IEEE Transactions on Biomedical Engineering

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.

Wearable devices for gravity balancing have high potential for impact across domains, including neuromotor rehabilitation and occupational systems. Devices made from compliant mechanisms, optimized to achieve specific compensation moments at target joints, have proven effective, but thus far have solely been optimized towards gravity compensation and not other wearability criteria. In this work, we propose a multi-objective optimization framework, based on particle swarm optimization, to design a soft, gravity balancing shoulder orthosis, while taking into account wearability constraints such as undesired loading directions and device size. Using this custom framework, we pursued multiple stages of orthosis design and optimization, selecting multiple solutions to be translated to real-world prototypes. These solutions were realized via 3D printing with thermoplastic polyurethane and evaluated for mechanical performance on benchtop and in-vivo. In-vivo testing on 6 healthy individuals demonstrated relative reductions in muscle activity for the anterior deltoid and upper trapezius, by 53 % and 71 % respectively when operating the orthosis for static tasks within functional shoulder ranges of motion. Changes in muscle activation were also were observed across other muscles, including the posterior deltoid, as well as in dynamic tasks at different speeds.

Reliable interpretation of electrocardiograms (ECGs) requires precise identification of P, QRS, and T (PQRST) wave boundaries. However, it remains challenging due to noise, signal quality variability, and inherent morphological diversity particularly in recordings from children. This study systematically compares the performance of leading deep neural networks (DNN) and heuristic-based delineation algorithms on ambulatory single-lead ECG signals focusing on temporal accuracy. Experiments were conducted using the publicly available LUDB dataset and a private validation dataset comprising 21,759 annotated single-lead wave segments from 611 children recorded using KardiaMobile ECG sensor. DNN were first trained on the LUDB dataset and subsequently tested on the validation dataset. The delineation performance was assessed using Sensitivity (Se) and positive-predictive-value (P+) metrics. The best-performing heuristic based and DNN models reached Se and P+ of (98.9% vs 97.9%) for P, (99.8% vs 99.2%) for QRS, and (98.7% vs 95.9%) for T wave fiducials, respectively. The lowest standard-deviation (in ms) of wave onset/offset delineation was achieved by attention based 1DU-Net model; {+/-}16.6/{+/-}16.3 for P-wave, {+/-}14.0/{+/-}16.3 for QRS, and {+/-}26.3/{+/-}18.8 for T-wave, respectively. The findings indicate that optimized heuristic models can perform comparably to complex DNN, highlighting their efficiency and suitability for real-time ECG delineation in digital health monitoring applications.

Class imbalance in clinical electrocardiogram (ECG) datasets limits the diagnostic sensitivity of automated arrhythmia classifiers, particularly for rare but clinically significant beat types. We propose a three-stage hybrid generative pipeline that combines a spectral-guided conditional Variational Autoencoder (cVAE), a class-conditional latent Denoising Diffusion Probabilistic Model (DDPM), and a Quantum Latent Refinement (QLR) module built on parameterized quantum circuits to augment minority arrhythmia classes in the MIT-BIH Arrhythmia Database. The QLR module applies a bounded residual correction guided by Maximum Mean Discrepancy minimization to align synthetic latent distributions with real class-specific latent banks. A lightweight 1D MobileNetV2 classifier evaluated over five independent random seeds and four augmentation ratios serves as the downstream benchmark. Our findings establish latent diffusion augmentation as an effective strategy for imbalanced ECG classification and motivate further investigation of quantum-classical hybrid methods in cardiac diagnostics.

Head rotation is the leading cause of diffuse brain injuries from cycling accidents, with severe, long-term or even fatal consequences. Here, we present a novel helmet safety technology, the Release Layer System (RLS), designed to enhance conventional helmets and reduce the likelihood of such injuries. RLS is located on the outer side of the helmet and thus gets impacted first. The force of the impact activates a rolling mechanism triggering the release of an outer polycarbonate panel, thereby dispersing and transforming a substantial portion of the incident rotational energy. To evaluate the effectiveness of the technology, we conducted oblique impact tests on three popular helmet types, in conventional and RLS-equipped configurations, at three impact locations. RLS-equipped helmets reduced Peak Angular Velocity (PAV) by 57-66%, averaged across impact locations, compared to their conventional counterparts. This corresponds to a 68-86% reduction in the probability of an AIS2+ brain injury, as estimated by the Brain Injury Criterion. The most notable improvement was observed at the pYrot location (front impacts, mid-sagittal plane), with up to 85% PAV reduction. Testing across headforms further demonstrated the effectiveness of the technology in mitigating head rotation irrespective of variations in evaluation setups. This work introduces a novel mechanism for rotational impact mitigation and provides evidence of its potential benefits compared with conventional helmets. As an outer-layer approach, RLS may offer an alternative pathway for managing rotational kinematics in future helmet designs.

