IEEE Transactions on Biomedical Engineering

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.

1ObjectivePassive cavitation imaging and control has explored various beamforming algorithms to balance resolution, imaging artifacts, and computational speed. Optimizing these parameters is essential for clinical translation, as precise cavitation localization and dosage control are critical for Focused Ultrasound (FUS)-based targeted therapies, minimizing unintended tissue damage. Among commonly used methods, Delay-Sum-Integrate (DSI) and Robust Capon Beamforming (RCB) have demonstrated effectiveness but are limited by either significant artifacts or the need for extensive parameter tuning. MethodsThis work introduces Passive Acoustic Dynamic Differentiation and Mapping (PADAM), which adapts the Multiple Signal Classification algorithm to the time domain to improve cavitation localization. ResultsPADAM achieves up to a 6-fold improvement in lateral beam-width compared to RCB, and a 4-fold reduction in mean-square intensity of artifacts. It further unveils a novel physical insight: its input parameter dynamically gauges the richness of an incoming signals frequency content. This feature enables a more physically defined and intuitive parameter for distinguishing between stable and inertial cavitation based on spectral characteristics, simplifying parameter selection and enhancing the framework for cavitation monitoring and control. ConclusionWith its ability to improve resolution, reduce artifacts, and provide computational efficiency, PADAM represents a promising advancement for precise cavitation localization and therapy monitoring. SignificanceThis work introduces PADAM, a novel time-domain passive cavitation imaging method that offers superior resolution and artifact reduction compared to DSI and RCB. Its physically intuitive input parameter enables dynamic differentiation between stable and inertial cavitation, enhancing precision in the monitoring and control of FUS therapy.

Transcranial focused ultrasound is an emerging noninvasive neuromodulation technique that offers high spatial precision and the potential for deep brain penetration. However, due to skull-induced attenuation and acoustic aberrations, precisely stimulating deep brain regions in mice remains challenging. To address this challenge, this study introduces a machine-learning-based computational framework to optimize single-element transducer designs for accurate deep-brain targeting in a mouse model. This framework includes a surrogate model consisting of a Random Forest regressor and classifier, trained on acoustic simulation results to predict performance from design parameters. A total of 72 transducer designs were simulated across coronal and sagittal planes, systematically varying frequency (1-6 MHz), radius of curvature (5-7 mm), and f-number (0.58-1.0). Each design was evaluated using five performance metrics: focal length, focal shape, maximum pressure at the focal region, pressure maximum location, and sidelobe suppression. The surrogate models were then combined with the Non-Dominated Sorting Genetic Algorithm II (NSGA-II) to perform multi-objective optimization and identify high-performing transducer designs. The optimized design produced a compact, symmetric focal region and accurate energy delivery to deep targets, with minimal off-target exposure, even in complex skull anatomy. Results show that lower f-numbers, moderate radius of curvature, and higher frequencies facilitate precise deep brain targeting. Overall, this data-driven approach enables practical design of single-element transducers for deep-brain neuromodulation in mice and provides a framework for designing transcranial transducers for other brain targets, potentially accelerating the clinical translation of focused ultrasound technologies.

Gastrointestinal magnetic resonance imaging (MRI) provides rich spatiotemporal data about the volume and movement of the food inside the stomach, but does not directly report on the muscular activity of the stomach itself. Here we describe a novel approach to characterize the motility of the stomach wall that drives the volumetric changes of the gastric content. In this approach, a surface template was used as a deformable model of the stomach wall. A neural ordinary differential equation (ODE) was optimized to model a diffeomorphic flow that ascribed the deformation of the stomach wall to a continuous biomedical process. Driven by this diffeomorphic flow, the surface template of the stomach progressively changes its shape over time or between conditions, while preserving its topology and manifoldness. We tested this approach with MRI data collected from 10 Sprague Dawley rats under a lightly anesthetized condition. Our proposed approach allowed us to characterize gastric anatomy and motility with a surface coordinate system common to every individual. Anatomical and motility features could be characterized for each individual, and then compared and summarized across individuals for group-level analysis. As a result, high-resolution functional maps were generated to reveal the spatial, temporal, and spectral characteristics of muscle activity as well as the coordination of motor events across different gastric regions. The relationship between muscle thickness and gastric motility was also evaluated throughout the entire stomach wall. Such a structure-function relationship was used to delineate two distinctive functional regions of the stomach. These results demonstrate the efficacy of using GI-MRI to measure and model gastric anatomy and function. This approach described herein is expected to enable non-invasive and accurate mapping of gastric motility throughout the entire stomach for both preclinical and clinical studies.

