IEEE Transactions on Biomedical Engineering最新文献

Electroencephalography (EEG) measures the summed electrical activity from pyramidal cells in the brain by using noninvasive electrodes placed on the scalp. Traditional, voltage-based measurements are done with differential amplifiers. Depending on the location of the electrodes used for the differential measurement, EEG can estimate electrical activity from radially (common or average reference) or tangentially (bipolar derivation) oriented neurons. A limitation of the bipolar derivation is that when the electrodes are too close together, the conductive solution used to improve electrode-skin impedance can short-circuit the electrodes. Magnetoencephalography (MEG) also enables measurements from tangentially oriented cells without concerns about short-circuiting the electrodes. However, MEG is a more expensive, and a less available technology. Measuring from both radial and tangential cells can improve the resolution to localize the origin of brain activity; this could be extremely useful for diagnoses and treatment of several neurological disorders. The work presented here builds on previous research that aims to record the electrical activity of the tangentially oriented cells with technology like that of EEG. The design of the device presented here has been improved from previous implementations. Characterization of the electronics, and validation in a saline phantom and with a steady state visually evoked potentials paradigm is presented along with a comparison to a voltage-based (vEEG) amplifier. The current-based (cEEG) amplifier satisfies suggested parameters for EEG amplifiers, and exhibited higher sensitivity to tangential dipoles in the phantom study. It measured brain activity using the same scalp electrodes as vEEG amplifiers with comparable performance.

: To develop and assess an automatic and robust knee musculoskeletal finite element (MSK-FE) modeling pipeline.

Methods: Magnetic resonance images (MRIs) were used to train nnU-Net networks for auto-segmentation of knee bones (femur, tibia, patella, and fibula), cartilages (femur, tibia, and patella), menisci, and major knee ligaments. Two different MRI sequences were used to broaden applicability. Next, we created MSK-FE models of an unseen dataset using two MSK-FE modeling pipelines: template-based and auto-meshing. MSK models had personalized knee geometries with multi-degree-of-freedom elastic foundation contacts. FE models used fibril-reinforced poroviscoelastic swelling material models for cartilages and menisci.

Results: Volumes of knee bones, cartilages, and menisci did not significantly differ (p>0.05) across MRI sequences. MSK models estimated secondary knee kinematics during passive knee flexion tests consistent with in vivo and simulation-based values from the literature. Between the template-based and auto-meshing FE models, estimated cartilage mechanics often differed significantly (p<0.05), though differences were <15% (considering peaks during walking), i.e., <1.5 MPa for maximum principal stress, <1 percentage point for collagen fibril strain, and <3 percentage points for maximum shear strain.

Conclusion: The template-based modeling provided a more rapid and robust tool than the auto-meshing approach, while the estimated knee biomechanics were comparable. Nonetheless, the auto-meshing approach might provide more accurate estimates in subjects with distinct knee irregularities, e.g., cartilage lesions.

Significance: The MSK-FE modeling tool provides a rapid, easy-to-use, and robust approach for investigating task- and person-specific mechanical responses of the knee cartilage and menisci, holding significant promise, e.g., in personalized rehabilitation planning.

Objectives: Monitoring cavitation during ultrasound therapy is crucial for assessing the procedure safety and efficacy. This work aims to develop a self-sensing and low-complexity approach for robust cavitation detection in moving organs such as the heart.

Methods: An analog-to-digital converter was connected onto one channel of the therapeutic transducer from a clinical system dedicated to cardiac therapy, allowing to record signals on a computer. Acquisition of successive echoes backscattered by the cavitation cloud on the therapeutic transducer was performed at a high repetition rate. Temporal variations of the backscattered echoes were analyzed with a Singular-Value Decomposition filter to discriminate signals associated to cavitation, based on its stochastic nature. Metrics were derived to classify the filtered backscattered echoes. Classification of raw backscattered echoes was also performed with a machine learning approach. The performances were evaluated on 155 in vitro acquisitions and 110 signals acquired in vivo during transthoracic cardiac ultrasound therapy on 3 swine.

Results: Cavitation detection was achieved successfully in moving tissues with high signal to noise ratio in vitro (cSNR = 25±5) and in vivo (cSNR = 20±6) and outperformed conventional methods (cSNR = 11±6). Classification methods were validated with spectral analysis of hydrophone measurements. High accuracy was obtained using either the clutter filter-based method (accuracy of 1) or the neural network-based method (accuracy of 0.99).

Conclusion: Robust self-sensing cavitation detection was demonstrated to be possible with a clutter filter-based method and a machine learning approach.

Significance: The self-sensing cavitation detection method enables robust, reliable and low complexity cavitation activity monitoring during ultrasound therapy.

