Resilient embedded system for classification respiratory diseases in a real time

Biomedical Signal Processing and Control, Journal Year: 2023, Volume and Issue: 90, P. 105876 - 105876

Published: Dec. 30, 2023

Language: Английский

---

Correlating spirometry findings with auscultation sounds for diagnosis of respiratory diseases

Biomedical Signal Processing and Control, Journal Year: 2023, Volume and Issue: 87, P. 105347 - 105347

Published: Aug. 30, 2023

Language: Английский

---

Wheeze Events Detection Using Convolutional Recurrent Neural Network

2022 Innovations in Intelligent Systems and Applications Conference (ASYU), Journal Year: 2023, Volume and Issue: unknown, P. 1 - 6

Published: Oct. 11, 2023

Chronic respiratory disorders (CRDs) affect the airways and other structures in lungs. According to WHO, CRDs are a major cause of death globally. Early diagnosis monitoring individuals with crucial due severity prevalence these disorders. Auscultation is common method used diagnose patients. However, classical auscultation procedure has some limitations, such as being subjective, depending on physician's expertise, inaccurate noisy environments. To tackle those this project aims implement for detection adventitious sounds, particularly wheeze using data derived from ICBHI open data. Short-time Fourier transforms (STFT) audio were applied feature extraction. The system was implemented perform sound recurrent neural network (RNN) based deep-learning model.

Language: Английский

---

AHP-CM: Attentional Homogeneous-Padded Composite Model for Respiratory Anomalies Prediction

2022 IEEE 10th International Conference on Healthcare Informatics (ICHI), Journal Year: 2023, Volume and Issue: unknown, P. 65 - 71

Published: June 26, 2023

Early diagnosis, treatment and regular monitoring can limit the spread adverse effects of respiratory diseases. Shortage trained physicians is one main obstacles to ensure early diagnosis which be overcome by making lung auscultations automated. To automate identify anomalies like crackles, wheezes and/or both, in this work, we propose a hybrid deep learning model combining Convolutional Neural Network (CNN) model, ResNet34 as feature extractor, Long Short-Term Memory (LSTM) predictor, along with novel augmentation technique called homogeneous padding over ICBHI-2017 dataset. We have also added an attention layer extractor allow learn important region vector. The proposed has outperformed recent state-of-the-art models regard. found that inclusion layer, LSTM predictor improved performance our 2-class 4-class anomaly predictions.

Language: Английский

---

Performance evaluation of lung sounds classification using deep learning under variable parameters

EURASIP Journal on Advances in Signal Processing, Journal Year: 2024, Volume and Issue: 2024(1)

Published: April 15, 2024

Abstract It is desired to apply deep learning models (DLMs) assist physicians in distinguishing abnormal/normal lung sounds as quickly possible. The performance of DLMs depends on feature-related and model-related parameters heavily. In this paper, the relationship between a DLM, i.e., convolutional neural network (CNN) analyzed through experiments. ICBHI 2017 selected dataset. sensitivity analysis classification DLM three parameters, length frame, overlap percentage (OP) successive frames feature type, performed. An augmented balanced dataset acquired by way white noise addition, time stretching pitch shifting. spectrogram mel frequency cepstrum coefficients are used features CNN, respectively. results training test show that there exists significant difference among various parameter combinations. OP sensitive. higher OP, better performance. concluded for fixed sampling 8 kHz, frame size 128, 75% optimum under which relatively no extra computation or storage resources required.

Language: Английский

---