<p>Cardiac ultrasound image segmentation is a crucial task in medical image analysis. It provides efficient and accurate support for early disease screening and treatment plans. Among current deep learning methods, images acquired by different devices exhibit differences in resolution and contrast, resulting in uneven image quality and increasing the difficulty of accurate segmentation. Furthermore, Convolutional Neural Networks (CNNs) excel at local feature extraction but lack global dependencies, and Transformers capture global context yet lack fine-grained spatial information, leading to coarse upsampled segmentation. To address these limitations, this paper proposes the bit plane enhanced fusion network (BPEFNet), a segmentation method for cardiac ultrasound images. This method integrates the grayscale information of the original image and the binary image information from the 3 MSB planes. First, dual-information feature extraction (DIFE) is employed to extract both local and global information from the original image and the bit plane. Second, by leveraging the tri-branch attention cooperative interaction (TBACI) to fuse local and global features, it enhances the ability to integrate multiscale and multispatial information. The final features are obtained by fusing through the hierarchical context multi-scale aggregation (HCMSA) module. Experiments on the CAMUS dataset show that the overall Dice coefficient of BPEFNet reaches 0.9364 and 0.9196 for end-diastole (ED) and end-systole (ES) phases, respectively. The EchoNet-Dynamic dataset shows that the overall Dice coefficient of BPEFNet reaches 0.9350 and 0.9127 for ED and ES phases, respectively. Compared with other methods, it performs the best in both subjective visualization results and objective quantitative metrics, enabling accurate segmentation of cardiac structures. Our model exhibits good segmentation capability and demonstrates potential in clinical applications.</p>