面向时间不耐受患者的脑MRI图像超分辨率算法研究

项目名称： 面向时间不耐受患者的脑MRI图像超分辨率算法研究

项目编号： No.61271374

项目类型： 面上项目

立项/批准年度： 2013

项目学科： 无线电电子学、电信技术

项目作者： 李建武

作者单位： 北京理工大学

项目金额： 75万元

中文摘要： 如何对时间不耐受患者进行有效的磁共振成像，是当前理论研究与实际临床中面临的难题。本项目面向脑部磁共振成像，提出对时间不耐受患者进行多次有间歇的快速扫描，得到多个分辨率较低的切片图像序列，分析与设计特定的超分辨率算法，重建出分辨率较高的切片图像序列。 研究内容包括四个方面：(1) 基于同组成像参数下的多个低分辨率切片图像的超分辨率重建；(2) 基于不同组成像参数下的多个低分辨率切片图像的超分辨率重建；(3) 在给定一组高分辨率切片图像序列的情况下，对其它不同组成像参数下的低分辨率切片图像序列进行超分辨率重建；(4) 降低切片层厚，实现切片选择方向的超分辨率。 在算法上拟提出基于自适应核回归的迭代反投影方法，以及特定的自适应正则化超分辨率算法，以充分利用同层切片的不同成像所共有的轮廓、边缘以及纹理等信息。 最终将实现脑部磁共振成像的超分辨率软件原型，为医生对疾病的诊断提供技术支持。

中文关键词： 磁共振成像；超分辨率重建；时间不耐受患者；压缩感知；图像去噪

英文摘要： How to perform effectively magnetic resonance imaging (MRI) for the patients with imaging time constraint is an open and critical problem in both theoretical research and clinical cases. This proposal attempts to solve brain MRI of the patients with imaging time constraint via performing several times of individual rapid MRI scans to obtain some low-resolution slice image sequences, and then designing super resolution algorithms to reconstruct a high-resolution slice image sequence. The research includes four aspects: (1) Super resolution enhancement based on several low-resolution slice images with the same set of imaging parameters; (2) Super resolution enhancement based on several low-resolution slice images with different sets of imaging parameters; (3) Given a low-resolution slice image sequence with a certain set of imaging parameters, performing super resolution for those low-resolution slice image sequences with other different sets of imaging parameters; and (4) Reducing slice thickness and realizing super resolution in the slice-select direction. The proposal will design iterative back projection algorithms based on self-adaptive kernel regression, and specific self-adaptive regularization super resolution algorithms to utilize the same information on sketch, edge, and texture from multiple images of t

英文关键词： magnetic resonance imaging；super resolution reconstruction；the patients with imaging time constraint；compressed sensing；image denoising