A Novel Lightweight Swin-Unet Network for Semantic Segmentation of COVID-19 Lesion in CT Images

Abstract

The Corona Virus Disease 2019 (COVID-19) is highly infectious, has been spread worldwide, caused a global pandemic, and seriously endangered human health and life. The most effective methods for halting and stopping the transmission of the Corona Virus include early detection, quarantine, and successful treatment. Because it exhibits significant imaging characteristics for COVID-19 lesions in chest computed tomography (CT), it can be used to diagnose COVID-19. Aiming at the inaccuracies of uneven gray distribution, irregular regions, multi-scale, and multi-region segmentation in COVID-19 CT images. This paper proposed a novel Swin-Unet network to improve the accuracy of multi-scale lesion segmentation in COVID-19 CT images. First, in the double-layer Swin Transformer blocks of the Swin-Unet, a residual multilayer perceptron (ResMLP) module was introduced and replaced the multi-layer perceptron (MLP) module to reduce the loss of features during the transmission process, thereby improving the segmentation precision of multi-scale lesion areas. Second, the uncertain region inpainting module (URIM) was added after Linear Projection, which can refine the uncertain regions in the segmentation features map, thereby improving the segmentation accuracy of different lesion regions. Third, a new loss function DF was designed. It can effectively improve the small target segmentation effect and thus improve the multi-scale segmentation result. Finally, the proposed method was compared to other methods on the public dataset. The Dice, Precision, Recall, and IOU of the proposed method are 0.812, 0.780, 0.848, and 0.683, respectively, which are better than the other models. Moreover, our model has fewer parameters and faster reasoning speed. The proposed method achieves excellent segmentation results for multi-scale and multi-region lesions, and it will be more beneficial in aiding COVID-19 diagnosis and treatment.

HMSAM-UNet: A Hierarchical Multi-Scale Attention Module-Based Convolutional Neural Network for Improved CT Image Segmentation

“…The proof set, in particular, assisted optimization and verification of the model by furnishing an unbiased means to analyze hyperparameters and track progress discrete from the test set. This rigorous methodology ensures HMSAM-UNet is properly trained and validated for optimal medical image segmentation [32]- [34].…”

Deep-learning-based pelvic automatic segmentation in pelvic fractures

“…Kai et al 15 utilized the Swin U-Net for tumor segmentation in breast cancer to enable rapid diagnosis. Gao et al 16 proposed a lightweight Swin U-Net model for segmenting COVID-19 lesions in CT images, achieving excellent segmentation results in multiple lesion regions. Urakawa et al 17 used the VGG model to classify pelvic fractures, focusing on straightforward lesion classification.…”

FKD-Med: Privacy-Aware, Communication-Optimized Medical Image Segmentation via Federated Learning and Model Lightweighting Through Knowledge Distillation

“…The study of medical image segmentation presents challenges due to the scarcity of high-quality medical imaging data, particularly in contexts involving user privacy concerns. Consequently, extensive research has been dedicated to enhancing segmentation model structures over the past few decades [4].To address this bottleneck, the focus has largely been on the innovation of neural network architectures. The U-net model [5], introduced in 2015, employed an encoderdecoder framework to amplify segmentation accuracy despite the constraints of limited datasets.…”