Irbm最新文献

Objectives

Intravascular optical coherence tomography (IVOCT) is a crucial micro-resolution imaging modality used to assess the internal structure of blood vessels. Lumen segmentation in IVOCT images is vital for measuring the location and the extent of vessel blockages and for guiding percutaneous coronary intervention. Obtaining such information in real-time is essential, necessitating the use of fast automated algorithms. In this paper, we proposed an innovative polynomial-regression convolutional neural network (CNN) for fast and automated IVOCT lumen segmentation.

Materials and methods

The polynomial-regression CNN architecture was uniquely crafted to enable single-pass extraction of lumen borders via IVOCT image regression, ensuring real-time processing efficiency without compromising accuracy. The architecture designed convolution for regression while omitting fully connected layers, leading to the spatial output of lumen representation as polynomial coefficients, thus enabling the formation of interconnected lumen points. The approach equipped the network to comprehend the intricate and continuous geometries and curvatures intrinsic to blood vessels in transverse and longitudinal dimensions. The network was trained on a dataset of 16,165 images and evaluated using 7,016 images.

Results

The predicted segmentations exhibited a distance error of less than 2 pixels (26.40 μm), Dice's coefficient of 0.982, Jaccard Index of 0.966, sensitivity of 0.980, specificity of 0.999, and a prediction time of 4 s (for a pullback containing 360 images). This technique demonstrated significantly improved performance in both accuracy and speed compared to published techniques.

Conclusion

The strong segmentation performance, fast speed, and robustness to image variations highlight the practical clinical utility of the proposed polynomial-regression network.

The Brain-Computer Interface (BCI) system based on motor imagery (MI) is a hot research topic nowadays, which can control external devices through the brain and has a wide range of applications in rehabilitation, gaming, and entertainment. Due to the non-smooth, non-linear, and low signal-to-noise ratio of the MI EEG signal, it is challenging to accurately decode the MI task intention. A new end-to-end deep learning method is proposed to decode raw MI EEG signals without preprocessing, such as filtering and feature reinforcement. The 1D convolution is used to learn the time-frequency features in MI signals, and a four-branch fusion network is used as the main body to add a 1D CNN-AE block and 1D SE-block to enhance the algorithm's performance. Experiments on two publicly available datasets demonstrate that our proposed algorithm outperforms the current state-of-the-art methods. It achieves 86.11% and 89.51% on the BCI Competition IV-2a and the BCI Competition IV-2b datasets, respectively, and a 6.9% improvement in the generalizability test. The proposed data enhancement method can effectively alleviate the overfitting of the algorithm and improve the decoding performance. Further analysis shows that 1D convolution can effectively extract the features associated with the MI task.

Introduction

Staying hydrated is an essential aspect of good health for people of all ages. Tracking fluid intake is important to ensure proper hydration and prompt users to drink as needed. Previous literature has attempted to measure the amount of fluid consumption, often using wearables or sensors embedded in containers.

Objective

In this paper, we introduce a novel vision-based method to estimate the amount of fluid consumed.

Methods

We trained different 3D Convolutional Neural Networks on data from 8 participants drinking from multiple containers and engaging in other activities in a simulated home environment.

Results

We show that it is possible to perform both drinking detection and volume intake estimation in a single algorithm with a Mean Absolute Percent Error (MAPE) of 28.5% and a Mean Percent Error (MPE) of 2.6% with 10-Fold and a MAPE of 42.4% and MPE of 25.4% for Leave-One-Subject-Out cross validation.

Conclusion

This shows that using video inputs does have the potential to detect and estimate the amount of fluid consumed throughout the day.

Background

Monitoring COVID-19 symptoms has become a critical task in controlling the spread of the virus and preventing hospitalizations. Aiming to contribute to efficient monitoring solutions, this article presents the development and testing of a wearable device capable of continuous monitoring biometric signals associated with the presence of COVID-19, such as the heart rate, the blood oxygen saturation, and the body temperature.

Methods

To ensure continuous monitoring the device is designed to be worn in the ear. Here, the temperature is measured through a non-contact infrared temperature sensor placed inside the ear canal while the heart rate and the pulse oximetry signals are monitored through a photoplethysmography reflective sensor positioned at the earlobe.

