Thirty years on from the introduction of deep brain stimulation as a therapy for Parkinson’s disease, adaptive deep brain stimulation (aDBS) is poised to transform neural stimulation for the treatment of motor disorders. Implementation of aDBS has been facilitated by the identification of biomarkers that reflect patients’ clinical states and the introduction of hardware allowing simultaneous sensing and stimulation. However, the most appropriate algorithms for aDBS remain unclear. This review provides an overview of control algorithms used in computational, animal and human studies of aDBS in Parkinson’s disease to date. Although shown to be effective and well-tolerated, only relatively simple aDBS algorithms have been used clinically. Computational studies have explored a wider range of more advanced algorithms which promise superior performance but need to be rigorously tested and compared against clinically-tested techniques before translating to patients. The need for comparative studies is further driven by advances in machine learning, which promise patient-specific programming solutions but require large datasets and have high computational requirements. Understanding the advantages and limitations of these different algorithms is critical for aDBS to reach its potential and enable personalized treatment to provide maximum patient benefit.
Microfluidic technology offers a transformative approach for biomedical diagnostics by enabling miniaturization and automation of analytical procedures. Its integration into Point-of-Care Testing (POCT) holds promise for rapid, decentralized, and accessible diagnostics. However, translating laboratory microfluidics into practical POCT devices faces key challenges, including the handling of complex clinical samples, integration of multiple functions on a single chip, ensuring reliable analytical performance in clinical deployment scenarios, and meeting regulatory requirements. This review explores the evolution of microfluidic technology and the unique requirements for POCT, while highlighting the main challenges in translating laboratory systems into practical applications. Representative case studies and recent technological advances are presented to illustrate progress and solutions, and future directions are discussed with a focus on strategies to bridge the gap between laboratory research and effective, real-world POCT deployment.
Ultrasound Computed Tomography (USCT) represents a paradigm shift in medical imaging, offering quantitative, high-resolution tissue characterization for diverse anatomical regions including breast, musculoskeletal system, brain, and lungs. By capturing the full ultrasonic wavefield through dedicated transducer arrays, USCT enables reconstruction of intrinsic tissue properties such as sound speed, attenuation, and acoustic impedance. Unlike conventional ultrasound, USCT provides three fundamental advantages: full-angle tomographic reconstruction, quantitative multi-parameter imaging capabilities, and operator-independent standardized acquisition—all while maintaining ultrasound’s inherent safety and cost-effectiveness. While clinical adoption is still evolving, the technology has achieved significant milestones with several commercially available systems receiving regulatory approval for breast imaging. This review synthesizes recent advances across five critical domains: system hardware design, reflection imaging, quantitative multi-parameter reconstruction (particularly through full-waveform inversion), precision calibration methodologies, and expanding clinical applications. Additionally, we have offered a comprehensive review of the application of deep learning-related technologies in USCT. By comprehensively analyzing current challenges and emerging trends, this work provides researchers and clinicians with an essential reference for understanding the state of the art and identifying pivotal pathways toward widespread clinical implementation of USCT technology.
Invasive medical interventions or abrupt reductions in ambient pressure can result in intravascular gas embolism. The accumulation of gas bubbles initiates a cascade of pathophysiological phenomena progressing from platelet activation to ischemia and neurological dysfunction. This review integrates current knowledge of the biophysical mechanisms of bubble nucleation, progression, and vascular occlusion into a framework aligned with the adverse physiological consequences on circulation. The discussion further addresses the present state of clinical practice, diagnostic approaches, and therapeutic interventions. Initial studies on gas embolism utilized in vivo models, and recent in vitro and in silico platforms have provided reproducible and cost-efficient experimental approaches. The initial symptoms of gas embolism often overlap with stroke, myocardial infarction, or sepsis. Reliable detection of intravascular gas bubbles is constrained by the sensitivity, resolution, and accessibility of existing imaging modalities, particularly in systemic cases. Current treatment frameworks emphasize hyperbaric oxygen therapy, while adjunct pharmacological strategies to improve clinical outcomes are under investigation. The challenges responsible for the persistent neglect of gas embolism in both clinical and academic contexts are discussed, and a forward-looking perspective on strategies to overcome these barriers is presented.