The clitoris is one of the least studied organs of the human body. The detailed anatomy of the clitoris is challenging to address through a gross dissection, as most of its parts are embedded internally, surrounded by pubic bone and several pelvic organs. While clinical imaging methods such as magnetic resonance imaging can capture the gross 3D morphology, they lack the spatial resolution required to resolve the detailed structures. In this study, we generated micron-scale computed tomography images of the female pelvises, leveraging a synchrotron radiation X-ray source. This unique data revealed the complex trajectory of the dorsal nerve of the clitoris, the main sensory nerve of the clitoris. Notably, the nerve trunks within the clitoral glans were revealed, with the maximum diameter ranging from 0.2 to 0.7 mm. They showed a tree-like branching pattern projecting towards the surface of the glans. We also revealed that some branches of the dorsal nerve of the clitoris ramify to innervate the clitoral hood and mons pubis. Finally, the posterior labial nerve, a branch of the perineal nerves, was shown to innervate the surroundings of the clitoris and the labial structures. These findings have an immediate impact on operations performed around the vulva area, such as gender-affirmation surgery and reconstruction surgery after genital mutilation.

Spectral similarity is often judged with a single metric such as RMSE, yet this can be misleading: physically different errors can produce similar scores. This is a critical limitation for computational biomechanics, where spectral agreement underpins both model validation and machine-learning loss design. Here, we develop a multi-metric framework for objective spectral biofidelity and test whether it better captures meaningful disagreement across complex frequency-domain responses. We evaluated 12 complementary similarity metrics, including CORA and ISO/TS 18571, using controlled spectral perturbations that mimic common real-world deviations such as resonance shifts, localized spikes, and broadband tilts. We then applied the framework to an SBI-tuned finite-element middle-ear model to assess convergence with training dataset size and robustness to measurement noise across repeated stochastic runs. No single metric performed reliably across all distortion types. Shape-based metrics tracked resonance morphology but could miss vertical scaling, whereas MaxError remained important for narrowband anomalies that smoother metrics underweighted. CORA and ISO 18571 did not consistently outperform simpler metrics. Rank aggregation using Borda count provided a robust consensus across metrics, enabling objective identification of training-data saturation and noise thresholds beyond which similarity rankings became unstable. These results show that spectral biofidelity cannot be reduced to a single norm. A multi-metric consensus provides a clearer and more physically meaningful basis for comparing experimental and simulated spectra, and offers a more defensible foundation for data-fidelity terms in physics-informed and simulation-based machine learning.

Brain-Computer interfaces (BCIs) offer a link between neural signals and external computation, enabling control of devices for the purposes of restoring function to motor-affected individuals and enhancing capabilities of a wider set of populations. Electroencephalography (EEG) offers a high temporal resolution for dynamic and potential real-time feedback for non-invasive systems. However, its practical efficacy remains limited due to low spatial resolution and poor signal-to-noise ratio, leading to insufficient decoding accuracy and unintuitive control paradigms that hinder reliable user interaction. In this study, we present a framework for an online EEG foundation model by creating a custom foundation model through spectrogram reconstruction of compact temporal windows and online constraints during pretraining. We evaluate the performance of the model in a challenging control paradigm of single-arm, directional motor imagery with dynamic movements for guided and free movement cursor control tasks. Our foundation model approach achieved a final average accuracy of 51.3% during a goal-oriented guided control task. This represents a 15.8% increase over a conventional deep learning framework and a 26.3% increase above chance level, evaluated in a cohort of 11 human participants. During the free movement task, the foundation model invoked a higher rate of completion and lower completion times. Furthermore, the custom EEG foundation model demonstrated superior adaptability from same-session finetuning and indicated an enhanced capability to assist subject learning. These findings highlight the potential of EEG foundation models to support more robust and intuitive non-invasive BCI systems, providing a promising modelling framework for future BCI development.

