中文摘要：

目的利用游程长方法研究多发性硬化表现正常脑白质的纹理特征。方法对宣武医院放射科提供的23例多发性硬化患者和23名健康志愿者T2WI图像选取感兴趣区,用游程长方法提取短游程因子等参量,然后依据这些特征参量用概率神经网络对样本分类。结果患者和健康志愿者分类正确率分别为84.6%和83.5%。结论对于本批数据游程长方法在一定程度上能反映出表现正常脑白质和正常脑白质的不同。

英文摘要：

Objective The run-length matrices were introduced in texture characteristic analysis on microcosmic pathological changes of normal appearing white matter （NAWM） in patients with multiple sclerosis （MS）. Methods The regions of interest （ROI） were gained from T2WI of 23 patients with MS and 23 normal controls from Xuanwu Hospital. Then the eigenvectors were obtained from run-length matrices, such as short run emphasis. And the probabilistic neural networks were used to classify these ROIs. Results The correctness rates of patients and controls were 84.6% and 83.5% respectively. Conclusion Run-length matrices could reflect some differences of two groups in a certain extent based on these images.