CovLIS-MUnet segmentation model for Covid-19 lung infection regions in CT images

One of the most serious health concerns facing the world is coronavirus disease (COVID-19). COVID-19 is a virus that is highly infectious and contagious. The RT-PCR test is not the only way to diagnose COVID-19; many other alternatives are available like Lung Computed Tomography (CT) imaging. Large variances in texture, size, and location of infections make manual segmentation of lung CT images time-consuming and difficult. We present an effective segmentation model Covid-19 Lung Infection Segmentation based on a multi-special block Unet (CovLIS-MUnet) to improve the segmentation process. The primary goal of our study is to segment the lung and infection parts of CT scan images. We integrate a multi-special block (MSB) with a Convolutional block in the encoder and bridge phases, which helps to study contextual information and COVID-19 infection-related characteristics, thereby resulting in accurate segmentation results and improved prediction accuracy. The proposed method has been evaluated on a COVID-19 CT segmentation dataset. The findings of the qualitative experiments suggest that the CovLIS-MUnet model can accurately segment lung and COVID 19 affected areas, with accuracy 0.9964 and 0.998. The proposed CovLIS-MUnet consistently achieves much better segmentation performance across four widely used evaluation parameters, according to experimental data. The proposed model is good as compared to other existing models. Medical professionals will benefit greatly from the usage of CovLIS-MUnet segmentation architecture because, in addition to aiding in the diagnosis of COVID-19, it allows them to determine how serious the illness is through infection projections.