Hypertension influences cardiovascular diseases, such as heart attacks and strokes. Blood pressure (BP) monitoring is essential for detecting hypertension and assessing its consequences. BP was traditionally measured using stethoscopes and pressure cuffs, which had several limitations. Additionally, automated blood pressure machines are not always accurate. Blood pressure measurement can be conducted more accurately and sensitively through a novel, non-invasive, and automated method. In this paper, a hybrid classification-mapping model is proposed to estimate Systolic (SBP) and Diastolic (DBP) blood pressure using 155 subjects from the University of New South Wales non-invasive BP (NIBP) dataset. In addition to exploring new beat-related features derived from oscillometric waveforms (OW), our study employs eight distinct feature ranking techniques to optimize the performance of different machine learning classifiers (K Nearest Neighbor (KNN), Ensemble KNN, Ensemble Bagged Tree, and Support Vector Machine (SVM)). As a comparison to existing methods for estimating DBP, which report a Mean Absolute Error (MAE) of 3.42 ± 5.38 mmHg, our approach achieves remarkably comparable results for estimating SBP, with an MAE of 1.28 ± 2.27 mmHg. Considering our promising results, implementing our methodology could provide a more reliable and convenient way to monitor blood pressure via remote healthcare.

The distinction between benign and malignant thoracic vertebral compression fractures (VCFs) on magnetic resonance imaging (MRI) is often subtle, and distinguishing between them is a fine-grained classification challenge. We propose a method for benign and malignant classification of thoracic VCFs based on multi-layer feature fusion and attention-guided patch reorganization to address the problem of low-level feature loss and noise associated with computing background patches when vision transformer (ViT) is applied to the task of fine-grained MRI image classification. The approach is based on the ViT architecture, which fuses low-level features with high-level features by selecting discriminative tokens using multiple layers of mutual attention weights between the classification tokens and the tokens. In addition, we incorporate an attention-guided patch recombination module that uses attention weights to select and combine patches of any two input images, which enhances the richness of the input images while reducing the noise computation. Experiments were conducted on the thoracic VCFs dataset with quantitative assessment metrics, achieving slice-level classification accuracy and AUC of 84.87% and 84.19%, respectively. Aggregating slice-level predictions, the patient-level classification accuracy reached 93.18%. Compared to other fine-grained ViT-based methods, our approach demonstrated varying improvements in slice-level accuracy, with a maximum increase of 3.65%. Ablation experiments further validated the effectiveness of the multi-layer feature fusion and patch recombination modules. The proposed MFAR-ViT utilizes the advantages of multi-layer feature fusion and patch reorganization to identify the fine-grained differences in MRI images of thoracic VCFs more efficiently, which is expected to help doctors diagnose the patient's condition quickly and accurately.

The integration of smartphones, wearable devices, and artificial intelligence (AI) has revolutionized mental health diagnostics, particularly for depression and anxiety, by enabling real-time data collection and early intervention. This review synthesizes the findings from recent studies on the use of these technologies for diagnostic precision and predictive modeling. Following the for Systematic Reviews and Preferred Reporting Items Meta-Analyses guidelines, a systematic search of PubMed, Scopus, and Web of Science was conducted for publications up to April 2025, resulting in the inclusion of 62 relevant studies. Our critical analysis revealed that, while artificial intelligence demonstrates high accuracy in detecting mental health symptoms, its performance is highly context-dependent. We examined significant challenges, including the lack of generalizability owing to disparate datasets, the critical yet often unstandardized role of feature engineering, and the "black box" nature of complex algorithms that hinder clinical trust. Addressing these limitations requires interdisciplinary collaboration, robust ethical and regulatory frameworks (e.g., GDPR and HIPAA), and scalable interpretable solutions. Future research must prioritize long-term validation, inclusivity across diverse populations, and development of explainable AI to bridge the gap between technological potential and clinical reality.

