Abstract: Thoracic and abdominal medical image segmentation plays an irreplaceable role in the screening and diagnosis of early lesions in clinical medicine. Due to the presence of artifacts and blurred boundaries in thoracic and abdominal medical images, existing models are prone to segmentation errors. This paper proposes a new semantic segmentation network, MQUNet, which solves the problems caused by the above factors by constructing a dual-branch feature interaction mechanism and combining cross-sample feature fusion. MQUNet utilizes a dual branch architecture based on CNN-Transformer to extract local and global features of the current image, in order to capture local details and spatial correlations between organs as comprehensively as possible. It also integrates the two features by adding a negative attention branch based on an improved activation function, while taking into account both similarities and differences between features, avoiding the loss of semantic information. In addition, a memory queue is established based on the category labels of the training samples to store anatomical prior features. Through cross sample feature retrieval and weighted fusion, stable anatomical prior features from the training samples are dynamically added to the current segmentation process, effectively correcting feature shifts caused by artifacts and blurring. The Dice coefficient index of this method reached 92.53% and 85.17% on publicly available datasets of cardiac cancer (ACDC) and abdominal multi organ (Synapse), respectively, which is superior to existing segmentation methods. This indicates that MQUNet has significant advantages in chest and abdominal medical image segmentation tasks.