基于特征融合视觉显著性的医学图像分割

Medical image segmentation based on visual saliency of feature fusion

医学图像分割结果的准确性对医生诊断病情并制定相应的治疗策略具有重要价值。针对现有的医学图像进行分割时由于没有考虑人类视觉显著性机制因素导致分割精度不高的问题,提出一种基于特征融合视觉显著性的医学图像分割方法。首先基于频率调谐生成待分割医学图像的显著图,得到图像的显著区域并突出医学图像的边缘轮廓,然后分别提取其颜色特征和纹理特征将其作为反向传播神经网络的输入向量,在此基础上用神经网络分类器模型对图像进行分割。通过实验进行验证,结果表明该方法获得了较好的分割精度和分割效率,本文所提方法为医学图像的准确分割提供了一种新途径。

The accuracy of medical image segmentation results is of great value for the doctor to diagnose the disease and develop appropriate treatment strategies. In view of the problem that the existing medical image segmentation methods have not considered the human visual saliency which leads to a lower segmentation accuracy, we propose a method for medical image segmentation based on the visual saliency of feature fusion. Firstly, based on frequency tuning, a saliency map is generated for the medical images to be segmented in order to obtain the salient regions and highlight the edge of the medical images. Then the color features and texture features are extracted separately to form the input vectors of back propagation neural network. Finally, the back propagation neural network model is used to achieve medical image segmentation. The proposed method is verified by experiments, and the results show that the proposed visual saliency of feature fusion algorithm for medical images segmentation could achieve a high efficiency and an ideal accuracy. The proposed method provides a new way for medical image segmentation.