A hybrid deep convolutional neural network model for improved diagnosis of pneumonia

Pneumonia is an infection that inflames the air sacs in lungs and is one of the prime causes of deaths under the age of five, all over the world. Moreover, sometimes it became quite difficult to detect and diagnose pneumonia by just looking at the chest plain X-ray images. Therefore, we propose a hy...

Full description

Saved in:
Bibliographic Details
Published in:Neural computing & applications 2024-02, Vol.36 (4), p.1791-1804
Main Authors: Mann, Palvinder Singh, Panchal, Shailesh D., Singh, Satvir, Saggu, Guramritpal Singh, Gupta, Keshav
Format: Article
Language:English
Subjects:
Accuracy
Artificial Intelligence
Artificial neural networks
Computational Biology/Bioinformatics
Computational Science and Engineering
Computer Science
Data Mining and Knowledge Discovery
Diagnosis
Image Processing and Computer Vision
Lungs
Medical imaging
Neural networks
Original Article
Pneumonia
Probability and Statistics in Computer Science
Probability density functions
Recall
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Pneumonia is an infection that inflames the air sacs in lungs and is one of the prime causes of deaths under the age of five, all over the world. Moreover, sometimes it became quite difficult to detect and diagnose pneumonia by just looking at the chest plain X-ray images. Therefore, we propose a hybrid deep convolutional neural network model (HDCNN) for improved diagnosis of pneumonia, to simplify the detection for medical practitioners and specialists. In this proposed model, image preprocessing is performed using Student's t distribution, a compact probability density function (cPDF), for better sampling and segregation between the healthy and infected part of lungs, to improve the predictions. Further, a hybrid deep convolutional neural network model is built to extract image features by fine-tuning the pretrained models, viz. Resnet-50, EfficientNet, VGG-16, MobileNetV2 and DenseNet to achieve better results of diagnosis. The proposed hybrid model is analyzed using Grad-CAM visualization, which produces a course localization map, highlighting the infected region in the image used for prediction. The proposed hybrid model is evaluated based on governing parameters, viz. precision, recall, F1-score and accuracy. The results show our proposed model achieves precision of 97.47%, recall 98.09%, F1-score of 97.77% and overall accuracy of 97.69% as compared to other existing models.
ISSN:0941-0643
1433-3058
DOI:10.1007/s00521-023-09147-y