医学X线图像配准技术研究进展及其存在的问题（英文）

Progress and Existing Problems of Registration Technique for Medical X-ray Image

医学X线图像受到探测器面积大小的限制,成像范围有限,对较大器官的扫描无法一次完成。在观察病变部位时,医生需要结合多幅图像来进行诊断或治疗,因此需要对多张影像进行拼接处理。作为图像拼接技术的核心,图像配准技术已被广泛应用于医学成像中,将那些从扫描中获得的多类型信息进行配准从而得到更详细的信息。首先,本文重点综述了目前面向X线图像的比较主流和新兴的配准技术,如基于互信息的配准法,基于特征的配准法和基于变换域的配准法。其次,指出了X线图像配准中存在的影像漂移问题、拍摄角度的限制、非刚性配准仍未成熟、没有绝对的配准评价标准等问题。最后,总结了基于FPGA等硬件的医学图像配准、采用超分辨率重建技术以获取更高质量的待配准图像从而提高图像配准的精度和速度等发展趋势与研究前景。

With the restriction of the size of detectors,the imaging range is limited and it is impossible to scan a large organ completely at a time. The doctors need to combine multiple images for a diagnosis or a treatment when observing the lesion area. Therefore the image mosaic technique is necessary. As the core of image mosaic technique,the image registration techniques have been found extensive use in medical imaging where information from multiple types of scans is registered to obtain information in greater detail. Firstly,the techniques of X-ray image registration are reviewed in this paper,and this paper pays an emphasis on the dominating as well as emerging techniques currently,such as the mutual information-based methods,feature-based methods,and the transform domain-based methods. Secondly,the problems existing in the X-ray image registration are pointed out,such as image shift problems,restriction of shooting angle,immaturity in non-rigid registration technique,no absolute standard for registration evaluation. Lastly,the research prospects are summarized,such as the medical image registration based on FPGA hardware,obtaining higher quality images for registration to achieve greater registration accuracy and speed by using super-resolution reconstruction technique,etc.