融合多层级特征的脑肿瘤图像分割方法北大核心CSCD

针对脑肿瘤图像分割中网络模型信息损耗、上下文信息联系不足及网络泛化能力较差导致分割精度较低的问题,提出了一种新型的脑肿瘤图像分割方法,该方法是通过深度门控卷积模块(depth gate convolution,DGC)和特征增强模块(feature enhancement module,FEM)组成的多层级连接(multi-level connection,MC)脑肿瘤分割模型。采用深度卷积模块降低特征信息在逐层传递的信息损耗;使用控制门单元(control gate unit,CGU)实现各个尺度的特征图的MC,其中组合池化来减少下采样过程中的信息丢失;通过FEM增强分割区域的特征权重。实验结果表明,预测分割脑肿瘤的整体肿瘤区(whole tumor,WT)、核心肿瘤区(tumor core,TC)和增强肿瘤区(enhancement tumor,ET)的Dice系数分别达到了0.92、0.84和0.83,Hausdorff距离达到了0.77、1.50和0.92,脑肿瘤分割精度相较于当前较多方法分割精度和计算效率较高,具有良好的分割性能。

Brain tumor image segmentation method based on multi-level features

Aiming at the problems of low segmentation accuracy caused by the loss of network model information,insufficient context information and poor network generalization ab ility in brain tumor image segmentation,a new brain tumor image segmentation method is proposed.Thi s method is a multi-level connected (MC) brain tumor segmentation model composed of depth gate co nvolution module (DGC) and feature enhancement module (FEM).The depth convolution module is used to reduce t he information loss of feature information transmitted layer by layer.The control gate unit (CGU) is used to realize the MC of each scale feature map,in which the combination pool ing is used to reduce the information loss in the down sampling process.The feature weight of the seg mented region is enhanced by the FEM.The experimental results show that t he Dice index of the whole tumor area (WT),tumor core area (TC) and enhanced tumor area (ET) predicted and segm ented brain tumors reaches 0.92,0.84 and 0.83 respectively,and the Hausdorff distance reac hes 0.77,1.50 and 0.92.Compared with many current methods,the segmentation accuracy and calculat ion efficiency of brain tumors are higher,and have good segmentation performance .