Osegnet-F-Unext: O-Segnet-Fusion-Unext for pulmonary lobe segmentation of Covid-19 using Computed Tomography image.

Abstract

PURPOSE: Recently, the severe acute respiratory syndrome coronavirus disease-2019 (COVID-19) has shown great health issues globally. The COVID-19 disease has caused economic issues, health issues and a greater number of deaths. However, many advanced techniques have been developed to segment the pulmonary lobe of COVID-19, but that techniques were inadequate for pulmonary lobe segmentation. The traditional models are not capable of capturing subtle changes in lung tissue and complex lung structure that are caused by COVID-19. Therefore, to improve the pulmonary lobe segmentation of COVID-19 using Computed Tomography (CT) image, an effective method, namely O-Segnet-Fusion-Unext (Osegnet-F-Unext), is proposed. METHOD: At first, the input COVID-19 CT image is fed to the image enhancement process. The image enhancement is performed by the power-law transformation technique. Finally, an enhanced image result is given to the pulmonary lobe segmentation. Therefore, the lobe segmentation is carried out with the proposed Osegnet-F-Unext, where the developed model is the combination of O-Segnet and Unext networks. Additionally, the evaluation of the performance of the Osegnet-F-Unext model is done by using segmentation accuracy, Jaccard coefficient and dice coefficient. RESULTS: Experimental output shows that the O-Segnet F-Unext method obtained the highest segmentation accuracy as 91.55%, Jaccard coefficient as 0.903 and dice coefficient as 0.898. CONCLUSION: The proposed Osegnet-F-Unext model provides a reliable and effective method for the pulmonary lobe segmentation of COVID-19.