Title: A flexible mobile application for image classification using deep learning: a case study on COVID-19 and X-ray images
Authors: Elilson Santos; Omar Andres Carmona Cortes; Bruno Feres De Souza
Addresses: Programa De Pós-Graduação em Engenharia da Computação e Sistemas (PECS), Universidade Estadual do Maranhão (UEMA), São Luis, Maranhão, Brazil ' Departamento De Computação (DComp), Instituto Federal De Educação, Ciência e Tecnologia do Maranhão (IFMA), São Luis, Maranhão, Brazil ' Departamento De Engenharia da Computação (ECP), Universidade Federal do Maranhão (UFMA), São Luis, Maranhão, Brazil
Abstract: This paper proposes a flexible mobile application for embedding any CNN-image-based classification model, providing a computer application to assist health professionals. Two approaches are suggested: an embedded offline and a running online model via web API. To present the applicability of the mobile software, we used a CNN COVID-19 classification based on X-ray images as a case study. Still, any other image-based classification application could have been used. We used a popular Kaggle database consisting of 7178 X-ray images divided into three classes: Normal, COVID-19, and Viral Pneumonia. We tested 14 state-of-art CNNs to decide which one to embed. The VGG16 achieved the best performance metrics; therefore, the VGG16 was embedded. The software production methodology was applied based on the built model, class diagram, use cases and execution flow, besides designing a web API to execute the back-end classification model.
Keywords: mobile application; medicine 4.0; CNN; COVID-19; X-ray.
International Journal of Computer Applications in Technology, 2022 Vol.69 No.2, pp.150 - 162
Received: 28 Jun 2021
Accepted: 16 Aug 2021
Published online: 11 Nov 2022 *