Osteoarthritis is the most common degenerative joint disease and a major cause of reduced physical function and the quality of life. Proper application of ultrasound has been proven to be effective for non-invasive osteoarthritis treatment. With a 2-D array transducer, spatial focusing of treatment pulses and three-dimensional (3-D) imaging of cartilage structures and intra-articular soft tissue are feasible for more effective treatment and diagnosis. However, supporting both imaging and therapy with a single 2-D ultrasound transducer is challenging due to the physical limitations caused by the array geometry. Given the number of active channels, increasing the element pitch can improve the lateral or elevational resolution and treatment efficacy, but introduces the grating lobe artifacts, degrading the overall image quality. To utilize a 2-D array configured with relatively large pitch elements for both 3-D imaging and low-frequency treatment, this study proposes a 3-D sub-pitch plane-wave imaging method. This method acquires channel RF data by physically translating the 2-D array transducer in the elevational and lateral directions and synthesizes all acquired RF data to reconstruct the single image, effectively maintaining the resolution while reducing grating lobe artifacts. We have demonstrated effective reduction in grating lobes through beam pattern analysis and quantitatively evaluated the imaging capabilities by Field II simulations and in-vitro experiments using a 2-D array with 8 × 8 elements centered at 2 MHz with 55% fractional bandwidth. These results could suggest that our approach may be useful in a theranostic ultrasound system supporting both treatment and diagnosis of osteoarthritic diseases.

This review article focuses on recent advancements and persistent challenges in artificial vision prostheses designed to restore sight for patients affected by retinal diseases. It comprehensively examines various approaches, including epiretinal, subretinal, and suprachoroidal implants, as well as optic nerve and visual cortex stimulation strategies. The critical role of the retina in visual perception is explored, emphasizing how retinal degeneration affects the transmission of visual information and how artificial devices aim to replicate this function. The review also discusses the technological complexities of artificial retina development, particularly challenges associated with enhancing resolution, minimizing the spread of electrical stimulation, and achieving reliable long-term device functionality within the biological environment. Practical clinical outcomes, such as surgical feasibility, device durability, and biocompatibility, are analyzed in light of these innovations. Furthermore, emerging trends are highlighted, including the adoption of flexible materials, photovoltaic structures, and 3D electrode architectures to improve the performance and longevity of implants. Ultimately, future advancements in artificial vision systems will depend on integrated approaches that combine cutting-edge engineering with a deep understanding of biological systems to achieve meaningful and lasting visual restoration.

As artificial intelligence (AI) becomes increasingly central to modern healthcare, medical education must move beyond passive knowledge transfer and adopt a system-wide approach to convergence training. This narrative review shares a 5-year case study from Seoul National University College of Medicine (SNU Medicine), which developed a comprehensive, multi-level model for integrating AI into medical education. Instead of relying on pilot programs or piecemeal curriculum updates, SNU Medicine established a governance-driven, modular framework that includes institutional infrastructure, interdisciplinary teaching strategies, cross-campus credit integration, and alignment with national digital health policies. Based on this long-term case, we propose four key design principles-modularity, transdisciplinary alignment, infrastructure-curriculum coupling, and policy embeddedness-as a framework for creating scalable and sustainable convergence education in medical AI. While rooted in Korea's unique policy environment, this model provides transferable insights for medical institutions worldwide, particularly those operating within public or policy-constrained environments.

Accurate, non-invasive prediction of muscle fatigue and coordination is essential for improving exercise performance and rehabilitation strategies. This study proposed a deep learning-based algorithm that integrates surface electromyography (EMG) and markerless motion analysis to estimate muscle fatigue and intermuscular coordination during dynamic upper-limb movement. Five healthy male participants (age: 26 ± 1.73 years) performed one-arm dumbbell curls at 50% of their one-repetition maximum (1RM), during which EMG signals were collected from the biceps brachii and lateral deltoid. Muscle fatigue was evaluated using median frequency (MDF) separately for each muscle, while intermuscular coordination was quantified via the Synergy Activation Ratio (SAR), derived from non-negative matrix factorization (NMF). Markerless motion data were captured using a Kinect V2 sensor, and both EMG and motion data were used to train an LSTM model. The model demonstrated high prediction accuracy (MDF: MSE 0.0081, MAE 0.0664 for biceps; MSE 0.0102, MAE 0.0728 for deltoid; SAR: MSE 0.0366, MAE 0.1230). Results showed a decline in biceps MDF across sets, indicating localized fatigue, while the deltoid exhibited increased MDF, possibly reflecting compensatory or inefficient activation. SAR values decreased over time, suggesting fatigue-induced reorganization of muscle synergy and increased reliance on stabilizer muscles. These findings demonstrate the feasibility of using LSTM models with synchronized EMG and motion data to detect both localized fatigue and coordination changes in real-time. The proposed framework may support future applications in personalized training, fatigue monitoring, and ergonomic assessment.

