IEEE Transactions on Biomedical Engineering

The potential to decode handwriting trajectories from brain signals has yet to be fully explored in clinical brain-computer interfaces (BCIs). One of challenges remains that the clinical BCIs mostly rely on imaginary movement due to motor deficit of the subject, which often leads to misalignment with neural activity and impedes accurate decoding. Here, we recorded intracortical neural signals from a paralyzed patient during imaginary handwriting of Chinese characters, from which the trajectories of handwriting were reconstructed and translated into texts using machine learning approach. We introduced an innovated decoding framework that incorporates a novel loss function, DILATE, to accommodate both shape and temporal distortions between movement and neural activity to account for the misalignment issue. Our method reconstructed closely resembled and human-recognizable handwriting trajectories, outperforming the conventional mean square error loss by 10% of recognition rate. Moreover, the new decoding framework enabled effective multi-day data fusion, resulting in further 15% enhancement. With a dynamic time warping approach, the recognition rate achieved up to 91.1% within a 1000-character database. Additionally, we applied our method to a previous subject who imaged handwriting of English letters, showcasing its capability for single-trail trajectory reconstruction and 13.5% higher recognition outcomes. Altogether, these findings demonstrated a new decoding scheme for BCIs that could accurately reconstruct the imaginary handwriting trajectory. This advancement paves the way for a universal brain-to-text communication system that is applicable to any written language, marking a significant leap forward in the field of neural decoding and BCI technology.

Continuously monitoring the Electrocardiogram (ECG) is an essential tool for Cardiovascular Disease (CVD) patients. In low-resource countries, the hospitals and health centers do not have adequate ECG systems, and this unavailability exacerbates the patients health condition. Lack of skilled physicians, limited availability of continuous ECG monitoring devices, and their high prices, all lead to a higher CVD burden in the developing countries. To address these challenges, we present a low-cost, low-power, and wireless ECG monitoring system with deep learning-based automatic arrhythmia detection. Flexible fabric-based design and the wearable nature of the device enhances the patients comfort while facilitating continuous monitoring. An AD8232 chip is used for the ECG Analog Front-End (AFE) with two 450 mi-Ah Li-ion batteries for powering the device. The acquired ECG signal can be transmitted to a smart-device over Bluetooth and subsequently sent to a cloud server for analysis. A 1-D Convolutional Neural Network (CNN) based deep learning model is developed that provides an accuracy of 94.03% in classifying abnormal cardiac rhythm on the MIT-BIH Arrhythmia Database. Index TermsWearable ECG, deep learning, arrhythmia detection.

Every human body is different, however, current movement analysis does not reflect that, as it heavily relies on generic musculoskeletal models. Usually, these models are scaled to match the participants body segment lengths and body weight, but not taking individual body shape into account. This can lead to errors in the estimation of joint forces and torques, which are important to accurately estimate musculoskeletal variables. Thus, we developed a method to estimate body segment inertial parameters based on smartphone pictures. From the pictures, we reconstruct the body hull, estimate the skeletal shape and pose, and then estimate the distribution of bone, lean, and fatty tissues. We then segment the body hull and assign each tissue type a density, which is used to calculate the body segment inertial parameters. These personalizated models were validated with an experiment including gait and magnetic resonance imaging measurements. We found that our method leads to a reduction of residual forces of up to 14.9 % and a reduction of metabolic cost of up to 12.8 % when compared to generic musculoskeletal models. Furthermore, the shape-based inertial parameter personalization method creates participant-specific musculoskeletal models that are closer to the MRI-derived ground truth than generic models. To allow for the use of our method with existing data, we also introduce two new generic musculoskeletal models, which are based on the average standing body shapes and show similar joint moment outcomes, but less reduction of residual forces as the personalized models. Author summaryOur method adresses the challenge of personalizing inertial properties of musculoskeletal models. The existing state-of-the-art methods involve scaling a generic model to match the participants body weight and segment lengths or widths, which does not take individual body shape into account. To overcome these limitations, we developed a method to personalize musculoskeletal models based on smartphone pictures by reconstructing their body hull. We then estimate where tissue types, e.g. fat, muscle, and bone, are located in the body hull estimate inertial parameters for each segment. The personalized models were validated with an experiment including gait and magnetic resonance imaging measurements, and we found that our method leads to reduced residual forces, which is important for accurate musculoskeletal simulations. As our method leads to different joint torques when compared to generic musculoskeletal models, it can influence the interpretation of biomechanical analyses. Our method can be used whenever a participants body shape is known, or smartphone pictures can be taken. For retrospective studies, we also introduce two new generic musculoskeletal models, SIPP-generic-female and SIPP-generic-male, which are based on average standing body shapes.