Ultrasound can penetrate centimetres of soft tissue, focus energy with millimetre precision, and operate safely under real-time image guidance. Leveraging these advantages, sonogenetics combines therapeutic ultrasound with genetic, cellular, and molecular engineering to create remotely programmable living systems. While widely applied in neuronal modulation, this review highlights recent progress in precision medicine, focusing on immunoengineering and genome engineering, illustrating how sonogenetics is moving beyond neuromodulation to enable precise control of immune responses and targeted genetic modifications. We introduce the fundamental principles of sonogenetics and key ultrasound-responsive biological actuators, including heat-shock promoters, thermosensitive and mechanosensitive ion channels, gas vesicles, microbubbles, and acoustically responsive nanoparticles. These convert acoustic signals into biological responses and are integrated into synthetic genetic circuits to control cell behaviour with high spatial and temporal precision. We then overview immunoengineering and genome engineering, covering cellular therapies such as CAR-T and engineered bacteria, synthetic materials, and CRISPR-based genome and transcriptome editors. This context supports recent applications, including ultrasound-guided immune modulation, remote control of CAR-T cells, tumour microenvironment reprogramming, targeted genome editing, and epigenetic regulation in vivo. We also discuss the emerging role of artificial intelligence in optimizing sonogenetic designs and outline translational challenges, including actuator safety and immunogenicity, ultrasound penetration limits, targeting accuracy, and regulatory pathways for device-biologic combinations. Key priorities include closed-loop dosimetry, scalable vector delivery, and actuator optimization. In summary, sonogenetics provides a programmable, non-invasive toolkit for controlling cellular functions, opening opportunities in basic research, diagnostics, and next-generation therapeutics.
Wearable ultrasound sensing systems are rapidly emerging for precise, continuous, and intuitive biomedical monitoring and human-in-the-loop interaction in healthcare, industry, and rehabilitation. These systems must operate under stringent constraints on size, weight, and power while delivering actionable physiological and functional information. Advances in micromachined transducers, conformable electronics, low-power signal processing, and edge artificial intelligence (AI) have enabled the first generation of wearable prototypes, yet integration of hardware and software at the system-level remains a major barrier to mass deployment. This review maps the technology readiness and architectures of wearable ultrasound systems, and examines critical design trade-offs, including edge versus cloud-based processing and pulse-echo versus coded signal approaches. We identify recurring design principles and argue that modular, scalable, and reusable platforms are key to lowering development barriers and accelerating translation from prototypes to commercial deployment across healthcare, industrial, and consumer domains.
Cardiovascular disease (CVD), the leading global cause of death, highlights the critical need for effective blood pressure management. Non-invasive blood pressure (NIBP) monitoring, compared with invasive methods, enables home-based and long-term use, supporting early detection and continuous care. Despite significant progress, challenges remain, including accuracy issues, insufficient validation in real-world settings, limited application-specific sensor designs, and inadequate calibration standards and validation platforms. These gaps call for a systematic review to clarify the unmet needs and future research directions. This article reviews current advances in four key areas: (1) novel NIBP estimation principles designed to minimize user intervention; (2) flexible and wearable electronics that improve accuracy and comfort; (3) integration with theranostic applications and broader healthcare scenarios enabled by NIBP technologies; (4) calibration and validation strategies that enhance reliability and accuracy. With the rapid growth of home healthcare and AI-enabled wearable systems, addressing these challenges is essential to advance personalized, precise and stable cardiovascular medicine.
Communication and control in biological systems is mediated by the timing of discharges -spikes- from excitable cells such as neurons and muscle fibers. Each spike is associated to a characteristic waveform that can be captured by sensors. The waveform's characteristics depend on the cell's biophysical properties and the recording modality. Depending on the technique, e.g., electrical recordings with electrodes, optical imaging, ultrasound, the observed signals are mixtures of waveforms emitted from active cells/sources (multiunit data/signals). Recovering the timing and identity of these sources (multiunit or spike decoding) is central to neuroscience, clinical diagnostics, and neural interfacing, yet it remains challenging due to waveform superposition, non-stationarity, limited training labels, and the computational demands of high-density recordings. This review provides a unified methodological perspective on spike decoding by formalizing the problem as a sparse source separation task under a convolutive mixing model. Rather than organizing the literature by application domain, we group and critically compare methods by their underlying principles: classical spike sorting, Bayesian and probabilistic inference, blind source separation, and data-driven approaches, including deep learning and hybrid schemes. For each class of methods, we present the core mathematical formulation and algorithmic strategies and discuss assumptions and limitations. Our synthesis highlights parallels in signal processing across physical recording modalities and clarifies when and why particular approaches succeed or fail. By bridging previously compartmentalized literature, this survey aims to accelerate crosspollination of ideas between application areas and to provide a roadmap for selecting, adapting, and advancing decoding methods across diverse multiunit recording modalities.
