Wheeze and crackle analysis are important parameters helpful in the clinical diagnosis of respiratory diseases. With the advent of deep learning techniques, objective medical diagnostics is made possible and its performances are evaluated over various parameters. In this work, the presence of wheeze and crackle sounds in lung audio recordings are analyzed tested for consistency, and presented. Mel-Spectrogram images which held by using Convolutional Neural Network models. One advantage of using deep learning for wheeze and crackle analysis is that it can improve accuracy and consistency compared to human auscultation. First, a large dataset ICBHI 2017 of lung sound recordings is collected, with corresponding labels indicating the presence and type of wheezing and crackling sounds. The model is subsequently tested on a distinct dataset to assess its performance. The model can then be optimized and refined based on the results of the testing. One major challenge is the availability of high-quality labeled data, as this is required for training the model. Overfitting presents an additional challenge, as the model may become excessively tailored to the training data, leading to subpar performance on novel data. Despite these challenges, there have been several successful applications of deep learning for wheeze and crackle analysis. In conclusion, deep learning has shown great promise for automating the analysis of wheeze and crackle sounds in respiratory diseases.
Wheeze and Crackle Analysis Using Deep Learning
2023-11-22
536278 byte
Conference paper
Electronic Resource
English
Source Mechanisms of Jet Crackle
AIAA | 2012
|Skewness as means for separating crackle from screech
AIAA | 2019
|Westergaard Stress Function of I - II Mixed Crackle
British Library Online Contents | 2003
|Crackle in the noise of high-performance aircraft
AIAA | 2018
|