This paper proposes a deep learning (DL) model for automatic sleep stage classification based on single-channel EEG data. The DL model features a convolutional neural network (CNN) and transformers. The model was designed to run on energy and memory-constrained devices for real-time operation with local processing. The Fpz-Cz EEG signals from a publicly available Sleep-EDF dataset are used to train and test the model. Four convolutional filter layers were used to extract features and reduce the data dimension. Then, transformers were utilized to learn the time-variant features of the data. To improve performance, we also implemented a subject specific training before the inference (i.e., prediction) stage. With the subject specific training, the F1 score was 0.91, 0.37, 0.84, 0.877, and 0.73 for wake, N1-N3, and rapid eye movement (REM) stages, respectively. The performance of the model was comparable to the state-of-the-art works with significantly greater computational costs. We tested a reduced-sized version of the proposed model on a low-cost Arduino Nano 33 BLE board and it was fully functional and accurate. In the future, a fully integrated wireless EEG sensor with edge DL will be developed for sleep research in pre-clinical and clinical experiments, such as real-time sleep modulation.

A number of recent and current efforts in brain machine interfaces are developing millimetre-sized wireless implants that achieve scalability in the number of recording channels by deploying a distributed swarm of devices. This trend poses two key challenges for the wireless power transfer: (1) the system as a whole needs to provide sufficient power to all devices regardless of their position and orientation; (2) each device needs to maintain a stable supply voltage autonomously. This work proposes two novel strategies towards addressing these challenges: a scalable resonator array to enhance inductive networks; and a self-regulated power management circuit for use in each independent mm-scale wireless device. The proposed passive 2-tier resonant array is shown to achieve an 11.9% average power transfer efficiency, with ultra-low variability of 1.77% across the network. The self-regulated power management unit then monitors and autonomously adjusts the supply voltage of each device to lie in the range between 1.7 V-1.9 V, providing both low-voltage and over-voltage protection.

Instrumented gait assessment in pediatric populations is often constrained by the complexity and lack of portability of traditional motion capture systems. In this article, we propose a deep learning approach utilizing a one-dimensional (1D) U-Net architecture to accurately estimate ankle and knee joint angles in the sagittal plane from surface electromyography (sEMG) signals. We analyzed data from the tibialis anterior and medial gastrocnemius of 25 typically developing children (ages 4-16) to evaluate the models performance and the influence of age-related gait maturation. The proposed 1D U-Net achieved high predictive accuracy for the ankle joint (RMSE: 3.6{degrees}) and the knee joint (RMSE: 4.1{circ}). Experimental results demonstrated that incorporating the toe-off event as a temporal marker significantly enhanced prediction stability during transitional gait phases. Furthermore, Statistical Parametric Mapping (SPM) was employed to identify systematic errors, which were primarily localized during initial contact and pre-swing but remained below clinically relevant thresholds. The findings reveal that prediction accuracy increases with age, reflecting more stable neuromotor patterns. This study demonstrates that a 1D U-Net can reliably decode lower-limb kinematics from sEMG alone, enabling the development of simplified, non-invasive, and portable pediatric gait assessment tools that can be integrated into the control strategies of assistive devices.

Real-time brain monitoring for neurosurgery and neuroscience research of natural behaviors demands portable imaging with high spatiotemporal resolution. Current technologies cannot simultaneously achieve the resolution, mobility, and real-time performance required. Here we present a miniaturized four-dimensional functional ultrasound system capturing volumetric brain hemodynamics in real-time using 3072 transceivers controlled by custom application-specific integrated circuits. The device achieves 450 Hz volumetric imaging up to 8 cm depth while maintaining a form factor suitable for direct cortical placement and potential sub-cranial implantation. We validated this technology across three clinical scenarios: through skull prosthesis, cranial defect, and during neurosurgery. The system mapped somatotopic finger representations and achieved decoding of individual finger movements during piano playing using machine learning, demonstrating single-trial detection. This portable platform establishes a new approach for brain monitoring bridging laboratory neuroscience and clinical applications, enabling research in natural behavioral settings, providing surgeons real-time hemodynamic feedback, and advancing brain-computer interface development.

