Target-free recognition of cable vibration in complex backgrounds based on computer vision

Mechanical Systems and Signal Processing, Год журнала: 2023, Номер 197, С. 110392 - 110392

Опубликована: Апрель 29, 2023

Язык: Английский

---

A Robust Framework Combining Image Processing and Deep Learning Hybrid Model to Classify Cardiovascular Diseases Using a Limited Number of Paper-Based Complex ECG Images

Biomedicines, Год журнала: 2022, Номер 10(11), С. 2835 - 2835

Опубликована: Ноя. 7, 2022

Heart disease can be life-threatening if not detected and treated at an early stage. The electrocardiogram (ECG) plays a vital role in classifying cardiovascular diseases, often physicians medical researchers examine paper-based ECG images for cardiac diagnosis. An automated heart prediction system might help to classify diseases accurately This study aims into five classes with using deep learning approach the highest possible accuracy lowest time complexity. research consists of two approaches. In first approach, models, InceptionV3, ResNet50, MobileNetV2, VGG19, DenseNet201, are employed. second integrated model (InRes-106) is introduced, combining InceptionV3 ResNet50. developed as convolutional neural network capable extracting hidden high-level features from images. ablation conducted on proposed altering several components hyperparameters, improving performance even further. Before training model, image pre-processing techniques employed remove artifacts enhance quality. Our hybrid InRes-106 performed best testing 98.34%. acquired 90.56%, ResNet50 89.63%, DenseNet201 88.94%, VGG19 87.87%, MobileNetV2 achieved 80.56% accuracy. trained k-fold cross-validation technique different k values evaluate robustness Although dataset contains limited number complex images, our based various techniques, fine-tuning, studies, effectively diagnose diseases.

Язык: Английский

---

Classification of Electrocardiogram Signals Based on Hybrid Deep Learning Models

Sustainability, Год журнала: 2022, Номер 14(24), С. 16572 - 16572

Опубликована: Дек. 10, 2022

According to the analysis of World Health Organization (WHO), diagnosis and treatment heart diseases is most difficult task. Several algorithms for classification arrhythmic heartbeats from electrocardiogram (ECG) signals have been developed over past few decades, using computer-aided systems. Deep learning architecture adaption a recent effective advancement deep techniques in field artificial intelligence. In this study, we new convolutional neural network (CNN) bidirectional long-term short-term memory (BLSTM) model automatically classify ECG into five different groups based on ANSI-AAMI standard. End-to-end (feature extraction work together) done hybrid without extracting manual features. The experiment performed publicly accessible PhysioNet MIT-BIH arrhythmia database, findings are compared with results other two models, which combination CNN LSTM Gated Recurrent Unit (GRU). performance also existing works cited literature. Using SMOTE approach, database was artificially oversampled address class imbalance problem. This trained validated tenfold cross-validation actual test dataset. experimental observations, outperforms terms recall, precision, accuracy F-score 94.36%, 89.4%, 98.36% 91.67%, respectively, better than methods.

Язык: Английский

---

Deep learning-based real-time diagnosis of cardiac diseases through behavioral changes in ECG signals

Biomedical Signal Processing and Control, Год журнала: 2025, Номер 104, С. 107532 - 107532

Опубликована: Янв. 28, 2025

Язык: Английский

---

Advanced Hybridization and Optimization of DNNs for Medical Imaging: A Survey on Disease Detection Techniques

Artificial Intelligence Review, Год журнала: 2025, Номер 58(4)

Опубликована: Фев. 4, 2025

Язык: Английский

---

A framework for detecting high-performance cardiac arrhythmias using deep inference engine on FPGA and higher-order spectral distribution

Mechanical Systems and Signal Processing, Год журнала: 2025, Номер 228, С. 112445 - 112445

Опубликована: Фев. 17, 2025

Язык: Английский

---

Design of Convolutional Neural Network Optimization Algorithm Based on Embedded System and Its Application in Real-Time Image Processing

Applied Mathematics and Nonlinear Sciences, Год журнала: 2025, Номер 10(1)

