Covid-19 detection: a Deep Learning Approach based on Wavelet Transform

Abstract

While being considered one of the most accurate and reliable techniques for detecting the Corona virus cases, the Reverse Transcription-Polymerase Chain Reaction (RT-PCR) remains quite expensive and requires advanced infrastructure and qualified manpower that are not always available in developing countries, fact that delays the diagnosis and increases the risks of mortality. Motivated by this concern and believing that applying AI techniques on X-Ray or Computed Tomography (CT) images can help detecting Covid 19 cases in a cheaper, faster, and accurate manner, a Wavelet Transform Enhanced deep learning Model (WTEM) is proposed to detect Covid-19 cases. More particularly, this paper presents a solution based on the combination of the wavelet transform technique with deep learning (DL) models. WTEM is compared to the DarkCovidNet model proposed by Ozturk et al. in (Ozturk et al., 2020) and to the VGG-19 model (Hansen, 2020). This solution outperforms both models in terms of accuracy, recall, and F1-Score in addition to significant reduction of the processing time and memory which makes it suited forY resource-constrained embedded systems.