Withdrawal StatementThe authors have withdrawn this manuscript because it contains fundamental errors and fabricated data. Therefore, the authors do not wish this work to be cited as reference for the project. If you have any questions, please contact the corresponding author.

In recent years, wearable electrocardiograms have risen in popularity as a solution for personal monitoring of heart activity. However, this technology has limitations in diagnostic capability and structural function monitoring. Meanwhile, auscultation of the heart remains a fundamental tool for physicians in diagnosis and monitoring of heart disease largely unaddressed in a convenient wearable format. The present work outlines a promising system currently under investigation, allowing user-initiated 10-second chest-sound recordings to be transmitted over Bluetooth-Low-Energy, with an innovative package design providing inherent noise reduction and a high signal-to-noise ratio. The device has been tested on healthy individuals, and system response has been validated against calibrated electrocardiogram recording equipment to analyze signal capture fidelity. O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=124 SRC="FIGDIR/small/601965v1_ufig1.gif" ALT="Figure 1"> View larger version (46K): org.highwire.dtl.DTLVardef@3605fdorg.highwire.dtl.DTLVardef@c362b3org.highwire.dtl.DTLVardef@184e799org.highwire.dtl.DTLVardef@811384_HPS_FORMAT_FIGEXP M_FIG C_FIG

Recognizing keyboard typing as a familiar, high information rate communication paradigm, we developed an intracortical brain computer interface (iBCI) typing neuroprosthesis providing bimanual QWERTY keyboard functionality for people with paralysis. Typing with this iBCI involves only attempted finger movements, which are decoded accurately with as few as 30 calibration sentences. Sentence decoding is improved using a 5-gram language model. This typing neuroprosthesis performed well for two iBCI clinical trial participants with tetraplegia - one with ALS and one with spinal cord injury. Typing speed is user-regulated, reaching 110 characters per minute, resulting in 22 words per minute with a word error rate of 1.6%. This resembles able-bodied typing accuracy and provides higher throughput than current state-of-the-art hand motor iBCI decoding. In summary, a typing neuroprosthesis decoding finger movements, provides an intuitive, familiar, and easy-to-learn paradigm for individuals with impaired communication due to paralysis.

This study investigates the deployability of commercially available impulse-radio ultra-wideband radar (UWB) sensors, in accurately detecting and analysing gait patterns during asymmetric overground locomotion. An adjustable knee brace was fitted on the right knee of 10 able-bodied participants, in five different confinement angles, during a 6-meter walking task to simulate asymmetric walking patterns. A computationally efficient signal processing framework extracts seven spatiotemporal gait parameters from six UWB radar signals, based on their joint Range-Doppler-Time representation. Gait asymmetry was quantified using primarily the symmetry ratio metric of step times. Validated against the gold standard motion capture system, radar-based gait parameters were estimated with 88.2-98.8% accuracy for all settings. By capturing step time symmetry ratios with 95.6{+/-}2.8% accuracy, the radar system can effectively distinguish between different gait impairment levels.

Transcranial focused ultrasound (tFUS) has emerged as a promising non-invasive modality for precision neuromodulation. However, the heterogeneous acoustic properties of the skull often induce phase aberrations that shift the ultrasound focus and compromise energy delivery. In this study, we developed and validated a phase-reversal based aberration correction method to enhance the targeting specificity of tFUS using a 128-element random array ultrasound transducer. Individual head models were constructed from T1-weighted magnetic resonance (MR) images and corresponding pseudo-computed tomography (pCT) data to accurately represent subject-specific skull geometries and the targeted left V5 (V5L) region. Acoustic simulations were conducted with the k-Wave toolbox by first acquiring free-field pressure waveforms and then recording the aberrated waveforms in the presence of the skull. The phase differences between these conditions were used to compute corrective delays for each transducer element. Quantitative evaluation using metrics such as focal overlap with the target region, axial focal positioning, and the delivered ultrasound energy demonstrated significant improvements: the overlap volume increased by 98.70%, mean axial positioning errors were reduced by up to 14.36%, and energy delivery to the target improved by 17.58%. We further demonstrated that the proposed approach outperforms the conventional ray-tracing methods. The results show that phase-reversal based aberration correction markedly increases the spatial targeting accuracy of tFUS and enhances the efficiency of focused ultrasound energy deposition for the customized random array transducer, paving a way for effective and personalized non-invasive neuromodulation therapies.

