Annals of Biomedical Engineering最新文献

Despite technological advancements in echocardiography (echo) systems, effectively utilizing these machines and achieving accurate and timely interpretation of the resulting image data still pose significant challenges. Hence, many researchers have sought to overcome these challenges by leveraging artificial intelligence (AI), particularly through the application of deep learning (DL) techniques. In this study, we provide a thorough analysis of studies aimed at leveraging DL to reshape the field of echocardiography, with a focus on their clinical impact, challenges, and opportunities for advancement. These studies can be categorized into two main groups, each encompassing multiple tasks: data acquisition and quality enhancement, and intelligent echo data analysis. The latter group is the most extensively studied, with key tasks including annotated data generation, cardiac abnormality diagnosis, cardiac structure segmentation, etc. Through thorough analysis of the selected studies, we highlight the transformative impact of DL techniques on the automatic diagnosis and monitoring of cardiovascular diseases, as well as the underexplored challenges and potential solutions and research tasks. Moreover, we introduce heart anatomy and important clinical parameters, thereby providing researchers of DL algorithms with clinical procedures associated with evaluating blood vessels and cardiac function. Furthermore, details of the commonly adopted deep learning models and their implementation procedures are presented. By providing a multidisciplinary perspective, we believe that our work paves the way for more collaborative efforts in the field and establishes a foundation for future innovations in DL-driven echocardiography.

Purpose: Limb compression is proposed as a supplement to cardiopulmonary resuscitation (CPR) because it has the potential to increase central organ blood flow rates. This is significant because CPR only has a survival rate of 10-15%, and survival depends on the amount of blood flow generated during CPR. However, the underlying physiological mechanisms are not understood, thereby limiting optimal application of limb compression during CPR. Therefore, this study addressed the hypothesis that circuit pathway reduction, volume displacement, and wave reflection mechanisms contribute to higher central organ blood flow rates. Mathematical models were utilized because modeling is the only way to isolate physiological mechanisms.

Methods: This study utilized two validated mathematical models that were minimally altered to investigate limb compression during CPR. A lumped-parameter model addressed circuit pathway reduction and volume displacement while the other model isolated wave reflections.

Results: Relative to standard CPR, the model simulating circuit pathway reduction via CPR with tourniquets predicted increases of cerebral and coronary flow rates by 3% each while the model simulating volume displacement via CPR with constant pressure cuffs predicted increased cerebral and coronary flow rates by up to 74% and 109%, respectively. Furthermore, the wave reflection model predicted increases in blood flow rates of major cerebral arteries up to 11% and across all major abdominal arteries up to 27%.

Conclusion: All limb compression configurations and modalities resulted in increased predicted central organ blood flow rates with volume displacement, wave reflection, and circuit pathway reduction mechanisms, in that order, being most influential.

During unexpected drop-like gait perturbations, the body's center of mass (CoM) energy must be absorbed reactively by the leg muscles, challenging muscle-tendon unit (MTU) function and body stability. Anticipation and prior experience may adjust muscle activation in advance to improve the perturbation response. The study's purpose was to investigate the interplay of muscle activation, MTU decoupling mechanisms, and contractile conditions for the CoM energy management during gait challenges.Kinematics, electromyographic activity (EMG), soleus fascicle length, and total CoM energy were measured during unperturbed walking, unpredictable and adapted (experience-based) drop-like perturbations as well as during hole negotiation. The soleus force-length and force-velocity relationships were also determined to assess the force-length-velocity potential.CoM energy decreased substantially after touchdown in the hole during both perturbations and hole negotiation, indicating energy absorption by the musculoskeletal system. During the unpredictable perturbation, a rapidly increased EMG activity after drop initiation and an almost isometric fascicle behavior close to optimal length throughout the CoM energy absorption phase was found, despite MTU lengthening. In the adapted perturbation, an initial isometric contraction accompanied by high EMG activity was observed, followed by active fascicle lengthening at decreasing EMG activity. Clear fascicle lengthening in combination with low EMG activity was found during hole negotiation.These novel findings suggest a regulation of muscle stiffness by scaled activation that tunes the contribution of muscle and tendon to the MTU length changes (i.e., tendon decoupling), to facilitate high fascicle force-length-velocity potentials and tendon energy buffering mechanisms in response to drop-like perturbations and hole negotiation gait.

Purpose: Examine sex-specific differences in Intracranial Aneurysms (IA) rupture at presentation and to retrospectively benchmark sex-stratified versus pooled classification models for their ability to discriminate rupture status, using a cross-sectional cohort.

Methods: We retrospectively analyzed 203 patients (46 males, 157 females) with 303 IAs from a single-center, IRB-approved registry. Of these, 76 IAs were ruptured and 227 unruptured at presentation. Clinical data and lesion characteristics were summarized into patient-level variables and used to develop sex-specific logistic regression models. Adjusted Odds ratios (ORs) were produced for clinical covariates. Cross-application assessed generalizability across sexes.