The proposed device's performance was evaluated by comparing it against a medical certified station. Usability and ergonomics were assessed through users' questionnaires. Additionally, experiments were performed to evaluate the hearing loss when the proposed device is in use. Data was acquired from 30 individuals of different sex, aged between 20 and 43 years old. In relation to usability and ergonomics the variation in ear dimensions was accessed and related to the device's comfort limitations.

Results

The temperature measurement produced a moderate correlation ($R=0.42$), despite a higher standard deviation was found in the proposed solution. This is due to the limited variability in temperature data, creating a short measuring range, as only healthy people were tested. The heart rate measurement also showed good correlation ($R=0.96$), with the proposed solution showing good repeatability with a standard deviation of 6.06 BPM, however, the SpO2 measurement was suboptimal ($R=0.14$).

The ergonomic evaluation revealed that most participants found the device shape comfortable, but some found the dimensions not adequate.

Additionally, the device was found to be user-friendly, with most participants reporting that they found it to be intuitive, and none reported a major loss in hearing in a normal conversation, however, there's a negligible loss of approximately 0.56 dB.

Conclusions

During this study, it was possible to develop and evaluate a wearable device that was suggested for monitoring biometric signals. The device demonstrated great reliability in temperature and heart rate measurement but showed limitations in the accuracy of pulse oximetry. The main contribution of this work is the evaluation of a continuous non-invasive monitoring concept for COVID-19 related biometric signals, which indicates good applicability in the case study.

Objective

Artificial Intelligence (AI) in electrocardiogram (ECG) analysis helps to identify persons at risk of developing atrial fibrillation (AF) and reduces the risk for severe complications. Our aim is to investigate the performance of AI-based methods predicting future AF from sinus rhythm (SR) ECGs, according to different characteristics of patients, time intervals for prediction, and longitudinal measures.

Methods

We designed a retrospective, prognostic study to predict AF occurrence in patients from 12-lead SR ECGs. We classified patients in two groups, according to their ECGs: 3,761 developed AF and 22,896 presented only SR ECGs. We assessed the impact of age on the overall performance of deep neural network (DNN)-based systems, which consist in a variation of Residual Networks for time series. Then, we analysed how much in advance our system can predict AF from SR ECGs and the performance for different categories of patients with AUC and other metrics.

Results

After balancing the age distribution between the two groups of patients, our model achieves AUC of 0.79 (0.72-0.86) without additional constraints, 0.83 (0.76-0.89) for ECGs recorded in the last six months before AF, and 0.87 (0.81-0.93) for patients with stable AF risk measures over time, with sensitivity of 90.62% (80.70-96.48) and diagnostic odd ratio of 20.49 (8.56-49.09).

Conclusion

This study shows the ability of DNNs to predict new onsets of AF from SR ECGs, with the best performance achieved for patients with stable AF risk score over time. The introduction of this time-based score opens new possibilities for AF prediction, thanks to the analysis of long-span time intervals and score stability.

Background

Stair walking requires to elevate or lower the body center of mass and results in increased muscle contractions and consumed energy compared to level walking. Mechanical work produced by the body can be quantified through Individual Limb Method and the summed lower limb joint work but there does not exist normative data of these works in stair ascent and descent compared to slope ascent and descent of the same individuals.

Methods

Upstair and downstair walking were investigated at 0%, 5% and 12% inclinations and compared to upslope and downslope walking for thirteen able-bodied volunteers. Lower limb joint and individual limb powers and works were compared across walking conditions.

Findings

Work production and absorption required to elevate or lower the center of mass directly depend on the inclination to be crossed (about 0.35 J/kg for 5% slope, 0.9 J/kg for 12% slope and 1.6 J/kg for stair). However, the distribution among joints and between gait phases is different when considering stair versus slope walking. In particular, the role of the knee is exacerbated for work production in stair ascent (45% of total work) as well as for work absorption in stair descent (61% of total work). Also, more work production/absorption is performed during the swing phase for stair walking then for slope walking.

Interpretation

This study provides reference data of the Individual Limb mechanical work performed during stair walking and show that this method can substitute to summed lower limb joint one during the stance phase of stair walking.