Recent years have demonstrated the feasibility of using intracortical Brain-Machine Interfaces (iBMIs), by decoding thoughts, for communication and cursor control tasks. iBMIs are increasingly becoming wireless due to the risk of infection and mechanical failure, typically associated with percutaneous connections. The wireless communication itself, however, increases the power consumption further; with the total dissipation being strictly limited due to safety heating limits of cortical tissue. Since wireless power is typically proportional to the communication bandwidth, the output Bit Rate (BR) must be minimised. Whilst most iBMIs utilise Multi-Unit activity (MUA), i.e. spike events, and this in itself significantly reduces the output BR (compared to raw data), it still limits the scalability (number of channels) that can be achieved. As such, additional compression for MUA signals are essential for fully-implantable, high-information-bandwidth systems. To meet this need, this work proposes various hardware-efficient, ultra-low power MUA compression schemes. We investigate them in terms of their BRs and hardware requirements as a function of various on-implant conditions such as MUA Binning Period (BP) and number of channels. It was found that for BPs [≤] 10 ms, the delta-asynchronous method had the lowest total power and reduced the BR by almost an order of magnitude relative to classical methods (e.g. to approx. 151 bps/channel for a BP of 1 ms and 1000 channels on-implant.). However, at larger BPs the synchronous method performed best (e.g. approx. 29 bps/channel for a BP of 50 ms, independent of channel count). As such, this work can guide the choice of MUA data compression scheme for BMI applications, where the BR can be significantly reduced in hardware efficient ways. This enables the next generation of wireless iBMIs, with small implant sizes, high channel counts, low-power, and small hardware footprint. All code and results have been made publicly available.

This work aims to evaluate a grip strength model, developed using a piecewise linear function based on the Woods and Bigland-Ritchie EMG-force model, which correlates the relationship between measured grip force and muscular activity. The grip strength model is compared against the results derived from an upper limb musculoskeletal model. Experimental results demonstrate the models efficacy in estimating surface electromyography (sEMG) readings from force measurements, with a mean root mean square error (RMSE) of 0.2035 and a standard deviation of 0.1207 for muscle activation (dimension-less). Moreover, incorporating sEMG readings associated with grip force does not significantly affect the optimization of muscle activation in the upper arm, as evidenced by kinematic data analysis from dynamic tasks. This validation underscores the models potential to enhance musculoskeletal model-based motion analysis pipelines without distorting results. Consequently, this research emphasizes the prospect of integrating external models into existing human motion analysis frameworks, presenting promising implications for physical Human-Robot Interactions (pHRI).

Robotic exoskeletons require human control and decision making to switch between different locomotion modes, which can be inconvenient and cognitively demanding. To support the development of automated locomotion mode recognition systems (i.e., high-level controllers), we designed an environment recognition system using computer vision and deep learning. We collected over 5.6 million images of indoor and outdoor real-world walking environments using a wearable camera system, of which ~923,000 images were annotated using a 12-class hierarchical labelling architecture (called the ExoNet database). We then trained and tested the EfficientNetB0 convolutional neural network, designed for efficiency using neural architecture search, to predict the different walking environments. Our environment recognition system achieved ~73% image classification accuracy. While these preliminary results benchmark Efficient-NetB0 on the ExoNet database, further research is needed to compare different image classification algorithms to develop an accurate and real-time environment-adaptive locomotion mode recognition system for robotic exoskeleton control.

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.

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.

Super-resolution ultrasound microvessel imaging based on ultrasound localization microscopy (ULM) is an emerging imaging modality that is capable of resolving micron-scaled vessels deep into tissue. In practice, ULM is limited by the need for contrast injection, long data acquisition, and computationally expensive post-processing times. In this study, we present a contrast-free super-resolution Doppler (CS Doppler) technique that uses deep generative networks to achieve super-resolution with short data acquisition. The training dataset is comprised of spatiotemporal ultrafast ultrasound signals acquired from in vivo mouse brains, while the testing dataset includes in vivo mouse brain, chicken embryo chorioallantoic membrane (CAM), and healthy human subjects. The in vivo mouse imaging studies demonstrate that CS Doppler could achieve an approximate 2-fold improvement in spatial resolution when compared with conventional power Doppler. In addition, the microvascular images generated by CS Doppler showed good agreement with the corresponding ULM images as indicated by a structural similarity index of 0.7837 and a peak signal-to-noise ratio of 25.52. Moreover, CS Doppler was able to preserve the temporal profile of the blood flow (e.g., pulsatility) that is similar to conventional power Doppler. Finally, the generalizability of CS Doppler was demonstrated on testing data of different tissues using different imaging settings. The fast inference time of the proposed deep generative network also allows CS Doppler to be implemented for real-time imaging. These features of CS Doppler offer a practical, fast, and robust microvascular imaging solution for many preclinical and clinical applications of Doppler ultrasound.