Results: Among females, 58.7% of ruptures were < 5 mm, with a median ruptured size of 4.2 mm at Anterior Communicating Artery (ACOM). Rupture likelihood in females peaked between ages 40 and 59 (aORs = 2.8 and 1.7), coinciding with perimenopause. In males, ACOM was the most frequent rupture site; although males had higher mean hemoglobin levels (14.1 vs. 12.5 g/dL, p < 0.0001), hemoglobin contributed less to rupture compared to sex-specific models incorporating age, location, and metabolic factors (hemoglobin concentration, blood glucose). Metabolic factors contributed significantly to the female-specific model, achieving strong discrimination (AUC-ROC: 0.80), while the male-specific model underperformed (AUC-ROC: 0.50), due to limited rupture events (n = 19). Cross-application of features between sexes drastically reduced performance, providing the first computational evidence that male and female rupture mechanisms represent distinct biological feature spaces requiring separate modeling architectures CONCLUSION: Women in this cohort more often presented with rupture IAs at smaller sizes and at midlife ages than men. These sex-specific patterns, though strictly cross-sectional and associative rather than predictive, highlight potential biological and clinical contributors to rupture presentation and may partly explain misclassification by pooled risk models. Future longitudinal, multicenter studies with balanced cohorts are required to validate these findings and to develop robust rupture-risk prediction tools that incorporate sex as a biological variable.

Purpose: 2D-3D registration in dynamic stereo-radiography is performed in biomechanics and orthopedics methods to acquire 6-degree of freedom poses of native anatomy. Conventional registration frameworks rely on subject-specific digitized anatomy segmented from computed tomography scans, which increases processing efforts associated with registration, and also presents additional radiation risks. Scan-free registration has thus been introduced to circumvent reliance on volumetric imaging by jointly estimating shape and pose in radiographs. Existing approaches utilize principal component analysis for efficient sampling of plausible anatomic candidates, but this technique involves tedious and error-prone mesh or scan registration as a developmental step. The current work proposes an alternative method for scan-free 2D-3D registration, referred to as Neural Implicit Shape and Intensity Models.

Methods: Neural Implicit Shape and Intensity Models were developed to capture population-level anatomic shape and intensity variability observed across a training set, without reliance on mesh registration. Trained models were integrated with a 2D-3D registration framework for pose estimation and anatomy reconstruction in stereo-radiographs.

Results: The framework was evaluated by performing 2D-3D registration of the native knee in stereo-radiographs capturing in vivo movement, and in additional frames capturing a synthetic leg phantom. Experiments revealed geometric errors consistently near or below 1 mm, and pose errors near or below 1 $_{}^{\circ }$ or mm in both evaluation cohorts.

Conclusions: This work proposed novel methods for scan-free 2D-3D registration in stereo-radiography. Results indicate the introduced framework may benefit registration applications by addressing dependence on volumetric medical imaging.

Purpose: Smart rings are emerging as a promising solution in the field of wearable devices, offering a compact and ergonomic solution for continuous physiological monitoring, yet one of their major limitations is that they are not adjustable, but rather come in different sizes. Another major gap in the literature is the lack of validation data during dynamic activities. This study presents the design, development, and validation under static and controlled-motion conditions of a research-grade adjustable smart ring for tracking pulse rate (PR) and peripheral blood oxygen saturation (SpO₂) using photoplethysmography.

Methods: The device features a rigid-flex printed circuit board (PCB) optimized for finger placement, ensuring accurate sensor positioning, and adaptability to different sizes. Data transmission occurs via the ANT protocol, allowing real-time visualization on a mobile application. The smart ring was evaluated on 30 healthy volunteers (13 women and 17 men, mean age 27.5 ± 7.4 years, mean height 172.2 ± 9.0 cm, mean weight 68.1 ± 13.1 kg) through a multi-phase experimental protocol involving spontaneous breathing, apnea, and physical activity, with measurements compared against a gold-standard pulse oximeter.

Results: Results demonstrate strong agreement in PR measurements both in static and dynamic conditions (r = 0.91, p < 0.001), while SpO₂ values exhibited an overestimation (mean bias = 1.04%).

Conclusion: This work demonstrates the feasibility of an adjustable, research-grade smart ring based on a rigid-flex PCB for finger palmar PR and SpO₂ monitoring and quantifies its agreement with a clinical-grade fingertip oximeter in healthy adults in static and dynamic conditions. The device is conceived primarily as a hardware and measurement site platform that can be reused across ring sizes, while future work is needed to develop open signal-processing algorithms.

Purpose: This study investigates the potential of large eddy simulation (LES) as a benchmark for validating Reynolds-averaged Navier-Stokes (RANS) models in the CentriMag centrifugal blood pump. We compare velocity field predictions from LES and RANS against particle image velocimetry (PIV) and quantify previously unreported non-rotatory components in the motion of the magnetically levitated impeller, assessing their impact on flow dynamics.