Multimodal medical image analysis exploits complementary information from multiple data sources (e.g., multi contrast Magnetic Resonance Imaging (MRI), Diffusion Tensor Imaging (DTI), and Positron Emission Tomography (PET)) to enhance diagnostic accuracy and support clinical decision making. Central to this process is the learning of robust representations that capture both modality invariant and modality specific features, which can then be leveraged for downstream tasks such as MRI segmentation and normative modeling of population level variation and individual deviations. However, learning robust and generalizable representations becomes particularly challenging in the presence of missing modalities and heterogeneous data distributions. Most existing methods address this challenge primarily from a statistical perspective, yet they lack a theoretical understanding of the underlying geometric behavior such as how probability mass is allocated across modalities. In this paper, we introduce a generalized geometric perspective for multimodal representation learning grounded in the concept of barycenters, which unifies a broad class of existing methods under a common theoretical perspective. Building on this barycentric formulation, we propose a novel approach that leverages generalized Wasserstein barycenters with hierarchical modality specific priors to better preserve the geometry of unimodal distributions and enhance representation quality. We evaluated our framework on two key multimodal tasks brain tumor MRI segmentation and normative modeling demonstrating consistent improvements over a variety of multimodal approaches. Our results highlight the potential of scalable, theoretically grounded approaches to advance robust and generalizable representation learning in medical imaging applications.

Resting-state electroencephalography (rs-EEG) offers a cost effective and portable alternative to conventional neuroimaging for dementia screening, yet the lengthy, multichannel nature of rs-EEG makes learning robust representations challenging. Convolutional and Transformer based architectures dominate current deep learning based approaches, but often struggle with long-range dependencies and may not properly preserve channel-dependent features. In this work, we propose EEG-ChiMamba, a state space model based architecture designed for the classification of mild cognitive impairment (MCI) and dementia from normal controls using raw channel-independent rs-EEG signals. Our method decouples channel-wise representation learning from modeling cross-channel interactions and leverages Mamba layers for effective long-sequence modeling. We evaluate our method on the Chung-Ang University EEG dataset (CAUEEG) with 1,155 subjects, the largest public rs-EEG dataset for challenging MCI and dementia differential diagnosis. We achieve a 3-class accuracy of 57.65% using a strict subject-wise split, and relate task-specific features learned by our model as revealed by feature occlusion-based explainability techniques to clinical literature, highlighting that state space models can facilitate interpretable and scalable clinical rs-EEG screening tools for cognitive degeneration. The code for the study is publicly available at: https://github.com/HealthX-Lab/EEG-ChiMamba

Reverse correlation is a widely-used and well-established method for probing latent perceptual representations in which subjects render subjective preference responses to ambiguous stimuli. Stimuli are purposefully designed to have no direct relationship with the target representation (e.g., they are randomly-generated), a property which makes each individual stimulus minimally informative toward reconstructing the target, and often difficult to interpret for subjects. As a result, a large number of stimulus-response pairs must be gathered from a given subject in order for reconstructions to be of sufficient quality, making the task fatiguing. Recent work has demonstrated that the number of trials needed can be substantially reduced using a compressive sensing framework that incorporates the assumption that the target representation can be sparsely represented in some basis into the reconstruction process. Here, we introduce an alternative method that incorporates the sparsity assumption directly into stimulus generation, which holds promise not only for improving efficiency, but also for improving the interpretability of stimuli from subjects perspective. We develop this new method as a mathematical variation of the compressive sensing approach, before conducting one simulation study and two human subjects experiments to assess the benefits of this method to reconstruction quality, sample size efficiency, and subjective interpretability. Results show that sparse stimulus generation improves all three of these areas relative to conventional reverse correlation approaches, and also relative to compressive sensing in most conditions.

Goal-oriented semantic communication has recently emerged in wireless sensor-actuator networks, emphasizing the meaning and relevance of information over raw data delivery, thereby enabling resource-efficient telecommunication. This paradigm offers significant benefits for intra-body or implantable sensor-actuator networks, including dramatic reductions in bandwidth requirements, latency, and power consumption. In this paper, we address a patch-based energy-efficient anomaly detection method for smart capsule endoscopy. We propose a deep learningbased algorithm that employs the similarity between features extracted from measured images and a reference (normal) image as the detection metric. The algorithm is evaluated using a clinical dataset of capsule-captured images, combined with a simulated intra-body channel model. The results demonstrate that even with only 60% of the transmission power (relative to a standard link design for QPSK modulation) and 65% of the light intensity, the probability of anomaly detection remains above 85%, and it gradually improves as power and illumination levels increase. This improvement translates into a potential battery life extension of over 43%. The findings highlight the potential of semanticaware, energy-efficient intra-body devices for more sustainable and effective medical interventions.

