A Study on efficient algorithms for automatic three-dimensional brain image analysis

我们的研究项目的目标是找到有效的医学图像处理算法的实际使用的图像诊断和神经临床研究。在前期的研究中，我们开发了一个基于PC机的脑图像数据库系统，并将其应用于神经临床研究，以定位脑桥锥体神经纤维的空间分布。在这项研究中，我们通过添加来自新安装的具有3.0特斯拉高磁场的MRI设备的直接图像数据集来改进系统，并实现了一种基于神经网络的算法来分割大脑中的小解剖结构。每个患者数据由图像数据集和数字信息组成。通过局域网连接到成像设备，可以在线获取直接的图像数据集，并介绍了MR-MR、MR-PET、PET-PET和CT-PET的三维配准新算法。基于人工神经网络实现了脑表面检测和脑沟回提取的有效算法，实现了精确的脑图像重建。所有这些步骤都是分割程序可以自动进行的。我们的下一个目标是自动分割小脑、海马和胼胝体。

The objective of our research project was to find efficient medical image processing algorithms for the practical use of image diagnoses and neuroclinical researches. In the previous study, we developed a PC-based brain image database system, and applied it to neuro-clinical research for localizing the spatial distribution of pyramidal nerve fibers in the pons. In this research, we improved the system by adding the direct image datasets from the newly installed MRI equipment with high magnetic field of 3.0 Tesla, and implementing a neural-network-based algorithm for segmentation of small anatomical structures in th brain. Each patient data consisted of image datasets and numerical information. The direct image datasets could be acquired on line with connection to the imagine equipments via local area network.Novel 3-D algorithms for MR-MR, MR-PET, PET-PET, and CT-PET registration were introduced. Efficient algorithms for detecting brain surfaces and extracting sulci and gyri based on artificial neural network were implemented to reconstruct precise brain images. All these steps were of segmentation procedures could be automatically carried out. Our next goal is to automatically segment the cerebellum, the hippocampus, and the corpus callosum.