Methods: We performed PIV on an optically accessible pump replica and compared phase-averaged velocity fields with computational fluid dynamics (CFD) predictions from LES and three unsteady RANS models. In addition, using custom optical trackers embedded in the PIV setup, we quantified the three-dimensional impeller motion and replicated its main non-rotatory components in the simulations.

Results: The impeller exhibited complex but small deviations from ideal rotation. When incorporated into the CFD model, these had negligible impact on the flow field. LES predictions agreed well with PIV data, with root-mean-square velocity errors of approximately 3%, while the RANS models showed larger local deviations, particularly in the outlet region.

Conclusion: LES demonstrated close agreement with PIV and proved to be a reliable reference for validating RANS models in this study. Although the analysis was limited to a single pump, LES warrants further consideration as a potential benchmark method that could reduce reliance on extensive experimental flow validation. Consistent with previous findings, the transition from the volute to the outlet emerged as a particularly sensitive region for RANS modeling and validation. The quantified non-rotatory impeller motion had negligible impact on the flow field, supporting the continued use of idealized rotor motion in CFD modeling of similar magnetically levitated devices.

Purpose: Despite falls accounting for the greatest number of fatal and non-fatal work-related traumatic brain injuries, current standards do not evaluate safety helmets under impact conditions representative of fall scenarios. This study's objective was to develop a test method that evaluates safety helmets under impact scenarios representative of falls. A Construction STAR rating system that quantitatively compares safety helmet performance in the context of concussion and skull fracture risk is also outlined.

Methods: A multi-step approach that combined information from previous literature, Occupational Safety and Health Administration (OSHA) accident reports, and oblique impact tests were used to develop a fall-specific safety helmet test methodology. The test methodology consisting of three impact locations (front boss, rear boss, and rear), two impact velocities (5.5 and 6.8 m/s), and a 25-degree anvil was executed on a representative subset of one Type I and four Type II models. STAR scores, combining concussion and skull fracture risk, were calculated for each model and compared.

Results: STAR scores demonstrated that Type II helmets reduced concussion risk by 32.7% and skull fracture risk by 57.5% when compared to the Type I model. Large variations in Type II performance were observed, with the top-performing Type II helmets reducing concussion risk by 28.7 and 33.2% compared to bottom-performing models.

Conclusions: Type II helmets offer substantial benefits in head protection compared to Type I models for oblique fall-related impacts. By including both skull fracture and concussion risk in the STAR score, the proposed methodology can differentiate high and low-performing safety helmets.

Purpose: Physiologic closed-loop controlled (PCLC) medical devices have the potential to enhance therapeutic precision and effectiveness. However, their complexity introduces potential failure modes that may pose risks to patients if not properly evaluated. This study assesses the effectiveness of a mathematical model of the cardiovascular system in predicting PCLC performance metrics through in vivo and in silico comparisons of blood pressure responses to automated fluid infusion.

Methods: A closed-loop control system for regulating mean arterial pressure (MAP) was implemented using a custom Java-based software application on a tablet. The system operated in two control speed modes, 'slow' and 'fast.' The study was conducted in an animal laboratory using thirteen swine subjects (two were excluded due to hardware-related issues). Following intubation, anesthesia, and splenectomy, hemorrhage was induced until MAP reached 45 mmHg, followed by fluid resuscitation using the closed-loop controller targeting 70 mmHg. Arterial blood pressure waveforms were continuously recorded, and cardiac output and hematocrit measurements were taken every 15 min. The same control algorithm and speed modes were applied to in silico subjects generated with a mathematical model simulating cardiovascular responses to fluid perturbation while the same protocol was simulated.

Results: The mathematical model effectively predicted key PCLC performance metrics, including rise time, %overshoot, settling time, and divergence. The simulated results closely matched experimental data and captured differences between slow and fast control speeds.

Conclusion: The ability of the mathematical model to predict PCLC performance metrics demonstrates its value in supporting the development and evaluation of automated fluid resuscitation controllers.

High myopia (HM) is a vision-threatening ocular disorder characterized by excessive axial elongation. While previous studies have primarily focused on structural and functional impairments in the posterior segment-including the optic nerve, posterior sclera, and macular region-recent evidence indicates that adaptive remodeling also occurs in anterior segment structures, such as the cornea, anterior sclera, and lens. In this review, we examine the molecular mechanisms underlying anterior segment remodeling in high myopia, investigate the associated biomechanical alterations using ex vivo and in vivo measurement techniques, and analyze the interrelationships among these changes. Furthermore, we highlight how multimodal imaging technologies, when integrated with artificial intelligence algorithms, enable the quantitative assessment of biomechanical parameters. These advances may contribute to improved prediction of myopia progression, risk stratification for associated complications, and the development of personalized therapeutic strategies. Finally, we discuss the current challenges and translational bottlenecks in this area.