【摘要】 脑肿瘤图像采用传统方法难以实现高精度分割，而手动分割图像费时费力，为此提出一种基于改进UNet的脑肿瘤图像分割算法。首先，在模型的上采样部分嵌入注意力机制，提高主要特征权重；其次，使用迁移学习增强模型泛化能力；最后，进行实验分析。实验结果表明，该算法在脑肿瘤图像分割上具有更好的效果。更多还原

【Abstract】 Traditional methods are difficult to achieve high-precision segmentation of brain tumor images, while manual image segmentation is time-consuming and laborious. Therefore, an improved UNet based brain tumor image segmentation algorithm is proposed. Firstly, embedding attention mechanism in the upsampling part of the model to improve the weight of the main features. Secondly, using transfer learning to improve the model’s generalization ability. Finally, conduct experimental analysis. The experimental results show that the algorithm has better performance in brain tumor image segmentation.更多还原