ResearchBib Share Your Research, Maximize Your Social Impacts
Sign for Notice Everyday Sign up >> Login

Automatic detection of infected areas of CT images in COVID-19 patients using Inf-Seg-Net

Journal: International Journal of Advanced Technology and Engineering Exploration (IJATEE) (Vol.10, No. 104)

Publication Date:

Authors : ; ;

Page : 930-946

Keywords : 930;

Source : Downloadexternal Find it from : Google Scholarexternal

Abstract

The swift spread of the coronavirus disease (COVID-19) makes it extremely difficult for early detection and diagnosis of the virus, necessitating timely care. Numerous research institutes, laboratories, diagnostic facilities, non-governmental organizations, and government-funded organizations collaborate daily to identify challenges that arise throughout the COVID-19 virus detection procedure. The first screening method utilized to locate COVID-19 was reverse transcription polymerase chain reaction (RT-PCR). However, advancements in technology have paved the way for the use of computed tomography (CT) imaging in early screening. Radiologists and research scientists are now exploring the potential of artificial intelligence (AI) and deep learning (DL) techniques to develop an automated disease detection model utilizing CT images for screening purposes. The aim of this research is to simplify infection segmentation by using Inf-Seg-Net, a network-based technique with dense UNets and residual blocks for infection classification. The proposed framework involves three stages: preprocessing CT images using contrast limited adaptive histogram equalization based on non-local mean filter (CLAHEN), followed by logarithmic non-maxima suppression (LNMS) for lung segmentation, and an infection segmentation network (Inf-Seg-Net) for infection classification. In this study, various DL models, including ResNet, SegNet, and UNet, were evaluated for their effectiveness in diagnosing COVID-19 infection using a real-time dataset of lung CT images. The proposed Inf-Seg-Net model demonstrated promising results, providing high-quality masking on the lung segmentation images. It achieved remarkable performance metrics, including 98.06% accuracy, 96.15% Jaccard index, 100% sensitivity, 98.1% precision, and 98.73% F1 score, indicating its potential for detecting infections from CT scans and outperforming existing models. These findings highlight the potential of AI and DL techniques in enhancing COVID-19 diagnosis and pave the way for more efficient and accurate screening methods.

Last modified: 2023-08-03 20:21:06