Advanced UNet++ Architecture for Precise Segmentation of COVID-19 Pulmonary Infections

In the realm of medical image processing using deep learning, accurate segmentation of lesion areas in COVID-19 CT scans remains challenging. We introduce an advanced Unet++ model, tailored for precise segmentation of pulmonary infections in COVID-19-related CT imagery. Our methodology begins with meticulous data preprocessing. Based on the Unet++ framework, we integrate the MulitRes module to amplify lesion edge details in low contrast scenarios. Following feature extraction, the SENet module is incorporated, enriching the receptive field and emphasizing relevant feature channels. Experiments on established datasets indicate our approach not only surpasses existing models in accuracy but also adeptly preserves critical details, especially around lesion boundaries, enhancing the overall segmentation efficacy.