基于BP神经网络的MRI分割

本文提出了基于ＢＰ神经网络的脑部多回波磁共振图像分割方法。输入图像包括Ｔ１，Ｔ２，ＰＤ三种加权特征图像，分割是基于各像素点灰度值进行的。首先通过一组已标号的训练样本训练神经网络分类器，不正确的分割结果通过交互方式修正训练点，然后再次训练，直至满意的分割结果。

BP-Based Artificial Neural Network of MRI Segmentation

This paper studies a method of supervised BP-based artificial neural network in segmentation multi-echo magnetic resonance images of the brain. Input data consisted of T1-weighted, T2-weighted and PD-weighted images. The segmentation was based on the pixel intensity in each images. A set of labeled training patterns has to be collected for each tissue type by an user to train the classifier . A special interactive environment must be developed for immediate correction of any unreasonable results that arise from the inappropriate choice of training samples.