Continuous long-term sensing of biopotential signals is vital to facilitate accurate diagnosis. The current state of the art in wearable health monitoring relies on radiative technology for communication. Due to their radiative nature, these systems result in lossy and inefficient transmission, limiting the devices life span. Human Body Communication has emerged as an energy-efficient secure communication modality, and literature has shown body communication to transmit biopotential signals at 100x lower power than traditional radiative technologies. Unlike radiative communication that uses airwaves, HBC, specifically Capacitive Electro-Quasistatic HBC (EQS-HBC), couple signals and confine them within the human body. In Capacitive EQS-HBC, the transmitter uses an electrode to modulate the body potential to transmit data. The modulation of body potential by HBC raises the following concerns. Will HBC transmissions affect the quality of biopotential signals sensed from the body? Additionally, since biopotential sensing systems commonly use Right Leg Drive (RLD) to bias body potential, there is also a concern if RLD can affect the quality of HBC transmissions. For the first time, our work studies the interactions between EQS-HBC and biopotential sensing. Our work is important since understanding HBC-RLD interactions is integral to developing EQS-HBC-based biosensors for Body Area Networks (BANs). For the studies, we conducted lab experiments and developed circuit theoretic models to back the experimental outcomes. We show that due to their higher frequency content and common-mode nature, HBC transmissions do not affect the differential sensing of low-frequency biopotential signals. We show that biopotential sensing using RLD affects HBC. RLD deteriorates the signal strength of HBC transmissions. We thus propose not to use RLD with HBC. We demonstrate our proposed solution by transmitting ECG signals using HBC with 96% correlation compared to the traditional wireless system at a fraction of the power.

The Jugular Venous Pulse (JVP) is a vital gauge of proper heart health, reflecting the venous pressure via the Jugular Vein observation. It offers crucial insights for discerning numerous cardiac and pulmonary conditions. Yet, its evaluation is often over-shadowed by the challenges in its process, especially in patients with neck obesity obstructing visibility. Although central venous catheterization provides an alternative, it is invasive and typically reserved for critical cases. Traditional JVP monitoring methods, both visual and via catheterization, present significant hurdles, limiting their frequent application despite their clinical significance. Therefore, there is a pressing need for a non-invasive, efficient JVP monitoring method accessible for home-based and hospitalized patients. Such a method could preempt numerous hospital admissions by offering early indicators. We introduce a non-invasive method using a frequency-modulated continuous wave (FMCW) radar for JVP estimation directly from the skin surface. Our signal processing technique involves an eigen beamforming method to enhance the signal-to-noise ratio for better estimation of JVP. By meticulously fine-tuning various parameters, we identified the optimal settings to enhance the JVP signal quality. In addition, we performed a detailed morphological analysis comparing the JVP and photoplethysmography signals. Our investigation effectively achieved signal localization within a Direction of Arrival (DoA) range from -20{degrees} to 20{degrees}. This initial study validates the effectiveness of using a 60 GHz far-field radar in measuring JVP. O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=110 SRC="FIGDIR/small/24308313v1_ufig1.gif" ALT="Figure 1"> View larger version (32K): org.highwire.dtl.DTLVardef@d95a7dorg.highwire.dtl.DTLVardef@1c3e6d4org.highwire.dtl.DTLVardef@678f28org.highwire.dtl.DTLVardef@e7a0d7_HPS_FORMAT_FIGEXP M_FIG C_FIG