Recent advancements in data-driven methods have enabled real-time estimation of biomechanical states for exoskeleton control. While biological joint moments can be directly used to scale exoskeleton assistance, this approach is often suboptimal. An optimized mapping between biological joint moments and exoskeleton assistance could enhance end-to-end controllers based on the users physiological state. We introduce a flexible parametrization of biological moment-based control using delay, scaling, and shaping terms to transform joint moment estimates into commanded torque. We performed human-in-the-loop optimization, using metabolic cost to evaluate each iterations controller parameters, for 9 subjects across three ambulation modes: level walking at 1.1 m/s, 1.5 m/s, and 5{degrees} inclined walking. We evaluated three methods of exoskeleton control: 1. Personalized/Task Dependent, 2. Task Dependent/Non-personalized, and 3. Task Agnostic/Non-personalized. On average, our personalized approach provided the greatest benefit of 18.3% reduction in metabolic cost compared to walking without the exoskeleton, with the task dependent and task agnostic controllers producing similar reductions of 8.6% and 8.4%, respectively. Our results show that while generalizable, task agnostic control parameters can improve user energetics across cyclic tasks, fully personalized exoskeleton control parameters yield larger metabolic reductions, highlighting the value of personalizing exoskeleton assistance to users across many diverse tasks.

Biomechanical biofeedback has the potential to enhance rehabilitation by providing clinicians with objective evaluation of patient performances. As feedback systems often depend on expensive and sophisticated motion capture technologies, researchers explore computer vision-based alternatives. Existing methods suffer from substantial joint angle errors, particularly in the upper limb, and neglect the scapular movements. We developed an approach for detecting bony landmarks and performing refined upper-limb kinematics assessments using a single consumer-grade depth-sensing camera. Unlike other markerless methods, our model incorporates the scapula, offering comprehensive shoulder joint kinematics. Annotated images from eight participants were used to fine-tune a convolutional neural network, which was subsequently evaluated on a hand-cycling motion. Our method showed a strong agreement with a reference marker-based system, with 3D bony landmark detection errors averaging 5 mm. The resulting kinematics closely aligned with the reference system, maintaining acceptable joint angle errors ([~]6.3{degrees}). Furthermore, the algorithm could provide real-time bony landmark positions and joint kinematics at a rate of 50 Hz. This study highlights the potential of using a single consumer-grade depth-sensing camera combined with an acromial 3D-shape to accurately estimate upper-limb kinematics through bony landmark detection, paving the way for more accessible clinical assessments.

Optimizing assistive wearable devices is crucial for their efficacy and user adoption, yet state-of-the-art methods like Human-in-the-Loop Optimization (HILO) and biomechanical modeling face limitations. HILO is time-consuming and often restricted to optimizing control parameters, while inverse dynamics assumes invariant kinematics, which is unreliable for adaptive human-device interaction. Predictive simulation offers a powerful alternative, enabling computational exploration of design spaces. However, existing approaches often lack systematic optimization frameworks and rigorous validation against experimental data. To address this, we developed a Design Optimization Platform that integrates predictive simulations within a two-level optimization structure for personalizing assistive device design. This paper primarily validates the platforms predictive simulations against a publicly available dataset of the passive Biarticular Thigh Exosuit (BATEX), assessing its reliability. Our findings show that the model can sufficiently predict the kinematics and major muscle activations, except for the pelvis tilt and some biarticular muscles. The key finding is that successful identification of personalized optimal BATEX stiffness parameters needs acceptable prediction of metabolic cost trends, not their precise values. Our analysis further reveals that the models accuracy in predicting Vasti muscle activation in the baseline condition is a significant indicator of its success in predicting metabolic cost trends. This demonstrates that accurate prediction of performance trends is more important for effective simulation-based design optimization than perfect biomechanical accuracy, advancing targeted and efficient assistive device development.