China's healthcare infrastructure faces growing population pressure and resource gaps. This review explores how AI applications, regulatory frameworks, and commercialization pathways are reshaping China's healthcare delivery system and global innovation standards. China's AI healthcare market is expected to grow from $900 million in 2020 to $1.59 billion in 2023, and is expected to reach $18.88 billion by 2030, at a Compound Annual Growth Rate (CAGR) of 42.5%. The National Medical Products Administration (NMPA) expects to approve 59 Class III AI devices by 2023, compared with just nine in 2020. Key applications include the widespread use of AI technology in lesion identification; a telemedicine platform serving 13 million users; and AI drug development that shortens the development cycle from 4 to 18 months. Regulatory pillars include the Personal Information Protection Law, which requires explicit consent before processing health data, and NMPA guidelines, which require all AI medical software to undergo three types of review. China's unique combination of centralized health data, policy incentives, and rapid commercialization has created a globally competitive AI medical ecosystem. Continued development requires addressing issues such as algorithm transparency, cross-border data governance, and international regulatory coordination.

This study proposes a visualization and analysis method for eye blinking pattern using high-frame-rate videos. The high-frame-rate video clips for visualization are taken using a camera without additional equipment. The partial video clips of eye blinking except for eyelid flutters and microsleeps are extracted from the entire video clip. The changes in shapes and positions of the upper eyelid during the eye blinking sequences are evaluated, and each eye blinking is visualized as a single image. The various parameters regarding eye blinking are calculated to analyze blinking patterns. The single eye blinking sequence is divided into phases to analyze and classify eye blinking patterns in more detail. In this experiment conducted on 80 volunteers, the proposed method was able to quantitatively analyze eyelid movements, and various parameters related to eye blinking were calculated. Additionally, different types of eye blinking patterns were visualized as graph images, and incomplete eye blinking and consecutive eye blinking were defined and detected. The proposed method can overcome the spatial and situational limitations of conventional bio-signal analysis methods, as it allows non-contact measurement in ordinary environments. In addition, since quantitative eye blink data obtained from high-frame-rate video contain more information than data obtained from bio-signals, it is expected that analysis methods using videos can be easily applied to a wider range of fields.

Microplastics have become ubiquitous in modern environments, entering the human body through multiple pathways, including air, water, and food. Recent evidence shows that microplastics penetrate deep into the human body and accumulate in tissues. Despite escalating exposure to microplastics and growing concerns about potential toxicity, strategies for microplastic clearance from the body have yet to be explored. This review summarizes current knowledge on exposure pathways, distribution, accumulation mechanisms, and health risks of microplastics and critically evaluates natural clearance mechanisms in human and their limitations. Further, we investigate potential biomedical strategies for microplastic clearance and detoxification and synthesize considerations for clinical translation.

The function of our brain is the result of the balanced interplay between billions of neurons forming a network of enormous complexity. However, the neurons themselves are also immensely complex entities, with many specialized macromolecular structures orchestrating signal processing and propagation. The postsynaptic density is an elaborate network of interconnected proteins, a dynamic yet highly organized molecular assembly beneath the dendritic membrane, and plays a pivotal role in learning, memory formation, and the development of a number of cognitive disorders. In this review, we argue that with the recent blooming of AI-assisted computational tools in structural biology, we might be able to get closer to understanding the molecular-level mechanistic aspects of this machinery. Nevertheless, we have to use these methods with caution as they are not yet capable of solving all the questions that arise for such a complex macromolecular system. First, we focus on the unique features of the postsynaptic protein network, highlighting those that pose particular challenges for such a modeling task, and put these in the light of the currently available deep learning-based approaches. We highlight the aspects that need specific attention and the areas where future developments could facilitate the detailed description of neural function at the molecular level.