Background

Alzheimer's disease can be diagnosed through various clinical methods. Among them, electroencephalography has proven to be a powerful, non-invasive, affordable, and painless tool for its diagnosis.

Objectives

In this study, eight machine learning (ML) approaches, including SVM, BLDA, DT, GNB, KNN, RF, and deep learning (DL) methods such as RNN and RBF, were employed to classify Alzheimer's disease into two stages: moderate Alzheimer's disease (ADM) and advanced Alzheimer's disease (ADA).

Material and methods

To this aim, electroencephalography data collected from five different hospitals over a decade has been used. A novel method based on neural networks has been proposed to increase accuracy and obtain fast classification times.

Results

Results show that deep neuronal networks based on radial basis functions initialized with fuzzy means achieved the best balanced accuracy with 96.66% accuracy in ADA classification and 93.31% accuracy in ADM classification.

Conclusion

Apart from improving accuracy, it is noteworthy that this algorithm had never been used before to classify patients with Alzheimer's disease.

Objectives

Magnesium and magnesium alloy materials have excellent potential as biodegradable bone plate implants. However, the practical application of magnesium alloys is limited by their high chemical activity and poor corrosion resistance. Here, we chose a microarc fluorination (MAF) treatment to improve corrosion resistance while enhancing aspects of magnesium alloy properties. The aim of this study was to identify the effect of fixed-point corrosion on the corrosion resistance as well as the mechanical properties of magnesium alloys and to design a new corrosion-oriented model that can provide absolute protection over a period of time.

Material and Methods

MAF treatment is used for surface modification of magnesium alloys to improve the corrosion resistance of magnesium alloys. To investigate the effect of the coating and indentation on the corrosion resistance of Mg alloy, electrochemical corrosion experiments were carried out. It is worth mentioning that in this experiment we measured and analyzed the mechanical properties of the samples, especially the tensile strength.

Results

In the innovative indentation sample test, the coated specimens showed lower tensile strength due to the occurrence of fixed-point corrosion. To avoid the loss of mechanical properties due to fixed-point corrosion, we proposed a new idea (Corrosion-oriented Design). Ultimately, the immersion experiments as well as the mechanical properties analysis concluded that the Corrosion-oriented Design samples could maintain the mechanical properties without detectable loss for a long time.

Conclusion

The Corrosion-oriented Design model can avoid the nuisance of fixed-point corrosion and control the centralized orientation of corrosion. This provides a new direction for the clinical application of magnesium alloys, which may offer a completely stable bone-healing condition in trauma treatment and avoid the drawbacks caused by the previous uncontrolled corrosion.

Background and Objective

Over 68,000 opioid-overdose related deaths occurred within the United States in 2020 alone, indicating a need to develop technologies to help curb this growing epidemic. The ability to detect respiratory rate (RR) depression in real-time has the potential to decrease adverse outcomes by alerting emergency medical services or willing bystanders to an overdose event. The aim of this investigation was to design, build, and test a novel photoplethysmography (PPG)-based measurement device capable of monitoring RR and identifying respiratory depression.

Materials and Methods

We developed a novel murine model for opioid-induced respiratory depression (OIRD) to demonstrate the PPG device's capabilities. We induced respiratory depression in mice using both isoflurane and opioid-overdose and initiated recovery events with injections of naloxone while monitoring respiration via PPG and a laboratory reference system.

Results and Discussion

The device accurately identified all anesthesia-induced respiratory depression (n = 5) and OIRD events (n = 3). Our PPG-based monitor showed significant correlation with a reference respiratory measurement system ($p<0.01$). The bias measured across the isoflurane trials was 0.6 breaths per minute (BrPM), while the bias measured across the oxycodone trials was −1.0 BrPM, with mean absolute errors of 1.5 and 3.6 BrPM, respectively, indicating that our device was able to accurately measure RR in a murine model.

Conclusions

These preliminary experiments suggest that our device could detect OIRD and could potentially be adaptable to humans with modifications to firmware and more extensive validation in human subjects. Our present study is a proof-of-concept for detecting OIRD and alerting bystanders and health professionals in real-time.