Eye fundus tissue three-dimensional segmentation method

The invention discloses a fundus tissue three-dimensional segmentation method, which comprises the steps of acquiring a three-dimensional medical fundus image data set, dividing an image into a training set and a verification set, inputting the data set into a proposed deep neural network for training and segmentation, and finally analyzing a segmentation result. According to the invention, an extrusion convolution block (Sconv) is introduced to replace the traditional three-dimensional convolution operation with less parameter quantity and calculation quantity, and a global attention mechanism module is combined to process the feature maps of the fused encoder and decoder blocks on the same layer. The optimized model can accurately segment the three-dimensional fundus tissue and assist doctors in clinical diagnosis, and has important clinical significance. 本发明公开一种眼底组织三维分割方法，步骤包括获取三维医学眼底图像数据集、对图像划分训练集和验证集，并将数据集输入到所提出的深度神经网络中进行训练与分割，最后对分割结果进行分析。本发明引入了挤压卷积块(SConv)用更少的参数量和计算量替代传统的三维卷积操作，并结合全局注意力机制模块对已融合的同一层编码器和解