No Detail Found!

Improved deep learning approaches for covid-19 recognition in ct images

Author(s)
HAMZA ABU OWIDA Department Of Medical Engineering, Faculty of Engineering, Al-Ahliyya Amman University, Zip-Code (Postal Address):19328, Amman, Jordan.
OMAR SALAH MOHAMED HEMIED Software Engineering Program, Computer Science Department, Faculty Of Computers And Information Assuit University, Zip-Code ( Postal Code): 71515, Assuit, Egypt.
RAMI S. ALKHAWALDEH Department Of Computer Information Systems, The University Of Jordan, Aqaba 77110, Jordan.
NAWAF FARHAN FANKUR ALSHDAIFAT Department Of Computer Science Prince Hussein Bin Abdullah, Faculty Of Information Technology, Al Al-Bayt University, Mafraq, 25113, Jordan.
SUHAILA FARHAN AHMAD ABUOWAIDA Department Of Computer Science Prince Hussein Bin Abdullah, Faculty Of Information Technology, Al Al-Bayt University, Mafraq, 25113, Jordan.
Abstract
Since the increasing risk of COVID-19, a set of actions have been achieved to develop tools to handle the spreading of the COVID-19 disease. Though testing kits were being used to diagnose the COVID19 infection, the process requires time and the test kits suffer from being lack. In COVID-19 management,the computed tomography (CT) is considered an important diagnostic method. Taking into account largenumber of exams performed in high case-load situations, an automated method may help to encourage and save time for diagnosing and identifying the disease. Several deep learning tools have recently beendeveloped for COVID-19 scanning in CT scans as a technique for COVID-19 automation and diagnosticassistance. This article aims to explore the rapid recognition of COVID-19 and proposes an advanced deep learning technique, derived from improving the ResNet architecture as a transfer learning model. The architecture design of the proposed model is based on alleviating the connections between the blocks of the ResNet-50 model. This reduces the training time for scale-ability and handles the problem of vanishing gradient with relevant features for recognizing COVID-19 from CT images. The proposed model is evaluated using two well-known datasets of COVID-19 CT examined with a patient-based split. The proposed model attains a total back-bone accuracy of 98.1% with 97%, and 98.6% specificity and sensitivity, respectively.
Publication Details
Page(s) 4925-4931
DOI DOI not available
Published Journal: Journal of Theoretical and Applied Information Technology, Volume: 100, Issue: 13, Year: 2022

Request the full text from PASTIC

Keywords
COVID-19, COVID-19, deep learning, Transfer Learning Model., CT Images, SARS-CoV-2
0

Citations

0

Downloads

30

Views