Journal of Medical Signals & Sensors最新文献

Heart disease in pregnancy is an important health issue worldwide which needs precise care to improve pregnant women health care and reduce maternal mortality rate (MMR). As we know registries play an important role in improvement of health care, so we decided to design a software to take the first step for having a national registry for pregnant women with heart disease in Iran and classify them in a more effective way to reduce mismanagements. A windows-based software with C# language programming was designed and implemented by a group of specialists included two experienced cardiologists, a skilled gynecologist, and a proficient medical doctor programmer. Since the launch of the software, information for 500 pregnant women with heart disease has been entered. The most common types of heart disease in order were congenital heart disease, prosthetic heart valves, valvular disease, and cardiomyopathies. The software developed by our team provides a comprehensive and efficient tool for managing patients with heart disease in pregnancy. The use of this software can help identify high-risk patients early on, leading to better patient outcomes and ultimately contributing to the global goal of reducing MMR. In the field of pregnant women with heart disease, gathering large and accurate data over time can be utilized in artificial intelligence for analysis.

Background: Optical coherence tomography (OCT) is a pivotal imaging technique for the early detection and management of critical retinal diseases, notably diabetic macular edema and age-related macular degeneration. These conditions are significant global health concerns, affecting millions and leading to vision loss if not diagnosed promptly. Current methods for OCT image classification encounter specific challenges, such as the inherent complexity of retinal structures and considerable variability across different OCT datasets.

Methods: This paper introduces a novel hybrid model that integrates the strengths of convolutional neural networks (CNNs) and vision transformer (ViT) to overcome these obstacles. The synergy between CNNs, which excel at extracting detailed localized features, and ViT, adept at recognizing long-range patterns, enables a more effective and comprehensive analysis of OCT images.

Results: While our model achieves an accuracy of 99.80% on the OCT2017 dataset, its standout feature is its parameter efficiency-requiring only 6.9 million parameters, significantly fewer than larger, more complex models such as Xception and OpticNet-71.

Conclusion: This efficiency underscores the model's suitability for clinical settings, where computational resources may be limited but high accuracy and rapid diagnosis are imperative.Code Availability: The code for this study is available at https://github.com/Amir1831/ViT4OCT.

The brain-computer interface (BCI) technology has emerged as a groundbreaking innovation with profound implications across diverse domains, particularly in health care. By establishing a direct communication pathway between the human brain and external devices, BCI systems offer unprecedented opportunities for diagnosis, treatment, and rehabilitation, thereby reshaping the landscape of medical practice. However, despite its immense potential, the widespread adoption of BCI technology in clinical settings faces several challenges. These include the need for robust signal acquisition and processing techniques and optimizing user training and adaptation. Overcoming these challenges is crucial to unleashing the complete potential of BCI technology in health care and realizing its promise of personalized, patient-centric care. This review work underscores the transformative potential of BCI technology in revolutionizing medical practice. This paper offers a comprehensive analysis of medical-oriented BCI applications by exploring the various uses of BCI technology and its potential to transform patient care.

Background: Deep learning has gained much attention in computer-assisted minimally invasive surgery in recent years. The application of deep-learning algorithms in colonoscopy can be divided into four main categories: surgical image analysis, surgical operations analysis, evaluation of surgical skills, and surgical automation. Analysis of surgical images by deep learning can be one of the main solutions for early detection of gastrointestinal lesions and for taking appropriate actions to treat cancer.

Method: This study investigates a simple and accurate deep-learning model for polyp detection. We address the challenge of limited labeled data through transfer learning and employ multi-task learning to achieve both polyp classification and bounding box detection tasks. Considering the appropriate weight for each task in the total cost function is crucial in achieving the best results. Due to the lack of datasets with nonpolyp images, data collection was carried out. The proposed deep neural network structure was implemented on KVASIR-SEG and CVC-CLINIC datasets as polyp images in addition to the nonpolyp images extracted from the LDPolyp videos dataset.

Results: The proposed model demonstrated high accuracy, achieving 100% in polyp/non-polyp classification and 86% in bounding box detection. It also showed fast processing times (0.01 seconds), making it suitable for real-time clinical applications.

Conclusion: The developed deep-learning model offers an efficient, accurate, and cost-effective solution for real-time polyp detection in colonoscopy. Its performance on benchmark datasets confirms its potential for clinical deployment, aiding in early cancer diagnosis and treatment.

Background: This study aimed to evaluate the effectiveness of clinical, dosimetric, and radiomic features from computed tomography (CT) scans in predicting the probability of heart failure in breast cancer patients undergoing chemoradiation treatment.