Conformal tissue-electrode interfaces play a vital role in the long-term high-performance operation of bioelectronic devices, enabling continuous health monitoring, precise diagnosis, and personalized therapeutics as well as human-machine interfaces in the form of electronic skin (E-skin) and prostheses. Softness and mechanical deformability of the tissue-electrode interface minimize the damage to the target tissue and allow long-term efficient signal transmission through conformal integration with dynamically moving and curved organs. We herein summarize the recent advances in the tissue-electrode interfaces for bioelectronic devices with a focus on materials, fabrication, and applications. First, we discuss material design strategies to achieve stretchable, conductive materials. Next, we present novel fabrication techniques that fulfill the requirements of tissue-electrode interfaces. Subsequently, we present the applications of these strategies to tissue-electrode interfaces, demonstrating the advancements in the functional properties of these interfaces. Finally, we conclude with a summary and a discussion on the remaining challenges and future prospects of tissue-electrode interfaces.

Augmented reality (AR) has emerged as a powerful tool for enhancing human spatial awareness by overlaying digital information onto the physical world. This paper presents a review of the methodologies that enable AR-based spatial perception, with a focus on challenging environments such as underwater and disaster scenarios. We review state-of-the-art deep learning approaches for 3D data interpretation and completion, including voxel-based, point-based, and view-based methods. As part of this review, we implement an AR-enabled spatial awareness system, where the investigated deep learning solutions can be tested directly. In our approach, a robotic arm with an ultrasound sensor performs 2D scans underwater, from which a 3D point cloud of the scene is reconstructed. Using the reviewed deep learning networks, the point cloud is segmented in order to identify objects of interest, and point cloud completion is performed to infer missing structure. We report experimental results from synthetic data and underwater scanning trials, demonstrating that the system can recover and augment unseen spatial information for the user. We discuss the outcomes, including segmentation accuracy and completeness of reconstructions, as well as challenges such as data scarcity, noise, and real-time constraints. The paper concludes that, when combined with robust sensing and 3D deep learning techniques, AR enhances human spatial awareness in environments where direct perception is limited. The need for more adequate metrics to describe point clouds and for more labeled sonar datasets is discussed.

Purpose: Idiopathic rapid eye movement (REM) sleep behavior disorder (iRBD) is a sleep disorder considered to be a prodromal stage of neurodegeneration disease and is often accompanied by cognitive impairments. The purpose of this study was to investigate spatiotemporal characteristics of abnormal oscillatory cortical activity associated with dysfunction of visuospatial attention in iRBD based on an explainable machine learning approach. Methods: EEGs were recorded from 49 iRBD patients and 49 normal controls while they were performing Posner's cueing task and transformed to cortical current density time-series. Spectral cortical activities for four frequency bands (theta, alpha, beta, and gamma) were estimated, and then converted to three-dimensional (3D) spatiotemporal data. A pattern classifier based on 3D convolutional neural network was devised to discriminate the cortical activities of iRBD patients and those of normal controls. Results: The location, time, and frequency which characterize the difference between the patients and normal controls, thereby deemed to be associated with cognitive impairment due to the iRBD, were identified by finding the input nodes which were most critical to the classifier's decision. Conclusion: Our results suggest that theta- and gamma-band activities in parietal and occipital regions, which may underlie efficient visuospatial processing and attentional reallocation, are impaired in iRBD patients, resulting in poor visuospatial attention performance.

