ABSTRACT :
Prediction of respiratory diseases such as COPD(Chronic obstructive pulmonary disease), URTI(upper respiratory tract infection), Bronchiectasis, Pneumonia, Bronchiolitis with the help of deep neural networks or deep learning. We have constructed a deep neural network model that takes in respiratory sound as input and classifies the condition of its respiratory system. It not only classifies among the above-mentioned disease but also classifies if a person’s respiratory system is healthy or not with higher accuracy and precision.
#To run the Flask local web application goto this repository
Index Terms:
Respiratory disease recognition, Deep neural network,GRU(Gated Recurrent Unit), sound data, data augmentation, feature extraction, classification
-> First download the git repo as a zip and extract it are else git clone to your environment.
-> Then get the dataset which's link is provided inside the Link_TO_Dataset.txt in the give repository and extract the downloaded dataset folder 'archive' to the main repository of our porject files present
About the Dataset:
These recordings were taken from 126 patients. It includes 920 annotated recordings of varying length from 10s to 90s.There are a total of 5.5 hours of recordings containing 6898 respiratory cycles - 886 contain wheezes, 1864 contain crackles and 506 contain both crackles and wheezes. The patients span all age groups - children, adults and the elderly. The respiratory sounds in the dataset are of different category such as Healthy, COPD(Chronic obstructive pulmonary disease), URTI(upper respiratory tract infection), Bronchiectasis, Pneumonia, Bronchiolitis, Asthma, LRTI(Lower respiratory tract infection) which would be classified or predicted by out neural network model.
-> To the accuracy of our model and data balancing we have droped the LRTI and ASTHMA from our dataseta and trained because of its very low level of number count in the dataset collection.
-> Iam using Python version 3.7.5 which is very flexible and capatible with all pip packages for me compared to other bug fitted new versionsss....!
-> So version is your choice and then run the command, to install all the requirements and the supporting libraries for our project. # pip install -r requirements.txt
-> After installing all these packages successfully just run the respiratory-disease.py for training the dataset and deploying a new model
python respiratory-disease.py
-> If you dont want to train again means, i have attached my locally trained deep learning model in the repo using that you can direct predict disease easily without any overhelming tasks or cpu hanging....!!!
python "result of model.py"
-> To get the result of first audio from the dataset of the patients run the file results of model.py, it will give you a predicted disease with probobility of other disease witht there percentages also.
-> To run in GUI format and to check a new audio file means, i have created a Simple GUI for getting the input as path of the audio file with the extension and return the results in the desktop manner using the kivy frame work in python.
python main.py
-> To run GUI mode just run the main.py inside the main repo of our project and give the input below:
#archive/Respiratory_Sound_Database/Respiratory_Sound_Database/audio_and_txt_files/104_1b1_Pr_sc_Litt3200.wav
nor like below formats
archive//Respiratory_Sound_Database//Respiratory_Sound_Database//audio_and_txt_files//104_1b1_Pr_sc_Litt3200.wav
or
archive\Respiratory_Sound_Database\Respiratory_Sound_Database\audio_and_txt_files\104_1b1_Pr_sc_Litt3200.wav
-> After giving input in click the Test button and you can view the predicted disease immediately in the GUI in the desktop easily.
INTRODUCTION:
In this research paper, we are going to discuss how deep learning could be used in the recognition of respiratory disease just from the respiratory sound. Respiratory audios are important indicators of respiratory health and respiratory disorder. For example, a wheezing sound is a common sign that a patient has an obstructive airway disease like asthma or chronic obstructive pulmonary disease (COPD). We have approached the problem with different neural network model architecture, and choose the model with would give us the best possible results, we also performed data augmentation over the data-set. The data-set we have used consists of respiratory sounds taken from different patients from different locations around the chest. We have used Accuracy score, Precision score, Recall score, f1-score, Cohen’s kappa score, Matthews correlation coefficient as metrics to evaluate and compare the performance of different models against the same data-set. With this model we have achieved an accuracy of 95.67 % ± 0.77 %,precision of 95.89 % ± 0.8 %,Sensitivity of 95.65 % ± 0.753 %,f1-score of 95.66 %±0.79 %, Cohen’s kappa score of 94.74 %±0.96 % and Matthews correlation coefficient of 94.79 % ± 0.96 %.
-> All the deep learning model structure and the architecture are attached a pic inside the screenshots folder
-> And The metrics that we have used to evaluate the performance of the model are Accuracy, Precision, Recall/sensitivity, F1-score, Cohens kappa score(CK), Matthews correlation coefficient(MCC). For all the above metrics if the score is 1, then the model is performing in the best way possible and predicting every class perfectly. Lower the score more is the model facing difficulty while predicting the correct class and also indicates poor model performance.
CONCLUSION
Medical research and medical science could be progressed
further with the help of artificial intelligence. The neural network architecture has performed better than our expectations. But still, it needs a lot of improvements to achieve higher accuracy during prediction. We hope our research would inspire future researchers to work on this subject and wish they would approach with a better and encouraging solution.
In case of any doubts are struggles contact me and feel free to ping me you stranger.....! phone: 9025421765 gmail: k.gokulappaduraikjgv@gmail.com