Опубликована: Янв. 1, 2025

Abstract With the rapid development of artificial intelligence technology, optimizing convolutional operation neural network (hereinafter referred to as CNN) adapt resource constraints embedded systems has become one current research hotspots. In this paper, we explain basic connotation CNN and platform Zynq, optimize Im2col-Gemm algorithm based on Darknet framework, so further model. The before after optimization under different hardware configurations are compared through acceleration tests, average time spent each layer total operations recorded, which clearly concludes that Zynq combining optimized can achieve 658.12 23.18 times with respect CPU GPU, respectively. Through character recognition detection traffic sign detection, Zynq’s achieves 220FPS less than 4.5W power consumption, it only takes about 4.5ms recognize a picture. Meanwhile, high rate 97.8% low leakage 8.28%, verifies is fast consumes power, advantageous for applications in real-time image processing. Optimizing helps promote continuous upgrading intelligence.

Язык: Английский

---

Identification of middle cerebral artery stenosis in transcranial Doppler using a modified VGG-16

Frontiers in Neurology, Год журнала: 2024, Номер 15

Опубликована: Окт. 16, 2024

Objectives The diagnosis of intracranial atherosclerotic stenosis (ICAS) is great significance for the prevention stroke. Deep learning (DL)-based artificial intelligence techniques may aid in diagnosis. study aimed to identify ICAS middle cerebral artery (MCA) based on a modified DL model. Methods This retrospective included two datasets. Dataset1 consisted 3,068 transcranial Doppler (TCD) images MCA from 1,729 patients, which were assessed as normal or by three physicians with varying levels experience, conjunction other medical imaging data. data used improve and train VGG16 models. Dataset2 TCD 90 people who underwent physical examination, verify robustness model compare consistency between human physicians. Results accuracy, precision, specificity, sensitivity, area under curve (AUC) best + Squeeze-and-Excitation (SE) skip connection (SC) dataset1 reached 85.67 ± 0.43(%),87.23 1.17(%),87.73 1.47(%),83.60 1.60(%), 0.857 0.004, while those dataset2 93.70 2.80(%),62.65 11.27(%),93.00 3.11(%),100.00 0.00(%), 0.965 0.016. kappa coefficient showed that it recognition level senior doctors. Conclusion improved has good diagnostic effect MCV expected help screening.

Язык: Английский

---

Classifying the heart sound signals using textural‐based features for an efficient decision support system

Expert Systems, Год журнала: 2023, Номер 40(6)

Опубликована: Фев. 7, 2023

Abstract Cardiovascular diseases have surpassed cancer as the leading cause of death on planet today. Numerous decision‐making systems with computer‐assisted support been developed to assist cardiologists detect heart disease, and thus, lowering mortality rate. The purpose this research is classify audio signals received from normal or abnormal. PhysioNet Computing in Cardiology (CinC) 2016 benchmark dataset, popularly known 2016, has used validate proposed methodology presented here. contains a total 3200 phonocardiogram (PCG) recordings divided into sub‐datasets A‐F. state‐of‐the‐art studies conducted till date not considered harmonic details beat that can be extracted its equivalent chromagram image. In work, textural features such linear binary pattern (LBP), adaptive‐LBP, ring‐LBP existing spectrogram combined chromagram. It observed combination both image variants resulted greater accuracy compared scenario where researchers were using only spectrogram. experiment yielded mean accuracy, precision, F1‐score 94.87, 93.11, 95.273, respectively. sound classification models employ spectrogram, scalogram, mel‐spectrogram images view analyse acoustic properties PCG signal. Although these visual tools provide useful information about signal, yet they are unable distinguish between pitch resonance generation. However, paper proposes an alternative approach signal representation allows for more precise measure pitch‐related changes sound. Its results highlight significance extracting time‐chroma (i.e., chromagram) explored domain related

Язык: Английский

---

An IoT-Based Telemedicine System for the Rural People of Bangladesh

Опубликована: Янв. 1, 2024

IoT devices can enable low cost and interactive health care services. In this paper we have proposed an affordable telemedicine system to bring healthcare services within the reach of rural people Bangladesh. Proposed enables transmission patient's body parameters in real-time a remote doctor. The also has patient monitoring capability which is based on ECG signal classification. A feed-forward neural network used for classification embedded ARM processor. For power operation, utilized fixed-point (integer) arithmetic instead floating-point task. implementation 1.06x faster than requires 50% less memory store model without loss accuracy.

Язык: Английский

---