Ito Koichi, Fujimoto Ryuichi, Huang Tzu-Wei, Chen Hwann-Tzong, Wu Kai, Sato Kazunori, Taki Yasuyuki, Fukuda Hiroshi, Aoki Takafumi
Annu Int Conf IEEE Eng Med Biol Soc. 2018 Jul;2018:694-697. doi: 10.1109/EMBC.2018.8512443.
The age of a subject can be estimated from the brain MR image by evaluating morphological changes in healthy aging. We consider using two-types of local features to estimate the age from T1-weighted images: handcrafted and automatically extracted features in this paper. The handcrafted brain local features are defined by volumes of brain tissues parcellated into 90 or 1,024 local regions defined by the automated anatomical labeling atlas. The automatically extracted features are obtained by using the convolutional neural network (CNN). This paper explores the difference between the handcrafted features and the automatically extracted features. Through a set of experiments using 1,099 T1-weighted images from a Japanese MR image database, we demonstrate the effectiveness of the proposed methods, analyze the effectiveness of each local region for age estimation and discuss its medical implication.
通过评估健康衰老过程中的形态变化，可以从脑部磁共振图像中估计受试者的年龄。在本文中，我们考虑使用两种类型的局部特征从T1加权图像估计年龄：手工制作的特征和自动提取的特征。手工制作的脑部局部特征由脑组织结构的体积定义，这些组织被分割成由自动解剖标记图谱定义的90个或1024个局部区域。自动提取的特征是通过使用卷积神经网络（CNN）获得的。本文探讨了手工制作的特征与自动提取的特征之间的差异。通过使用来自日本磁共振图像数据库的1099张T1加权图像进行的一组实验，我们证明了所提出方法的有效性，分析了每个局部区域对年龄估计的有效性，并讨论了其医学意义。