Three-dimensional (3D) handheld photoacoustic tomography typically relies on bulky and expensive external positioning trackers to correct motion artifacts, which severely limits its clinical flexibility and accessibility. To address this challenge, we present PA-SfM, a tracker-free framework that leverages exclusively single-modality photoacoustic data for both sensor pose recovery and high-fidelity 3D reconstruction via differentiable acoustic radiation modeling. Unlike traditional Structure-from-Motion (SfM) methods that formulate pose estimation as a geometry-driven optimization over visual features, PA-SfM integrates the acoustic wave equation into a differentiable programming pipeline. By leveraging a high-performance, GPU-accelerated acoustic radiation kernel, the framework simultaneously optimizes the 3D photoacoustic source distribution and the sensor array pose via gradient descent. To ensure robust convergence in freehand scenarios, we introduce a coarse-to-fine optimization strategy that incorporates geometric consistency checks and rigid-body constraints to eliminate motion outliers. We validated the proposed method through both numerical simulations and in-vivo rat experiments. The results demonstrate that PA-SfM achieves sub-millimeter positioning accuracy and restores high-resolution 3D vascular structures comparable to ground-truth benchmarks, offering a low-cost, softwaredefined solution for clinical freehand photoacoustic imaging. The source code is publicly available at https://github.com/JaegerCQ/PA-SfM.

Conventional temporal interference stimulation (TI, TIS, or tTIS) leverages two pairs of electrodes to induce an interfering electrical field in the brain. Both computational and experimental studies show that TI can stimulate deep brain regions without significantly affecting shallow areas. While promising, optimization of the locations and dosages on these two pairs of electrodes for maximal focal modulation remains computationally challenging. We are the first to propose two arrays of electrodes instead of two or multiple pairs of electrodes to boost modulation focality. However, the optimization algorithm outputs too many electrodes with overlaps across two frequencies, making it difficult to implement in practice. Based on recent progress in developing multi-channel TI devices and computational work on TI optimization, here we again advocate two-array TI, but with solid software and hardware evidence to show the feasibility. Specifically, we show that the latest optimization algorithm for two-pair TI innately works for two-array TI with the fastest speed (under 30s) among all major algorithms. With a similar amount of electrodes, two-array TI could achieve better focality (3.03 cm) at the hippocampus even than TI using up to 16 pairs of electrodes (3.19 cm) that takes days to optimize. We also show a hardware implementation of two-array TI using 10 electrodes on our 8-channel TI device. We argue that two-pair TI is only preferred when one does not care about modulation focality and promote two-array TI for its advantages in focality and lower cost in terms of both optimization time and electrodes needed. We restate the focality-intensity tradeoff but in the context of TI and provide a first voxel-level map of achievable focality and modulation strength by TI in the MNI-152 head template. We hope this work will pave the way for future adoptions of two-array TI for more focal non-invasive deep brain stimulation.

Neurodegenerative diseases such as Mild Cognitive Impairment (MCI), Multiple Sclerosis (MS), Parkinson s Disease (PD), and Amyotrophic Lateral Sclerosis (ALS) are becoming more prevalent. Each of these diseases, despite its specific pathophysiological mechanisms, leads to widespread reorganization of brain activity. However, the corresponding neurophysiological signatures of these changes have been elusive. As a consequence, to date, it is not possible to effectively distinguish these diseases from neurophysiological data alone. This work uses Magnetoencephalography (MEG) resting-state data, combined with interpretable machine learning techniques, to support differential diagnosis. We expand on previous work and design a Riemannian geometry-based classification pipeline. The pipeline is fed with typical connectivity metrics, such as covariance or correlation matrices. To maintain interpretability while reducing feature dimensionality, we introduce a classifier-independent feature selection procedure that uses effect sizes derived from the Kruskal-Wallis test. The ensemble classification pipeline, called REDDI, achieved a mean balanced accuracy of 0.81 (+/-0.04) across five folds, representing a 13% improvement over the state-of-the-art, while remaining clinically transparent. As such, our approach achieves reliable, interpretable, data-driven, operator-independent decision-support tools in Neurology.

