COVID-19 CT SCAN IMAGE SEGMENTATION USING TRANSFER LEARNING APPROACH

Abstract

The Computer tomography Scan imaging technique becomes an alternative approach of RTPCR test in COVID-19 diagnosis, disease staging and monitoring of treatment response evolution. The sensitivity of RTPCT is low neat about 70% with compare to the sensitivity of chest CT Scan images technique up to 98%. In this work, COVID-19 CT Scan image segmentation has been performed. The UNET Model has been selected as the baseline model for CT Scan Image Segmentation. The CT Scan Image segmentation performance of the deep learning models has been improved using the transfer learning approach. The IoU values have been improved from 86.94 to 88.90, 93.56, 94.34 using MobileNet as an encoder, DenseNet201 as an encoder, DenseNet169 as an encoder in Baseline UNET Models respectively.