Accès ouvert

Multiscale aware classification of COVID-19 from Chest X-Ray using a spatially weighted atrous spatial pyramid pooling CNN

Article scientifique 2023 Anglais

Résumé

Abstract COVID-19 is a severe respiratory tract infections which can range from mild to lethal. COVID-19 caused by SARS-CoV-2 can readily spread through direct or indirect contact with an infected person. This high spread rate pressure on the health care systems and requires non time-consuming methods for diagnosing. Convolutional Neural Networks (CNN) show a great success for various computer vision tasks. However, CNN like many computer vision models is a scale-variant model and requires expensive computation. In this paper, a novel micro architecture is proposed for multiscale feature extraction and classification. Proposed CNN learns multiscale features using a pyramid of shared convolution kernels with different dilation, atrous, rates. Proposed CNN is an attention based mechanism that is used to guide and select correct scale for each input. Proposed CNN is an end-to-end trainable Network. It achieved a 0.9929 for F1-score tested on QaTa-Cov19 benchmark dataset with a total of 5,040,571 trainable parameters.

Citer ce document

Fetoh, M., Amin, K., Hamad, A. (2023). Multiscale aware classification of COVID-19 from Chest X-Ray using a spatially weighted atrous spatial pyramid pooling CNN. https://doi.org/10.21203/rs.3.rs-3043485/v1

Accès au document

Texte intégral en lecture en ligne, réservé aux abonnés SPHAERO et aux membres de l'institution. Se connecter

Voir l'article sur le site de la revue

Auteur(s)

Mahmoud Z. Fetoh Khalid Amin Ahmed Mahmoud Hamad

Statistiques

Consultations : 1

Téléchargements : 0