Speckle is the major artifact in ultrasound imaging, and it is well-known that speckle can be reduced by compounding (averaging) images taken either at different frequencies or from different angles. By averaging images of a phantom taken over a frequency range [Formula] and a 90{degrees} span of angles, the combined speckle reduction is demonstrated to be ~ 9x compared to non-compounded images, while the reduction with frequency or angle averaging resulted in reductions of ~ 3x individually. The rf input to the transducer is altered to vary the sound frequency and the phantom is rotated with respect to the transducer to obtain different imaging angles. Numerical simulations of sound scattered by randomly distributed point scatterers showed quantitative agreement with the experiment. Using a commercial system, a 6x reduction in speckle is demonstrated imaging a human wrist. A robot arm is used to move the transducer along a circular path to acquire images at 9 angles separated by 10{degrees}. The commercial system does not allow direct control of the input to the transducer, so the broadband signal detected is Fourier filtered to obtain images at different frequencies with ~ 2x reduced frequency range. Images taken at different angles contain distortions from speed of sound variations and pressure induced by the probe. Two forms of non-rigid image registration are applied to correct for the distortions and create a higher resolution composite image. A design for achieving ~10x speckle reduction with essentially no loss in imaging speed is described.

ObjectiveIndividuals with autism may perform repetitive or stereotyped rhythmic movements known as "stereotypies" or "stims." Per first-person reports, stereotypies provide benefits including sensory and emotional regulation. These movements are not often convenient to engage in and for some may be self-injurious. We propose the design of a device that discreetly provides rhythmic stimulation through vibrating motors that are incorporated into a vest designed to be worn under clothing. MethodsWe designed a vest that provides sensory stimulation through vibrations from 18 coin motors arranged on the interior surface of a compression vest. The vibration frequency and duty cycle are controlled by the user via a Bluetooth smartphone application, allowing them to optimize the stimulation for their unique needs. The vest is designed to be worn discreetly under the users clothing. It is rechargeable and can be used at full power for up to 4 hours. We performed experiments to quantify target values for the vibration pressure, signal frequency, signal duty cycle, battery life, noise level, and temperature. ResultsOur device accurately controls frequency and duty cycle with marginal error and meets all our engineering requirements except for battery life. It also conforms to ethical and regulatory guidelines. ConclusionFuture work, such as variable device weight, pressure and temperature control, and vibration patterns synchronized to music, could refine the product and deliver more value to the user. Significance: We have designed a vest using vibration to provide rhythmic sensory input to supplement the benefits of stereotyped movements.

Magnetic resonance elastography (MRE)-based muscle force estimation methods have been proposed to estimate individual muscle forces based on measurements of shear wave speed and joint torque in multiple postures. To estimate both the slope and offset parameters of the relationship between shear wave speed and muscle force, it is necessary to collect measurements in a plethora of postures in case of substantial muscle redundancy. However, anisotropic MRE requires a structural and diffusion-tensor imaging scan in each posture, which is infeasible given the time constraints of MRI imaging. The objectives of this work were to develop a muscle force estimator with sufficient accuracy that would only require a limited set of postures, and to evaluate its effectiveness under a variety of measurement conditions. We developed a novel MRE-based muscle force estimator, which decouples shear wave speed into its active and passive components and solves for the slope and offset parameters, independently. We assessed the effectiveness of the proposed estimator under different simulated measurement conditions, with varying levels of noise, and compared it to the original MRE-based estimator. The proposed estimator results in a reduction in the estimation error for the offset parameter, for all muscles, without significant degradation in the estimation error for the slope parameter. However, the proposed muscle force estimator does not improve the goodness-of-fit or the cross-validation error compared to the original estimator. In conclusion, the proposed MRE-based muscle force estimator improves the estimation of muscle-specific parameters and may yield increased muscle force estimation performance.

Abdominal ultrasound is a crucial first-line diagnostic tool, yet its efficacy is inherently constrained by a strong dependency on operator skill, leading to significant inter-operator variability and limiting its widespread adoption. While robotic ultrasound systems with artificial intelligence have emerged to mitigate this, current approaches face critical limitations: they often rely on simulated or phantom data for training, which hampers generalizability, and employ control strategies that lack the flexibility to adapt to complex human anatomy. To address these challenges, this paper introduces a novel, end-to-end robotic system for autonomous abdominal scanning. Our methodology is systematic: it begins with coarse scan planning via RGB and 3D point cloud-based localization of key anatomical landmarks (e.g., xiphoid process). The core of our approach involves training a reinforcement learning scanning policy directly on live human volunteers, enabling the development of a strategy that is robust to anatomical diversity. This high-level policy is executed by a sophisticated hybrid force-position controller, enhanced with real-time normal vector calibration and 3D point cloud-based respiratory phase detection to dynamically adjust contact force, particularly during inspiration. This innovation proves critical for improving scan quality in subjects with higher Body Mass Index (BMI). Extensive validation against expert sonographers demonstrates that our system achieves high precision in standard plane recognition and superior scanning efficiency. Furthermore, to holistically assess performance where standard planes are difficult to acquire, we introduce a 3D reconstruction-based coverage metric. Results confirm that our system delivers strong adaptability and significantly improved consistency, marking a substantial step toward clinically viable, operator-independent ultrasound robotics.

