心脏病图像分割方法研究

Heart Disease Image Segmentation Method

研究心脏造影图像准确分割问题。心脏病变区域存在大量水状的囊肿,囊肿区域的像素会造成病变区域图像像素灰度大幅下降,与正常区域的灰度差减少。传统的图像分割方法多是基于边沿像素灰度差进行分割,大量来自水状病变区域的干扰像素使得病变区域边沿像素与正常区域像素发生灰度混淆,造成利用灰度差方法进行图像分割的准确率比较低的问题。为此提出了一种基于改进神经网络算法的心脏造影图像分割方法。提取图像中的关键细节特点,运用改进后的神经网络模型,对水状杂质像素的干扰进行迭代分割过滤。实验证明,运用神经网络方法能够有效提高心脏造影图像病变部位分割的准确率,具有很好的应用价值。

Study heart angiogram accurate segmentation problem. There are a lot of aquiform cyst in heart disease area, the cyst area can cause the pixel grayscale greatly decreased in disease area, and the grayscale difference be- tween normal area and the cyst area is reduced. The paper put forward an improved heart image segmentation method based on neural network algorithm. The key detail characteristics in the images were extracted, and the improved neural network model was used to segmente the aquiform impurities pixels. The experiment shows that this method can effectively improve image segmentation accuracy of the lesion.