Accurate interaction force estimation can play an important role in optimization human-robot interaction in exoskeleton. In this work, we propose a novel approach for system identification of exoskeleton dynamics in presence of interaction forces as a whole multi-body system regardless of gait phase or any assumption on human-exoskeleton interaction. We hanged the exoskeleton through a linear spring and excited the exoskeleton joints with chirp commands while measuring the exoskeleton-environment interaction force. Several structures of neural networks have been trained to model the exoskeleton passive dynamics and estimate the interaction force. Our testing results indicated that a deep neural network with 250 neurons and 10 time delays can obtain sufficiently accurate estimation of the interaction force, resulting in 1.23 of RMSE on Z-normalized applied torques and 0.89 of adjusted R2.

Many neurological conditions negatively affect a persons walking quality, which is a vital aspect of their quality of life. Gait quality, through the collection of spatiotemporal variables, can also help infer disease status; however, in-clinic access to these metrics is limited or cannot be assessed frequently enough to proactively monitor disease progression (i.e., improvement, maintenance, worsening). To address these limitations, we developed a framework that analyzes spatiotemporal gait metrics using healthy and neurodegenerative walking data collected from instrumented shoe insoles. The Insole Framework (IF) identifies ambulatory activities using an artificial neural network, identifies gait events using logic, fuses the inertial measurement unit (IMU) data, standardizes the analysis to every ten seconds, and calculates spatiotemporal metrics categorized into core, pace, percentage, and asymmetry metrics. Activity classification algorithms had excellent accuracy and F1-score ([≥] 93%). The spatiotemporal metrics obtained from the IF were validated against a gold standard motion capture system using ICCs, limits of agreement, and statistical testing. All core and pace metrics had good to excellent reliability and acceptable bias compared to the motion capture system, regardless of neurological function. Of the 19 spatiotemporal metrics assessed, system-independent statistical tests showed that similar population-level interpretations (i.e., one disagreement) and post-hoc differences (i.e., three disagreements) with similar levels of explained variance (absolute 2 difference between systems across all tests was 0.046) would be found regardless of the system used. The IF was considered valid and can appropriately capture ambulatory activities and spatiotemporal gait metrics in healthy, multiple sclerosis, and Parkinsons disease populations. Author SummaryGait assessments are used by clinicians to infer the severity and progression of neurological diseases. These assessments aim to quantify gross walking quality (i.e., patient perception, visual observations, speed, and distance) rather than the spatiotemporal metrics (e.g., double support time, stride length, cadence, etc.) that differentiate people from controls, conditions, and severity levels. Although spatiotemporal metrics can be powerful digital biomarkers to assess disease severity and monitor progression, traditional motion capture methods are limited due to high costs, the need for specialized expertise, time-consuming analysis/operations and infrequent patient collections. To overcome these limitations, we propose a framework that uses instrumented shoe insoles (inertial measurement unit + pressure) to identify activities and analyze gait. With our framework, gait assessments can be done several times a month in free-living conditions instead of infrequent clinical gait assessments, reducing healthcare barriers and promoting objective decision-making. This work describes our activity recognition, gait detection, and fusion methods and demonstrates our frameworks ability to produce results comparable to a gold-standard motion capture system in participants with multiple sclerosis, Parkinsons disease, and healthy individuals. Our Insole Framework is deemed valid due to high reliability, similar between-group interpretations across systems, and the activity recognition algorithms performance.