Objective: In Long-Term Monitoring (LTM), noise significantly impacts the quality of the electrocardiogram (ECG), posing challenges for accurate diagnosis and time-consuming analysis. The clinical severity of noise refers to the difficulty in interpreting the clinical content of the ECG, in contrast to the traditional approach based on quantitative severity. In a previous study, we trained Machine Learning (ML) algorithms using a data repository labeled according to the clinical severity. In this work, we explore Deep Learning (DL) models in the same database to design architectures that provide explainability of the decision making process.

Methods: We have developed two sets of Convolutional Neural Networks (CNNs): a 1-D CNN model designed from scratch, and pre-trained 2-D CNNs fine-tuned through transfer learning. Additionally, we have designed two Autoencoder (AE) architectures to provide model interpretability by exploiting the data regionalization in the latent spaces.

Results: The DL systems yield superior classification performance than the previous ML approaches, achieving an F1-score up to 0.84 in the test set considering patient separation to avoid intra-patient overfitting. The interpretable architectures have shown similar performance with the advantage of qualitative explanations.

Conclusions: The integration of DL and interpretable systems has proven to be highly effective in classifying clinical noise in LTM ECG recordings. This approach can enhance clinicians' confidence in clinical decision support systems based on learning methods, a key point for this technology transfer.

Significance: The proposed systems can help healthcare professionals to discriminate the parts of the ECG that contain valuable information to provide a diagnosis.

Perfusion index (PI), the ratio between variable pulsatile (AC) and non-pulsatile (DC) components in a photoplethysmographic (PPG) signal, is an indirect and non-invasive measure of peripheral perfusion. PI has been widely used in assessing sympathetic block success, and monitoring hemodynamics in anesthesia and intensive care. Based on the principle of dual-wavelength depolarization (DWD) of skin tissues, we propose to investigate its opportunity in quantifying the skin perfusion contactlessly. The proposed method exploits the characteristic changes in chromaticity caused by skin depolarization and chromophore absorption. The experimental results of DWD, obtained with the post occlusive reactive hyperemia test and the local cooling and heating test, were compared to the PI values obtained from the patient monitor and photoplethysmography imaging (PPGI). The comparison demonstrated the feasibility of using DWD for PI measurement. Clinical trials conducted in the anesthesia recovery room and operating theatre further showed that DWD is potentially a new metric for camera-based non-contact skin perfusion monitoring during clinical operations, such as the guidance in anesthetic surgery.

Objective: To develop a novel synthetic multi-modal variable capable of capturing cardiovascular responses to acute mental stress and the stress-mitigating effect of transcutaneous median nerve stimulation (TMNS), as an initial step toward the overarching goal of enabling closed-loop controlled mitigation of the physiological response to acute mental stress.

Methods: Using data collected from 40 experiments in 20 participants involving acute mental stress and TMNS, we examined the ability of six plausibly explainable physio-markers to capture cardiovascular responses to acute mental stress and TMNS. Then, we developed a novel synthetic multi-modal variable by fusing the six physio-markers based on numerical optimization and compared its ability to capture cardiovascular responses to acute mental stress and TMNS against the six physio-markers in isolation.

Results: The synthetic multi-modal variable showed explainable responses to acute mental stress and TMNS in more experiments (24 vs ≤19). It also exhibited superior consistency, balanced sensitivity, and robustness compared to individual physio-markers.

Conclusion: The results showed the promise of the synthetic multi-modal variable as a means to measure cardiovascular responses to acute mental stress and TMNS. However, the results also suggested the potential necessity to develop a personalized synthetic multi-modal variable.

Significance: The findings of this work may inform the realization of TMNS-enabled closed-loop control systems for the mitigation of sympathetic arousal to acute mental stress by leveraging physiological measurements that can readily be implemented in wearable form factors.

Objective: Chronic pain necessitates early intervention and accurate evaluation. Current subjective questionnaire -based methods have limitations. This study aims to develop a chronic pain assessment method based on multi-modal biosignal and to validate its feasibility.

Methods: We present a model utilizing electroencephalogram (EEG), photoplethysmogram (PPG), electrocardiogram (ECG), and facial temperature (FT) data from 59 subjects (26 chronic pain patients). A total of 112 features were derived from all signals, and 17 of them showed a significant difference between the chronic pains and the normal control.

Results: By optimizing signal types and feature combinations, our pain classification model significantly enhanced chronic pain assessment (AUROC: 0.802 to 0.864). Notable features included PPG systolic length (12.3%), EEG alpha band power (11.1%), and delta band power (9.4%).

Conclusion: This multi-modal biosignal approach holds promise for effective chronic pain quantification.