Materials and methods: We selected 54 breast cancer patients who received left-sided chemoradiation therapy and had a low risk of natural heart failure according to the Framingham score. We compared echocardiographic patterns and ejection fraction (EF) measurements before and 3 years after radiotherapy for each patient. Based on these comparisons, we evaluated the incidence of heart failure 3 years postchemoradiation therapy. For machine learning (ML) modeling, we first segmented the heart as the region of interest in CT images using a deep learning technique. We then extracted radiomic features from this region. We employed three widely used classifiers - decision tree, K-nearest neighbor, and random forest (RF) - using a combination of radiomic, dosimetric, and clinical features to predict chemoradiation-induced heart failure. The evaluation criteria included accuracy, sensitivity, specificity, and the area under the receiver operating characteristic curve (area under the curve [AUC]).

Results: In this study, 46% of the patients experienced heart failure, as indicated by EF. A total of 873 radiomic features were extracted from the segmented area. Out of 890 combined radiomic, dosimetric, and clinical features, 15 were selected. The RF model demonstrated the best performance, with an accuracy of 0.85 and an AUC of 0.98. Patient age and V5 irradiated heart volume were identified as key predictors of chemoradiation-induced heart failure.

Conclusion: Our quantitative findings indicate that employing ML methods and combining radiomic, dosimetric, and clinical features to identify breast cancer patients at risk of cardiotoxicity is feasible.

Background: Functional near-infrared spectroscopy (fNIRS) is a valuable neuroimaging tool that captures cerebral hemodynamic during various brain tasks. However, fNIRS data usually suffer physiological artifacts. As a matter of fact, these physiological artifacts are rich in valuable physiological information.

Methods: Leveraging this, our study presents a novel algorithm for extracting heart and respiratory rates (RRs) from fNIRS signals using a nonstationary, nonlinear filtering approach called cumulative curve fitting approximation. To enhance the accuracy of heart peak localization, a novel real-time method based on polynomial fitting was implemented, addressing the limitations of the 10 Hz temporal resolution in fNIRS. Simultaneous recordings of fNIRS, electrocardiogram (ECG), and respiration using a chest band strain gauge sensor were obtained from 15 subjects during a respiration task. Two-thirds of the subjects' data were used for the training procedure, employing a 5-fold cross-validation approach, while the remaining subjects were completely unseen and reserved for final testing.

Results: The results demonstrated a strong correlation (r > 0.92, Bland-Altman Ratio <6%) between heart rate variability derived from fNIRS and ECG signals. Moreover, the low mean absolute error (0.18 s) in estimating the respiration period emphasizes the feasibility of the proposed method for RR estimation from fNIRS data. In addition, paired t-tests showed no significant difference between respiration rates estimated from the fNIRS-based measurements and those from the respiration sensor for each subject (P > 0.05).

Conclusion: This study highlights fNIRS as a powerful tool for noninvasive extraction of heart and RRs alongside brain signals. The findings pave the way for developing lightweight, cost-effective wearable devices that can simultaneously monitor hemodynamic, heart, and respiratory activity, enhancing comfort and portability for health monitoring applications.

Background: Retinal imaging employs various modalities, each providing distinct perspectives on ocular structures. However, the integration of information from these modalities, which often have differing resolutions, requires effective image registration techniques. Existing retinal image registration methods typically rely on rigid or affine transformations, which may not adequately address the complexities of multimodal retinal images.

Method: This study introduces a nonrigid fuzzy image registration approach designed to align optical coherence tomography (OCT) images with fundus images. The method employs a fuzzy inference system (FIS) that uses vessel locations as key features for registration. The FIS applies specific rules to map points from the source image to the reference image, facilitating accurate alignment.

Results: The proposed method achieved a mean absolute registration error of 44.57 ± 39.38 µm in the superior-inferior orientation and 11.46 ± 10.06 µm in the nasal-temporal orientation. These results underscore the method's precision in aligning multimodal retinal images.

Conclusion: The nonrigid fuzzy image registration approach demonstrates robust and versatile performance in integrating multimodal retinal imaging data. Despite its straightforward implementation, the method effectively addresses the challenges of multimodal retinal image registration, providing a reliable tool for advanced ocular imaging analysis.

Background: Optical coherence tomography (OCT) is a biomedical imaging technique used to achieve high-resolution images from human tissues in a noninvasive manner.