ObjectiveUltrasound can detect individual motor unit (MU) activity during voluntary isometric contractions based on their subtle axial displacements. The detection pipeline, currently performed offline, is based on displacement velocity images and identifying the subtle axial displacements. This identification can preferably be made through a blind source separation (BSS) algorithm with the feasibility of translating the pipeline from offline to online. However, the question remains how to reduce the computational time for the BSS algorithm, which includes demixing tissue velocities from many different sources, e.g., the active MU displacements, arterial pulsations, bones, connective tissue, and noise. ApproachThis study proposes a fast velocity-based BSS (velBSS) algorithm suitable for online purposes that decomposes velocity images from low-force voluntary isometric contractions into spatiotemporal components associated with single MU activities. The proposed algorithm will be compared against stICA, i.e., the method used in previous papers, for various subjects, ultrasound- and EMG systems, where the latter acts as MU reference recordings. Main resultsWe found that the spatial and temporal correlation between the MU-associated components from velBSS and stICA was high (0.86 {+/-} 0.05 and 0.87 {+/-} 0.06). The spike-triggered averaged twitch responses (using the MU spike trains from EMG) had an extremely high correlation (0.99 {+/-} 0.01). In addition, the computational time for velBSS was at least 50 times less than for stICA. SignificanceThe present algorithm (velBSS) outperforms the currently available method (stICA). It provides a promising translation towards an online pipeline and will be important in the continued development of this research field of functional neuromuscular imaging.

ObjectiveUltrasound (US) images during a muscle contraction can be decoded into individual motor unit (MU) activity, i.e., trains of neural discharges from the spinal cord. However, current decoding algorithms assume a stationary mixing matrix, i.e. equal mechanical twitches at each discharge. This study aimed to investigate the accuracy of these approaches in non-ideal conditions when the mechanical twitches in response to neural discharges vary over time and are partially fused in tetanic contractions. MethodsWe performed an in silico experiment to study the decomposition accuracy for changes in simulation parameters, including the twitch waveforms, spatial territories, and motoneuron-driven activity. Then, we explored the consistency of the in silico findings with an in vivo experiment on the tibialis anterior muscle at varying contraction forces. ResultsA large population of MU spike trains across different excitatory drives, and noise levels could be identified. The identified MUs with varying twitch waveforms resulted in varying amplitudes of the estimated sources correlated with the ground truth twitch amplitudes. The identified spike trains had a wide range of firing rates, and the later recruited MUs with larger twitch amplitudes were easier to identify than those with small amplitudes. Finally, the in silico and in vivo results were consistent, and the method could identify MU spike trains in US images at least up to 40% of the maximal voluntary contraction force. ConclusionThe decoding method was accurate irrespective of the varying twitch-like shapes or the degree of twitch fusion, indicating robustness, important for neural interfacing applications.

Accurate detection of tumors is critical for the success of oncologic surgical intervention. Along with any modality for the localization of tumors, surgeons achieve important information with direct palpation as long as direct touch of the tumor is possible. However, often due to using surgical tools in confined cavities or employing robotic endeffectors, direct palpation is not possible. We propose a low-frequency vibro-acoustic vibration sensing method to identify tumors based on their mechanical properties. The tumor detection problem was cast into a binary classification as a complementary modality. The method involves a wavelet-based multilayer perceptron neural network that is trained in a supervised manner. Some phantoms of healthy tissue with tumor model inclusion were used for experiments and data collection. From the 120 overall number of experiments, 18 data (15% of the whole data samples) were used as test data to evaluate the performance. The results show an 83.3% accuracy related to the confusion matrix, reflecting the model performance on previously unseen data. Performance of the classifier was evaluated using the confusion matrix and the receiver operating characteristic (ROC). It is discussed that the advantage of the proposed system is that the sensor is not directly touching the specimen and can be integrated into the probe end due to its small dimension.

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.