Heart sounds provide essential information about cardiac function; however, their clinical meaning and potential for minimally invasive hemodynamic monitoring in real world clinical settings remain underexplored. This study assessed relationships between heart sound indices and hemodynamic parameters during liver transplant surgery. Data from 80 liver transplant recipients were analyzed across five procedural phases (approximately 1,680k cardiac beats). The heart sound indices (S1 amplitude, S2 amplitude, systolic time interval, systolic time variation (STV)) were compared with hemodynamic parameters (mean blood pressure, peak arterial pressure gradient, stroke volume, systemic vascular resistance (SVR), stroke volume variation (SVV)). Relationships were assessed using Pearson's correlation, Bland-Altman analysis, and concordance correlation coefficient (CCC). The heart sound indices showed significant correlations with hemodynamic parameters during liver transplantation. S1 amplitude had positive correlations with dP/dt_max (r = 0.467-0.548), while S2 amplitude was correlated with SVR (r = 0.364-0.406). The STV showed the strongest and most consistent correlations with SVV across surgical phases (r = 0.687-0.721). Agreement metrics between STV and SVV showed mean biases ranging from - 0.34 to 0.28 with limits of agreement ranging from - 6.20 to 6.10, and the CCC ranged from 0.55 to 0.69. The amplitudes of S1 and S2 and their interval variation may reflect changes in dP/dt_max, SVR and SVV, respectively. These results suggest that heart sound parameters can serve as valuable minimally invasive indicators of hemodynamic changes during complex surgical procedures such as liver transplantation.

Foundation models, including large language models and vision-language models (VLMs), have revolutionized artificial intelligence by enabling efficient, scalable, and multimodal learning across diverse applications. By leveraging advancements in self-supervised and semi-supervised learning, these models integrate computer vision and natural language processing to address complex tasks, such as disease classification, segmentation, cross-modal retrieval, and automated report generation. Their ability to pretrain on vast, uncurated datasets minimizes reliance on annotated data while improving generalization and adaptability for a wide range of downstream tasks. In the medical domain, foundation models address critical challenges by combining the information from various medical imaging modalities with textual data from radiology reports and clinical notes. This integration has enabled the development of tools that streamline diagnostic workflows, enhance accuracy (ACC), and enable robust decision-making. This review provides a systematic examination of the recent advancements in medical VLMs from 2022 to 2024, focusing on modality-specific approaches and tailored applications in medical imaging. The key contributions include the creation of a structured taxonomy to categorize existing models, an in-depth analysis of datasets essential for training and evaluation, and a review of practical applications. This review also addresses ongoing challenges and proposes future directions for enhancing the accessibility and impact of foundation models in healthcare.

Cage subsidence is a common complication following transforaminal lumbar interbody fusion (TLIF) that can lead to poor clinical outcomes, including recurrent pain and segmental instability. Conventional TLIF cage designs often fail to distribute stress evenly, increasing the risk of endplate damage and subsequent subsidence. This study aims to evaluate the effect of a modified TLIF cage with upper and lower open windows (lattice structure) in reducing cage subsidence in patients with lumbar degenerative disc disease (LDDD). A finite element (FE) model of the lumbar spine was developed and validated. Three TLIF cage designs (Open, Lattice, Closed) were simulated under various loading conditions (flexion-extension, lateral bending, axial rotation), and von Mises stresses were analyzed within the TLIFs, endplates, and cancellous bone. The FE model demonstrated ROMs consistent with cadaveric studies. Elevated stresses were found in all cages, especially Open and Closed designs. The Lattice TLIF showed improved stress distribution, reducing peak stress on endplates. However, increased contact area had a limited effect on reducing subsidence under physiological loads. While contact area alone does not significantly mitigate subsidence risk, incorporating lattice structures may enhance resistance to physiological stress. These findings suggest that optimized TLIF designs integrating lattice structures can improve stability and reduce the likelihood of subsidence, leading to better clinical outcomes (e.g., reduced pain, improved fusion success, long-term stability) in LDDD patients.