Methods: In this article, a practical approach is proposed for designing ultrahigh-resolution spectral-domain OCT (UHR SD-OCT) devices. At first, block diagram of a typical SD-OCT is introduced in detail. At second, internal components of each arm are introduced where the key parameters of each component are highlighted. At third, the effects of these key parameters on the overall performance of the UHR SD-OCT are investigated in a comprehensive manner. At fourth, the most important requirements of a UHR SD-OCT are explained, where suitable optical equipment is selected for each arm based on these requirements. At fifth, optical accessories as well as the electrical devices required for managing and control of the performance of a UHR SD-OCT are introduced in brief.

Results: Performance of the proposed device is assessed through various simulations, and finally, the implementation cost and implementation challenges are investigated in detail.

Conclusions: Simulation results indicate that the proposed UHR SD-OCT has acceptable axial resolution and imaging depth; hence, it is a good candidate for use in retinal applications that require UHR imaging.

Background: Clinical decisions for stroke treatments, such as thrombolytic drugs for ischemic strokes or anticoagulants for hemorrhagic strokes, rely on accurate diagnosis and severity assessment. Our study uses diffusion-weighted magnetic resonance imaging and Convolutional Neural Networks (CNNs) to differentiate healthy and stroke samples, classify stroke types, and predict severity, aiding in decision-making for stroke management.

Methods: We evaluated 143 patients: 85 with ischemic stroke and 58 with hemorrhagic stroke. For stroke diagnosis, we compared multimodal (apparent diffusion coefficient and diffusion-weighted imaging [DWI]) and single-modal (using separate images) preprocessing techniques. Our study introduced two models, Added CNN Layer-ResNet-50 (ACL-ResNet-50) and Added CNN Layer-MobileNetV1 (ACL-MobileNetV1), based on transfer learning (MobileNetV1 and ResNet-50), enhancing performance through reinforced layers. We compared our proposed models with a scenario in which only the final layer was replaced in ResNet-50 and MobileNetV1. Furthermore, we predicted National Institutes of Health Stroke Scale (NIHSS) scores in three ranges based on DWI images to gauge stroke severity. Evaluation criteria for the models included accuracy, sensitivity, specificity, and area under the curve (AUC).

Results: In stroke classification (normal, ischemic, and hemorrhagic), ACL-MobileNetV1 outperformed other models, achieving 98% accuracy, 99% sensitivity, 98% specificity, and 99% AUC. For assessing ischemic stroke severity using NIHSS ranges, ACL-ResNet-50 showed the optimal performance with an accuracy of 0.92, sensitivity of 0.84, specificity of 0.92, and AUC of 0.95.

Conclusion: Our study's proposed method effectively classified stroke type and severity based on multimodal MR images, potentially as a practical decision support tool for stroke treatments.

Background: The global increase in diabetes prevalence necessitates advanced diagnostic methods. Machine learning has shown promise in disease diagnosis, including diabetes.

Materials and methods: We used a dataset collected from the Medical City Hospital laboratory and the Specialized Center for Endocrinology and Diabetes at Al-Kindy Teaching Hospital in Iraq. This dataset includes 1000 physical examination samples from both male and female patients. The samples are categorized into three classes: diabetic (Y), nondiabetic (N), and predicted diabetic (P). The dataset contains twelve attributes and includes outlier data. Outliers in medical studies can result from unusual disease attributes. Therefore, consulting with a specialist physician to identify and handle these outliers using statistical methods is necessary. The main contribution of this study is the proposal of two hybrid models for diabetes diagnosis in two scenarios: (1) Scenario 1 (presence of outlier data): Hybrid Model 1 combines the K-medoids clustering algorithm with a Gaussian naive Bayes (GNB) classifier based on kernel density estimation (KDE) to handle outliers and (2) Scenario 2 (after removing outlier data): Hybrid Model 2 combines the K-means clustering algorithm with a GNB classifier based on KDE with suitable bandwidth. We performed principal component analysis to minimize dimensionality and evaluated the models using fivefold cross-validation.

Results: All experiments were conducted in identical settings. Our proposed hybrid models demonstrated superior performance in two scenarios, handling and rejecting outliers, compared to other machine-learning models in this study, including support vector machines (with radial-based, polynomial, linear, and sigmoid kernel functions), decision trees (J48), and GNB classifiers for diabetes prediction. The average accuracy for Scenario 1 with Hybrid Model 1 was 0.9743, and for Scenario 2 with Hybrid Model 2, it was 0.9867. We also evaluated precision, sensitivity, and F1-score as performance metrics.

Conclusion: This study presents two hybrid models for diabetes diagnosis, demonstrating high accuracy in distinguishing between diabetic and nondiabetic patients and effectively handling outliers. The findings highlight the potential of machine-learning techniques for improving the early diagnosis and treatment of diabetes.