Multiclass Deep Transfer Learning for Covid-19 Classification
5th International Conference on Vocational Education and Electrical Engineering, ICVEE 2022
; : 60-64, 2022.
Article
in English
| Scopus | ID: covidwho-2136340
ABSTRACT
Over 500 million people have been infected with COVID-19 since it first appeared, including more than 6 million cases in Indonesia. Although COVID-19 has the potential to cause pneumonia, COVID is not always the sole cause of the illness, necessitating the need for another rapid and precise approach to disease classification. Additionally, it is not only dependent on the Polymerase Chain Reaction (PCR) technique, which is costly and labor-intensive. The study of chest X-ray images can be one quick and accurate way of helping to confirm the disease. It is necessary to investigate the multiclass classification of diseases with comparable clinical characteristics since COVID-19-related diseases can vary. This investigation chose pneumonia, COVID-19, and Normal as the deep learning model's three target classes. The mobileNet-based deep transfer learning accuracy obtained was 0.95%, while the recall obtained was 0.93%, 0.97%, and 0.96%, respectively, where the targets were three classes (COVID, Pneumonia, and Normal). Additionally, the Covid class precision value received the perfect score, while the Normal and Pneumonia classes received the same for the f1-score. © 2022 IEEE.
Full text:
Available
Collection:
Databases of international organizations
Database:
Scopus
Language:
English
Journal:
5th International Conference on Vocational Education and Electrical Engineering, ICVEE 2022
Year:
2022
Document Type:
Article
Similar
MEDLINE
...
LILACS
LIS