Hypertrophic cardiomyopathy (HCM) is a common hereditary heart disease and is the leading cause of sudden cardiac death in adolescents. Septal hypertrophy (SH) and apical hypertrophy (AH) are two common types. The former is characterized by abnormal septal myocardial thickening and the latter by left ventricular apical hypertrophy, both of which significantly increase the risk of heart failure, arrhythmias, and other serious complications. Identifying hypertrophic sites in HCM patients using 12-lead electrocardiography (ECG) is crucial for early diagnosis, staging, and prognosis. However, most deep learning methods rely on 1D one-dimensional ECG signal detection, or 2D two-dimensional ECG image or spectrogram recognition, which may result in the loss of spatial or temporal information, thus limiting diagnostic accuracy. Therefore, an optimized multi-stage network with multi-dimensional spatiotemporal interactions (Ms-MdST) is proposed for detecting AH and SH in HCM. The optimized Ms-MdST model combines the advantages of different dimensional convolutions to capture the spatiotemporal characteristics of ECG and consists of a 1D convolution branch for overall temporal features and a 2D convolution branch for similar spatial features across multiple leads. Moreover, a global-local interactive attention mechanism (GLIA) and a multi-loss joint optimization strategy are employed to facilitate multi-stage multi-scale feature fusion. Experimental results show that Ms-MdST achieves F1-scores of 0.9672, 0.7250, and 0.8009 in the CONTROL, SH, and AH groups, respectively, demonstrating its superiority compared to existing ECG classification methods. In addition, the proposed model is interpretable and can be further extended to clinical applications.

This study investigates the antibacterial and anticancer activity of previously reported iron oxide (FeO)-based nanoparticles (NPs) conjugated with chlorin e6 and folic acid (FCF) in photodynamic therapy (PDT) using a human bladder cancer (BC) (T-24) cell line and three bacterial strains.

Generative models have become innovative tools across various domains, including neuroscience, where they enable the synthesis of realistic brain imaging data that captures complex anatomical and functional patterns. These models, such as Variational Autoencoders (VAEs), Generative Adversarial Networks (GANs), and diffusion models, leverage deep learning to generate high-quality brain images while maintaining biological and clinical relevance. These models address critical challenges in brain imaging, e.g., the high cost and time required for data acquisition and the frequent imbalance in datasets, particularly for rare diseases or specific population groups. By conditioning the generative process on variables such as age, sex, clinical phenotypes, or genetic factors, these models enhance dataset diversity and provide opportunities to study underrepresented scenarios, model disease progression, and perform controlled experiments that are otherwise infeasible. Additionally, synthetic data generated by these models offer a potential solution to data privacy concerns, as they provide realistic non-identifiable data. As generative models continue to develop, they hold significant potential to substantially advance neuroscience by augmenting datasets, improving diagnostic accuracy, and accelerating the development of personalized treatments. This paper provides a comprehensive overview of the advancements in generative modeling techniques and their applications in brain imaging, with a particular emphasis on conditional generative methods. By categorizing existing approaches, addressing key challenges, and highlighting future directions, this paper aims to advance the integration of conditional generative models into neuroscience research and clinical workflows.

This study aims to demonstrate that surface engineering of cytotoxic T cells with drug-loaded nanoparticles enhances nanoparticle delivery to induce a more potent combinatorial chemotherapeutic and immunotherapeutic effect, as well as enabling spatial tracking through the use of non-invasive, real-time ultrasound-guided photoacoustic imaging. Ovalbumin (OVA)-targeting OT-1 T cells were functionalized with doxorubicin-loaded, mesoporous silica-coated gold nanorods. In vitro toxicity and synergistic effects were assessed using antigen-matched OVA-expressing melanoma cells, while in vivo studies evaluated therapeutic efficacy. Ultrasound-guided photoacoustic imaging was employed to confirm the targeted delivery of the nanoengineered cells. The integration of optically active, drug-loaded nanoparticles with T cells facilitates precise image-guided delivery and enhances nanoparticle accumulation within the tumor environment, thereby maximizing the combinatorial chemo-immunotherapeutic effect. The integration of optically active, drug-loaded nanoparticles with T cells facilitates precise image-guided delivery and enhances nanoparticle accumulation within the tumor environment, thereby maximizing the combinatorial chemo-immunotherapeutic effect.