Wireless communication using electro-magnetic (EM) fields acts as the backbone for information exchange among wearable devices around the human body. However, for Implanted devices, EM fields incur high amount of absorption in the tissue, while alternative modes of transmission including ultrasound, optical and magnetoelectric methods result in large amount of transduction losses due to conversion of one form of energy to another, thereby increasing the overall end-to-end energy loss. To solve the challenge of wireless powering and communication in a brain implant with low end-end channel loss, we present Bi-Phasic Quasistatic Brain Communication (BP-QBC), achieving < 60dB worst-case end-to-end channel loss at a channel length of ~55mm, by using Electro-quasistatic (EQS) Signaling that avoids transduction losses due to no field-modality conversion. BP-QBC utilizes dipole coupling based signal transmission within the brain tissue using differential excitation in the transmitter (TX) and differential signal pick-up at the receiver (RX), while offering ~41X lower power w.r.t. traditional Galvanic Human Body Communication (G-HBC) at a carrier frequency of 1MHz, by blocking any DC current paths through the brain tissue. Since the electrical signal transfer through the human tissue is electro-quasistatic up to several 10s of MHz range, BP-QBC allows a scalable (bps-10Mbps) duty-cycled uplink (UL) from the implant to an external wearable. The power consumption in the BP-QBC TX is only 0.52 W at 1Mbps (with 1% duty cycling), which is within the range of harvested power in the downlink (DL) from a wearable hub to an implant through the EQS brain channel, with externally applied electric currents < 1/5th of ICNIRP safety limits. Furthermore, BP-QBC eliminates the need for sub-cranial interrogators/repeaters, as it offers better signal strength due to no field transduction. Such low end-to-end channel loss with high data rates enabled by a completely new modality of brain communication and powering has deep societal and scientific impact in the fields of neurobiological research, brain-machine interfaces, electroceuticals and connected healthcare.

This study presents a novel, computationally efficient training framework demonstrated through bio-signal processing on edge medical devices. The approach integrates conventional full training with an innovative {micro}-Training technique, wherein the encoder and decoder of a compact model remain frozen while only the middle layer is updated. This design is further enhanced by a novel Future-Guided Self-Distillation mechanism that leverages the models anticipated future state in training to boost performance and improve generalization on unseen data, using electrocardiogram (ECG) signals as the primary case study. Additionally, {micro}-Fine-Tuning facilitates ondevice adaptation under resource-constrained conditions. We validate our framework using in-sample data from the Telehealth Network of Minas Gerais (TNMG) and out-of-sample testing on the China Physiological Signal Challenge 2018 (CPSC) datasets. Experimental results demonstrate that our integrated strategy (combining full training, self-distilled {micro}-Training, and {micro}-Fine-Tuning) consistently matches or surpasses conventional methods while significantly improving computational efficiency and mitigating catastrophic forgetting. Deployment on Radxa Zero hardware underscores the approachs practical applicability and scalability. Moreover, a demonstration incorporating the proposed self-distilled {micro}-Training into standard training procedures reveals performance improvements. This highlights the techniques potential for broader applications beyond medical diagnostics and TinyML systems, paving the way for its integration into existing training mechanisms to elevate overall model performance.

Neurological diseases and neuropsychiatric disorders are often characterized by abnormal neural oscillations, such as exaggerated synchronization or suppression within a narrow frequency band and complex oscillation coupling which disrupt normal brain function and contribute to debilitating symptoms. Phase-dependent stimulation (PDS) offers a promising solution by synchronizing electrical stimulation with specific phases of neural oscillations, thereby enhancing therapeutic precision and efficacy. However, the widespread clinical adoption of PDS is hindered by technological challenges, including the need for accurate detection and prediction of neural oscillatory phases in real-time, stimulation management, stimulus artifact removal, fast communication, and adaptable hardware for dynamic neural environments. This study aims to address some of these challenges by leveraging adaptive System-on-Chip and Field-Programmable Gate Array (FPGA) technology, which offers the computational power and flexibility required for real-time neural signal processing and management. Specifically, we propose to optimize, integrate, and validate our PDS technique within this advanced hardware framework to develop a unified, closed-loop phase-dependent neuromodulation system. We evaluated our devices performance by assessing its latency and accuracy in targeting specific phases of stimulation on both simulated signals and intraoperative cortical and subcortical recordings. Our findings indicate that the device successfully sent stimulation commands in time with the occurrence of target phases with both high accuracy and low latency for extended time periods. This work has the potential to transform therapeutic approaches for disorders with well-described brain network dysfunction, offering a precise, adaptable, and safer alternative to traditional stimulation techniques.