Biomedical AI is increasingly shaped by policy-bound, multi-step clinical workflows and non-stationary, multimodal data and tools. In this setting, the field is moving beyond static predictors toward agentic systems, enabled by foundation models that maintain task-relevant state and operate through a closed perceive$\rightarrow$plan$\rightarrow$act$\rightarrow$observe loop under explicit oversight. However, the field lacks a coherent account that defines biomedical agency, relates foundational model capabilities to agent behaviors, and traces the pathway from pretraining to domain-adapted, deployable systems. This survey offers such an account by synthesizing operational boundaries of agency and framing six core components (memory, planning, reflection, tool use, dialogue, and collaboration) as foundational agent-enabling capabilities that drive the transition from isolated pipelines to fully realized agents. This survey situates these perspectives along the model-building pathway, from pretraining through post-training adaptation to the orchestration mechanisms that operationalize agents. We highlight safety and governance considerations for high-stakes settings, emphasizing the fidelity of process and reasoning, uncertainty and abstention, privacy and provenance, and human oversight. Taken together, this survey provides a structured synthesis of how recent work connects foundation models to governable biomedical agentic systems and distills the recurring challenges and directions identified in the literature for reliable, accountable deployment.
Microcirculation is essential for maintaining tissue health and overall physiological function. Over the past few decades, various optical techniques have been developed to measure, visualize, and assess microvasculature. The skin has easily an accessible vascular bed allowing for noninvasive evaluation of microvascular function. Alterations in cutaneous microcirculation have been linked to dysfunctions in other target organs and vascular regions reinforcing the idea that cutaneous microcirculation can provide insights into systemic vascular conditions. Currently, there is no unified review focusing specifically on microcirculation-related optical techniques nor comprehensive analyses connecting these technological innovations to clinical evidence. This review aims to bridge that gap by systematically examining the wide spectrum of optical technologies used in assessing cutaneous microvascular function. We review techniques based on non-coherent light including oximetry, photoplethysmography, and microscopic methods and coherent light-based techniques, including speckle contrast imaging, diffuse correlation spectroscopy, photoacousting imaging, laser Doppler flowmetry and self-mixing interferometry. We emphasize cardiovascular research and evaluate the clinical relevance and technical maturity of the techniques. Additionally, brief explanation of skin structure and skin microvasculature while explaining light skin interaction is discussed. Lastly, we discuss these findings on wider context by including discussions and advancements in multimodal monitoring and machine learning.
Lower limb assistive exoskeletons (LLEs) show great potential to reduce metabolic cost, improve walking performance, and correct abnormal gait patterns. Among their core control architectures, assistive trajectory planning is key to determining system responsiveness and effectiveness under varying locomotion conditions. Many trajectory planning methods were proposed in either laboratory or unstructured real-world environments. To clarify the scope and challenges of existing research, this review categorizes trajectory planning strategies into two major types based on application scenarios: (1) strategies for laboratory settings with controllable disturbances, which usually involve optimal control for gait under stable or controllable conditions; and (2) strategies for real-world environments characterized by varying terrain, individual differences, and gait fluctuations, which usually involve adaptive control for gait under diverse or unstructured conditions. Given the foundational role of gait phase detection in trajectory planning, this review also systematically examines mainstream algorithms for gait phase recognition and estimation. Finally, the paper analyzes the limitations of existing methods and discusses the potential of advanced algorithms, intelligent multimodal sensing systems, novel sensing technologies, and embedded deployment to enhance the performance of exoskeleton assistive trajectory planning.
Pathological tremor affects over 40 million people worldwide, significantly impairing daily activities and quality of life. Pharmacological treatments show limited efficacy, with up to 30% discontinuation rates, while surgical interventions like deep brain stimulation achieve significant tremor reduction but are often unsuitable due to age, comorbidities, or personal preference. Recently, the need for safe and effective alternatives has led to the development of innovative, noninvasive, and patient-friendly technologies for tremor management. This clinical application review analyzed 134 studies (1969-2025), categorizing them into three major modalities, focusing on their underlying neurophysiological mechanisms, with a special emphasis on the clinical perspective. The three considered modalities were force-controlling (orthoses and functional electrical stimulation), central neuromodulation (transcranial magnetic stimulation, transcranial electrical stimulation, low-intensity focused ultrasound, and transcutaneous spinal cord stimulation), and peripheral neuromodulation (afferent stimulation and vibration). Force-controlling strategies showed promising acute effects, though clinical translation remains limited by poor wearability and the development of muscle fatigue. Central neuromodulation produced moderate effects, while peripheral neuromodulation has gained clinical traction, with several devices now being commercially available. However, heterogeneity in study design, patient populations, and technology maturity remain the main obstacles for the direct comparison of techniques. Future research should prioritize larger multicenter trials, standardized outcome measures, and accessibility considerations to enable personalized, evidence-based treatment selection for diverse tremor populations.