Deep Brain Stimulation (DBS) is an effective therapy for Parkinson's disease (PD), but clinical programming of stimulation parameters remains a time-consuming process largely guided by subjective symptom assessment. The increasing availability of sensing-enabled neurostimulators and wearable motion sensors provides an opportunity to introduce objective biomarkers into DBS titration. In this work, we present DBSgram, a multimodal framework designed to support data-driven DBS programming by integrating neurophysiological and kinematic measurements acquired during routine clinical titration. The proposed system combines subthalamic nucleus local field potential (STN-LFP) recordings from sensing-enabled neurostimulators with hand kinematic data acquired using wearable inertial measurement units (IMUs). A two-stage synchronization strategy aligns independent data streams from implanted and wearable devices, followed by automated signal processing pipelines for extracting electrophysiological and motor biomarkers. Patient-specific beta-band power is derived from LFP recordings, while tremor, rigidity, and bradykinesia metrics are computed from multi-axis IMU signals using symptom-specific processing algorithms. These synchronized features are then integrated into the DBSgram visualization framework, which maps stimulation amplitude to simultaneous changes in neural activity and objective motor performance. The framework was implemented in a standardized 40-minute clinical titration protocol conducted in a cohort of 18 PD patients implanted with sensing-enabled DBS systems. We present here the analysis of aligned multimodal datasets from different patients to demonstrate proof-of-concept feasibility. The resulting DBSgram visualizations capture stimulation-dependent suppression of pathological beta activity alongside quantitative motor improvements, enabling intuitive identification of patient-specific therapeutic windows. These results demonstrate the technical feasibility of integrating implanted neurophysiological recordings with wearable kinematic sensing during DBS programming. By providing synchronized physiological and motor biomarkers within a unified framework, the DBSgram approach may support more objective and data-driven DBS titration, and contribute to future closed-loop neuromodulation strategies.

Objective detection of evoked potentials (EPs) is central to digital diagnostics in hearing assessment and clinical neurophysiology, yet current approaches remain time-intensive and sensitive to inter-individual noise variability. Many existing detection methods rely on population-based assumptions or computationally demanding procedures, limiting robustness and efficiency in real-world clinical settings. We present Fmpi, a digital EP detection framework enabling individualised, real-time response detection through analytical modelling of the spectral colour and temporal dynamics of background noise within each recording. Using extensive simulations and large-scale human electroencephalography datasets spanning brainstem, steady-state, and cortical EPs recorded in adults and infants, we demonstrate performance comparable or superior to state-of-the-art bootstrapped methods while operating at a fraction of the computational cost and maintaining well-controlled sensitivity with improved specificity. Importantly, Fmpi incorporates a futility detection mechanism enabling early termination of uninformative recordings, reducing testing time without compromising diagnostic reliability.

Deep learning model classification on large datasets is often limited in countries with restricted computational resources. While transfer learning can offset these limitations, standard architectures often maintain a high memory footprint. This study introduces HybridNet-XR, a memory-efficient and computationally lightweight hybrid convolutional neural network (CNN) designed to bridge the domain gap in medical radiography using autonomous self-supervised learning protocols. The HybridNet-XR architecture integrates depthwise separable convolutions for parameter reduction, residual connections for gradient stability, and aggressive early downsampling to minimize the video RAM (VRAM) footprint. We evaluated several training paradigms, including teacher-free self-supervised learning (SSL-SimCLR), teacher-led knowledge distillation (KD), and domain-gap (DG) adaptation. Each variant was pre-trained on ImageNet-1k subsets and fine-tuned on the ChestX6 multi-class dataset. Model interpretability was validated through gradient-weighted class activation mapping (Grad-CAM). The performance frontier analysis identified the HybridNet-XR-150-PW (Pre-warmed) as the optimal configuration, achieving a 93.38% average accuracy and 99% AUC while utilizing only 814.80 MB of VRAM. Regarding class-wise accuracy, this variant significantly outperformed standard MobileNetV2 and teacher-led models in critical diagnostic categories, notably Covid-19 (97.98%) and Emphysema (96.80%). Grad-CAM visualizations confirmed that the teacher-free pre-warming phase allows the model to develop sharper, anatomically grounded focus on pathological landmarks compared to distilled models. Specialized pre-warming schedules offer a viable, computationally autonomous alternative to knowledge distillation for medical imaging. By eliminating the requirement for high-performance teacher models, HybridNet-XR provides a robust and trustworthy diagnostic foundation suitable for clinical deployment in resource-constrained environments. Author summaryTraditional deep learning models for medical imaging are often too large for the low-power computers available in many global health settings. We developed a new model to bridge this computational gap. We designed HybridNet-XR, a highly efficient AI architecture, and trained it using a "teacher-free" method that doesnt require a massive supercomputer. We found a specific version (H-XR150-PW) that provides high accuracy while using very little memory. Our results show that high-performance diagnostic AI can be deployed on standard, low-cost hardware. Furthermore, using visual heatmaps (Grad-CAM), we proved that the AI correctly identifies medical landmarks like lung opacities, ensuring it is safe and reliable for real-world clinical use.

