基于BP神经网络的脑肿瘤MRI图像分割

Brain Tumor Segmentation of MRI based on BP Neural Network

利用BP神经网络技术对MR脑肿瘤图像中的肿瘤区域与正常组织区域进行分割,以辅助医疗诊断与治疗。首先,人工分割出部分影像中的肿瘤组织与正常组织作为已知样本;其次,在BP神经网络模型中输入已知样本中进行训练;最后,用训练好的BP神经网络处理其他脑肿瘤图像。BP神经网络能够有效分割MR脑肿瘤图像,辨别出肿瘤与周围正常组织的差异,但模糊区域也常被误判为肿瘤。因此,本研究提出进一步对模糊区域样本进行针对性训练与特殊的滤波处理,所得结果有较大改进。BP神经网络能有效地进行脑肿瘤MRI图像分割,但在使用时仍需正确选择输入样本的区域和范围并结合特殊的滤波处理。

To distinguish normal tissue area and the tumor area in MRI brain images by using BP neural network technology First-ly,the sample image segmentation with the image processing software was made.Secondly,the sample image segmentation was put into the neural network model as the known sample for training.Finally,the unknown MRI with the trained neural network was classified. Applying the BP neural network in the segmentation of brain MRI could identify the difference between the tumor and the surrounding normal tissues effectively.To improve the accuracy,we input samples of fuzzy area with specific training in the model and used special filtering method.The BP neural network can realize accurate and effective segmentation of brain MRI,but it’s important to choose in-put sample correctly and use special filtering method.