Fast Healthcare Interoperability Resources (FHIR), developed by Health Level Seven International (HL7), has emerged as the leading healthcare data standard to address persistent barriers in interoperability, fragmented exchange, and inconsistent data harmonization. As health systems worldwide undergo digital transformation, FHIR offers a flexible framework for integrating electronic health records, analytics platforms, and decision-support tools. Its growth has been accelerated by policy mandates such as the 21st Century Cures Act, as well as the availability of application programming interfaces (APIs), software development kits (SDKs), and web standards. Globally, FHIR has been adopted or piloted by national health systems in the United States, United Kingdom, Canada, and Australia, and incorporated into World Health Organization data initiatives, underscoring its role in global digital health strategy. Documented outcomes of this review include comprehensive mapping of FHIR applications across clinical, research, and public health domains; identification of adoption barriers and enablers; insights into integration with generative AI and large language models for predictive modeling, automated documentation, and decision support; and guidance for future innovations such as blockchain-enabled infrastructure and cloud-native scalability. Nonetheless, challenges remain, including uneven implementation, workforce training gaps, scalability limitations, and unresolved concerns around privacy, security, and regulatory compliance. This synthesis provides actionable insights for providers, researchers, policymakers, and developers to advance global health interoperability.
Focused ultrasound combined with intravenously infused microbubbles has been shown to effectively enhance the permeability of the blood-brain barrier, facilitating drug delivery to the brain. A wide range of technical parameters has been evaluated through preclinical studies and clinical trials. Generally, a low frequency between 200 and 300 kHz is preferred for the transcranial approach, while 1 MHz is used in implantable devices. Standard parameters include a burst length of 5 to 10 ms, a pulse repetition frequency of 0.2 to 10 Hz, and sonication durations of 90 to 180 seconds. A pressure magnitude around 0.46 mechanical index appears to be near the threshold for BBB permeability enhancement at standard microbubble dosage without causing hemorrhage. Various microbubble and nanobubble types have been tested at different doses, which in principle can be normalized by gas volume. Control methods that use harmonic emmisions for power feedback have been proposed to enhance consistency and account for patient variability, and these methods are currently being tested in several clinical trials.
Transcranial focused ultrasound (tFUS) is an emerging neuromodulation and therapeutic technology offering noninvasive, submillimeter precision for targeting deep brain structures. Unlike transcranial magnetic stimulation (TMS) and transcranial electric stimulation (tES), which are limited by depth-focality tradeoffs, or deep brain stimulation (DBS), which is invasive and costly, tFUS enables precise modulation with minimal risk. Its applications include ablation for movement and psychiatric disorders, blood-brain barrier opening (BBBO) for drug delivery in neuro-oncology and neurodegeneration, and neuromodulation for circuit-based interventions in addiction, mood/anxiety disorders, and chronic pain. Advances in phased-array transducers, holographic focusing, and real-time imaging continue to refine its accuracy and safety. Ongoing research explores closed-loop systems and wearable devices to expand clinical accessibility. This review outlines the physics, current applications, and future directions of tFUS, positioning it as a transformative tool in personalized neuromodulation and neurotherapeutics.
Backed by a century of research and development, Hill-type models of skeletal muscle, often including a muscle-tendon complex and neuromechanical interface, are widely used for countless applications. Lacking recent comprehensive reviews, the field of Hill-type modeling is, however, dense and hard-to-explore, with detrimental consequences on innovation. Here we present the first systematic review of Hill-type muscle modeling. It aims to clarify the literature by detailing its contents and critically discussing the state-of-the-art by identifying the latest advances, current gaps, and potential future directions in Hill-type modeling. For this purpose, fifty-eight criteria-abiding Hill-type models were assessed according to a completeness evaluation, which identified the modelled muscle properties, and a modeling evaluation, which considered the level of validation and reusability of the models, as well as their modeling strategy and calibration. It is concluded that most models (1) do not significantly advance beyond historical foundational standards, (2) neglect the importance of parameter identification, (3) lack robust validation, and (4) are not reusable in other studies. Besides providing a convenient tool supported by extensive supplementary materials for navigating the literature, the results of this review highlight the need for global recommendations in Hill-type modeling to optimize inter-study consistency, knowledge transfer, and model reusability.