Objective: Abdominal aortic aneurysms (AAA) affect more than 1% of adults over 50 and carry significant mortality risk. Current surveillance relies on intermittent imaging (ultrasound or MRI) at 6--24 month intervals, which may miss rapid growth acceleration between visits. We investigate the feasibility of continuous aneurysm diameter tracking using peripheral pulse waves, like those detected by photoplethysmography (PPG) devices. Approach: We use a simplified one-dimensional hemodynamic model that simulates pulse wave propagation from the heart to the pedal digital artery. We first demonstrate diameter estimation when the hemodynamic model parameters defining systemic circulation are known within bounds for an individual, aggregating thousands of observations over hours or days. We then address the more challenging scenario where systemic circulation parameters are only known to be within wider population-level physiological bounds, using a sequential Monte Carlo approach that combines ensemble MCMC with Kalman filtering to marginalise over unknown parameters while tracking the aneurysm diameter. Both approaches are validated through 12-month tracking simulations with constant and accelerating aneurysm growth rates. Main results: While single-observation diameter estimation is fundamentally limited by noise and confounding variables, aggregating 1,600 measurements under baseline noise conditions reduces diameter uncertainty to 0.8~mm when patient-specific hemodynamic parameters are known within bounds. In this setting, tracking simulations across eight virtual patients achieve average root-mean-square error (RMSE) of $\sim$0.3~mm. When systemic parameters are known only within population-level bounds, joint Bayesian estimation over the full parameter space achieves a median RMSE of 0.65~mm (1.4$\pm$0.3~mm, mean$\pm$standard error) across 50 virtual patients, remaining within clinically relevant ranges despite the underlying parameters being only partially identifiable. Significance: These physically-grounded, computational results suggest that peripheral pulse wave monitoring through wearable PPG sensors could complement traditional imaging for aneurysm surveillance, potentially enabling earlier detection of growth acceleration and more timely clinical intervention.

Bidirectional interfaces combined with neural de-coding algorithms are essential for closed-loop (CL) neuromodulation, enabling simultaneous neural monitoring and responsive optogenetic stimulation. However, implementing these capabilities in compact wireless headstages for freely moving animals remains challenging, as most existing platforms rely on tethered setups and external processors to execute computationally intensive decoders. This work presents the design and optimization of a neural decoder integrated into a bidirectional wireless system for CL optogenetic experiments in rodents. The proposed platform combines 32-channel electrophysiological recording with neuromorphic feature extraction, dimensionality reduction, and a nonlinear support vector machine (NL-SVM) decoder implemented on a resource-constrained Spartan-6 FPGA. Temporal dynamics are captured using spike-count features and leaky integrators, while principal component analysis (PCA) reduces the feature space to six components, enabling sub-millisecond inference with minimal memory and power requirements. Model size is further reduced using k-means clustering during training to limit the number of support vectors. Decoder performance was validated using datasets from non-human primate and rat motor cortex recordings. The proposed decoder achieved accuracy comparable to convolutional neural networks (R2 =0.85 vs. 0.87) and outperformed Wiener filters (R2 = 0.81) while requiring significantly fewer computational resources. The full system was further demonstrated in vivo through wireless closed-loop optogenetic stimulation in rats, achieving a variance accounted for (VAF) of 0.9148. Overall, this work introduces a versatile, fully self-contained, and resource-efficient platform for real-time untethered closed-loop neuroscience experiments.