【摘要】 提出一种基于BP网络分割CT图像序列中肝实质的方法。首先选取训练样本,提取样本图像中肝脏的纹理特征,作为输入向量,以对训练样本手工分割的结果作为导师信号,对BP神经网络进行训练,再用训练好的网络对CT图像序列中的肝实质进行分割,最后对分割后的结果进行三维区域生长及孔洞填充处理。实验结果表明:该方法能够有效的对肝脏纹理特征明显的CT图像序列进行分割,可用于CT图像序列的自动分割。更多还原

【Abstract】 It puts forward a method to segment liver parenchyma in CT image sequence based on BP network. Firstly, select the training sample, extract the texture features of liver in sample image as input vector, take the result of manual segmentation on training sample as teacher signal to train BP neutral network, and then make segmentation on liver parenchyma in CT image sequence by trained network, and finally, make 3-D domain growth and hole filling and treat-ment for the result after segmentation. The experimental result shows that this method can effectively segment CT image sequence with obvious liver texture features, which can be applied to the automatic segmentation of CT image sequence.更多还原