基于GAN与轴向区块注意力的心脏磁共振图像分割

Cardiac Magnetic Resonance Image Segmentation Based on GAN and Axial Block Attention

心脏磁共振图像分割对心功能分析和心脏疾病诊断具有十分重要的意义。针对传统心脏分割方法对心脏MR图像特征提取不全面,以及基于注意力机制的深度学习方法参数量过大的问题,设计一种基于GAN与轴向区块注意力的心脏磁共振图像分割模型,对图像特征进行多尺度、全方面地提取,结合GAN策略提升模型性能。实验结果表明,模型实现了图像的有效分割并提高了分割结果与真实标签之间的一致性。

Cardiac magnetic resonance image segmentation is of great significance for cardiac function analysis and cardiac disease diagnosis.Aiming at the problem that traditional cardiac segmentation methods can not fully extract features from cardiac MR images,and the Deep Learning method based on the Attention Mechanism has too many parameters.A cardiac magnetic resonance image segmentation model based on GAN and axial block attention is designed to extract image features at multiple scales and in all aspects,and combined with GAN strategy to improve model performance.Experimental results show that the model achieves effective segmentation of images and improves the consistency between segmentation results and real labels.