Transfer Learning for COVID-19 Detection in Medical Images

As of late, the COVID infection 2019 (COVID-19) has caused a pandemic sickness in more than 200 nations, therefore impacting billions of people. To control the spread of the coronavirus, it is crucial to detect infected individuals and ensure their complete isolation to prevent further infection. Ch...

Full description

Saved in:
Bibliographic Details
Published in:SN computer science 2024-04, Vol.5 (4), p.344, Article 344
Main Author: Azhari, Maryam El
Format: Article
Language:English
Subjects:
Chest
Computed tomography
Computer Imaging
Computer Science
Computer Systems Organization and Communication Networks
Coronaviruses
COVID-19
Data Structures and Information Theory
Datasets
Disease control
Image classification
Information Systems and Communication Service
Learning
Medical imaging
Original Research
Pattern Recognition and Graphics
Performance evaluation
Software Engineering/Programming and Operating Systems
Viral diseases
Vision
X-rays
Citations: Items that this one cites
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:As of late, the COVID infection 2019 (COVID-19) has caused a pandemic sickness in more than 200 nations, therefore impacting billions of people. To control the spread of the coronavirus, it is crucial to detect infected individuals and ensure their complete isolation to prevent further infection. Chest X-rays and CT-scans have been proven to be very promising as signals of the infection can be clearly shown in lung areas. Transfer learning from ImageNet dataset has become the latent trend in medical imaging applications. However, there are major differences between ImageNet and medical imaging datasets. Therefore, the feasibility of transfer learning in medical applications remains questionable. This paper investigates the performance of five fine-tuned pre-trained models for chest X-rays and CT-scans classification in contrast with a deep CNN model built from scratch. DenseNet121, Resnet-50, Inception v2, Resnet101-V2 and VGG16 are selected and initialized with either random or pre-trained weights to classify augmented images into two classes: COVID and non-COVID. The performance evaluation proves the minuscule impact of training transfer learning models for good quality results, as all CNN models contribute almost equally to the classification and achieve considerable results in terms of precision, accuracy, recall and F1 score.
ISSN:2661-8907
2662-995X
2661-8907
DOI:10.1007/s42979-024-02675-x