体素形态学方法与人工神经网络结合在Alzheimer病诊断中的应用研究

Combining Voxel-based Morphometry with Artificial Neural Network Theory in the Application Research of Diagnosing Alzheimer's Disease

目的将体素形态学方法与人工神经网络方法相结合,用于阿尔茨海默氏病的诊断,探索自动、客观的Alzheimer临床诊断新途径。方法研究对象包括10名可能的Alzheimer病人和12例年龄与性别与之相匹配的正常老年人,在Siemens Sonata1.5T超导MR成像系统上获得精细的三维结构图像,首先利用体素形态学方法得到了Alzheimer患者灰质发生缺失的10个区域,然后统计每个区域的灰度值,在此基础上,利用反向传播神经网络对数据进行分析处理。结果海马、海马旁回、内嗅皮层、杏仁体、尾状核头部、颞中回、扣带回、顶下小叶、岛叶和前额背外侧区域,在训练样本数为18的情况下,利用神经网络能够达到100%的判别效果。结论将人工神经网络方法和体素形态学(VBM)方法相结合诊断Alzhei-mer患者,具有良好的临床应用前景,可能成为一种实用和可靠的临床诊断工具。

Objective Combine voxel-based morphometry（VBM） with artificial neural network in diagnosing Alzheimer＇s disease（AD）,and explore the new way of diagnosing AD for clinical applications. Methods Ten probable AD and 12 normal controls with matched age and gender were studied. The data were collected on Siemens 1.5 T Sonata MRI systems. Areas of the gray matter loss in AD patients were obtained with VBM. The data was processed and analyzed by back-propagation neural network（BP）. Results The ten specific areas comprising hippocampus,parahippocampal gyrus,entohinal cortex,amygdale,the head of the caudate nucleus,middle temporal gyrus,cingulated gyrus,inferior parietal lobule,insula and dorsolateral prefrontal cortex could completely differentiate AD from normal controls by back-propagation neural network with eighteen training samples. Conclusion The method using VBM in combination with artificial neural network to diagnose AD patient will have a good perspective for clinical applications and is expected to become a useful and reliable clinical tool.