an Airbus A320 simulator from 14 volunteers, including experienced pilots and novices. The framework combines contact availability, R-peak reliability, PQRST morphology, and complementary signal quality indices into graded usability classes for heart rate (HR)-oriented monitoring. The iECG channel remained accessible for most of the analyzed recording time, with 93.8% of windows showing full or partial contact. Several windows classified as low quality by individual SQIs were retained as HR-usable when contact and R-peak timing remained reliable, indicating that single-metric rejection can be overly conservative for external-contact ECG. HR agreement supported the physiological relevance of the proposed classes: concordant wECG–iECG windows showed a mean absolute error (MAE) of 1.1 bpm, compared with 7.1 bpm for discordant windows. Bland–Altman analysis for iECG windows classified as usable for HR estimation showed a small mean bias of 1.4 bpm. These findings indicate that incomplete ECG morphology does not necessarily imply loss of HR usability, and that contact-aware, task-specific classification can preserve useful physiological information from unobtrusive cockpit interfaces.
Monitoring prognostic biomarkers is essential for evaluating cancer treatment efficacy in real time. Here, we present a rapid, proof-of-concept microfluidic assay for cancer biomarker quantification and functional therapy evaluation. The assay requires only 5 µL of sample, eliminates tedious sample manipulation, and uses commercial antibodies to track both biomarker concentration and functional activity. We demonstrated the successful detection of VEGF and EML4-ALK proteins in both lung cancer cell culture supernatant and serum-spiked samples, achieving a detection sensitivity 15 times greater than standard ELISA using the same antibody pairs for the respective antigens. Crucially, we showcase the platform’s distinct applicability to functional assays by monitoring dynamic phosphorylation changes in EML4-ALK following treatment with the tyrosine kinase inhibitor alectinib, successfully capturing a significant, rapid decrease in phosphorylation levels. At this preclinical stage, the platform demonstrates analytical and functional feasibility for highly sensitive and rapid biomarker monitoring, providing a foundation for future validation in patient-derived ALK-positive samples for therapeutic-response assessment.
Wearable brain-imaging devices have been developed to meet the growing demand in the healthcare industry for long-term monitoring of brain signals in natural conditions, such as monitoring brain diseases and emotions. However, conventional EEG and fNIRS (functional near-infrared spectroscopy) devices are often expensive, bulky and difficult to operate, making it difficult to monitor patients for long periods in natural conditions. To address these issues, this article proposes a low-cost, portable and multimodal wearable brain signal acquisition scheme. It combines EEG (electroencephalography) and fNIRS to reflect brain activity from different perspectives. In order to make it more wearable, a conductive rubber material is used as the electrode for the EEG. In this study, the corresponding experiments were used to verify the performance of the device. The first is the measurement of internal system noise, which satisfies the data acquisition of EEG and fNIRS at different gain levels. The α-rhythm experiment and the SSVEP (steady-state visual evoked potentials) experiment were used to validate the performance of EEG data acquisition. The performance of the fNIRS was verified by measuring changes in cerebral blood oxygen during breath-hold and breathing. In addition, by decomposing the raw fNIRS data with the VMD (variational mode decomposition) algorithm and performing correlation analysis, heart rate information was separated from the data. The performance of the proposed device was validated in the above experiments, confirming the feasibility of the design for multimodal data acquisition and meeting the requirements for portability and wearability. Furthermore, the proposed device was tested with 31 subjects (15 depressive subjects) to detect depression. Experiments proved the effectiveness of the multimodal signals, which outperformed single modal and surpassed EEG by 8.4% and fNIRS by 23.5%.
Recently, plant-based bioelectronic systems have been explored for environmental sensing applications. However, their intrinsically low electrical conductivity often limits signal sensitivity and measurement reliability. In this work, the electrothermal behavior of living Epipremnum aureum plants incorporating Ag/AgCl nanoparticles supported on nanocellulose was investigated. Electrical and thermal responses were simultaneously measured under controlled environmental conditions using external shunt resistances of 1, 10, 100, and 1000 Ω. Compared with the control without nanoparticle incorporation, the nanoparticle-incorporated plant exhibited stronger electrical responses and distinct electrothermal behavior over the studied temperature range. The measured signals showed nonlinear responses, temporal asymmetry, and resistance-dependent modulation, suggesting changes in charge transport within the plant tissues. Silver-enriched regions and the co-detection of chlorine within the nanoparticle-incorporated plant tissues were identified by environmental scanning electron microscopy and energy-dispersive X-ray spectroscopy. Five machine-learning regression models were trained to estimate temperature using the measured electrothermal voltage signals as predictors. The best-performing model, MLP FitRNet, achieved a mean absolute error of 0.598 °C, a root mean square error of 0.748 °C, and an R2 value of 0.974. These results demonstrate the potential of nanoparticle-incorporated biohybrid plant systems for electrothermal signal analysis and data-driven temperature estimation, while providing a foundation for future intelligent environmental monitoring applications.
