MRIのテクスチャ特徴量を用いた健忘型軽度認知障害の検出方法

抄録

In this study we propose a new method of texture analysis to detect aMCI (Amnestic Mild Cognitive Impairment) by using an MRI (Magnetic Resonance Imaging). The problem in the conventional texture analysis for aMCI diagnosis is that is only one pixel was applied to the detection. we therefore proposed to reconsider the pixel distance of texture features and gray scales by examining 70 MRI images of 37 aMCI patients against a 33 healthy elderly control group (70 subjects in total). In our method we increased the pixel distance to 2 or more and lowered considerably the graduation level. This allowed us to observe a clear difference between aMCI patients and the healthy control group. This discriminating analysis using the proposed texture features demonstrated an overall accuracy of 100% in our dataset. Thus, this indicates that our proposed method can be used as the new detection technique for aMCI.