Neurodegenerative diseases are characterized by the accumulation of misfolded proteins and widespread disruptions in brain function. Computational modeling has advanced our understanding of these processes, but efforts have traditionally focused on either neuronal dynamics or the biological processes underlying disease. One class of models uses neural mass and whole-brain frameworks to simulate changes in oscillations, connectivity, and network stability. A second class focuses on biological processes underlying disease progression, particularly prion-like propagation through the connectome, glial responses and vascular mechanisms. Each modeling tradition has provided important insights, but experimental evidence shows these processes are interconnected: neuronal activity modulates protein release and clearance, while pathological burden disrupts neuronal function. Modeling these domains in isolation limits our understanding, although recent studies have begun to bridge the two by coupling neuronal and pathological processes. To determine where and why disease emerges, how it spreads, and how it might be altered, mathematical models that capture feedback between neuronal dynamics and disease biology are needed. This review surveys the two modeling approaches and highlights efforts to unify them, emphasizing that linking neuronal activity and disease progression is key to identifying strategies that slow, halt, or reverse degeneration and restore neural function.
Neurorehabilitation conventionally relies on the interaction between a patient and a physical therapist. Robotic systems can improve and enrich the physical feedback provided to patients after neurological injury, but they under-utilize the adaptability and clinical expertise of trained therapists. In this position paper, we advocate for a novel approach that integrates the therapist's clinical expertise and nuanced decision-making with the strength, accuracy, and repeatability of robotics: Robot-mediated physical Human-Human Interaction. This framework, which enables two individuals to physically interact through robotic devices, has been studied across diverse research groups and has recently emerged as a promising link between conventional manual therapy and rehabilitation robotics, harmonizing the strengths of both approaches. Although current findings are largely based on pilot studies and conceptual frameworks, integrating therapists' expertise with the functionalities offered by robotic systems represents a promising direction for improving rehabilitation outcomes. This paper presents the rationale of a multidisciplinary team-including engineers, doctors, and physical therapists-for conducting research that utilizes: a unified taxonomy to describe robot-mediated rehabilitation, a framework of interaction based on social psychology, and a technological approach that makes robotic systems seamless facilitators of natural human-human interaction.
Endovascular procedures have revolutionized vascular disease treatment, yet their manual execution is challenged by the demands for high precision, operator fatigue, and radiation exposure. Robotic systems have emerged as transformative solutions to mitigate these inherent limitations. A crucial moment has arrived, where a confluence of pressing clinical needs and breakthroughs in AI creates an opportunity for a paradigm shift toward Embodied Intelligence (EI), enabling robots to navigate complex vascular networks and adapt to dynamic physiological conditions. Data-driven approaches, leveraging advanced computer vision, medical image analysis, and machine learning, drive this evolution by enabling real-time vessel segmentation, device tracking, and anatomical landmark detection. Reinforcement learning and imitation learning further improve navigation strategies and replicate expert techniques. This review systematically analyzes the integration of EI into endovascular robotics, identifying challenges such as the heterogeneity in validation standards and the gap between human mimicry and machine-native capabilities. Based on this analysis, a conceptual roadmap is proposed that reframes the ultimate objective away from systems that supplant clinical decision-making. This vision of augmented intelligence, where the clinician's role evolves into that of a high-level supervisor, provides a principled foundation for the future of the field.
Neurological disorders pose major global health challenges, driving advances in brain signal analysis. Scalp electroencephalography (EEG) and intracranial EEG (iEEG) are widely used for diagnosis and monitoring. However, dataset heterogeneity and task variations hinder the development of robust deep learning solutions. This review systematically examines recent advances in deep learning approaches for EEG/iEEG-based neurological diagnostics, focusing on applications across 7 neurological conditions using 46 datasets. For each condition, we review representative methods and their quantitative results, integrating performance comparisons with analyses of data usage, model design, and task-specific adaptations, while highlighting the role of pre-trained multi-task models in achieving scalable, generalizable solutions. Finally, we propose a standardized benchmark to evaluate models across diverse datasets and improve reproducibility, emphasizing how recent innovations are transforming neurological diagnostics toward intelligent, adaptable healthcare systems.
