Modality-based Image Registration Using Multiple Classifiers in Image-guided Medical Diagnosis Model

概要

論文の詳細を見る
In clinical applications, medical images are frequently fused together to improve the diagnostic accuracy, treatment planning and aid in surgical procedure. Image registration of medical images from different imaging modality according to diagnostic prescription has played an important role as a diagnostic-assisted tool in supporting diagnosis findings and their accuracy. This paper proposed a novel method for the modality-based medical image registration using multiple classifiers. In our approach, pre-processing is first employed to identify and preview input images by their modality and usage. Three classifiers are then chosen from the current literature : motion estimation, chamfer matching and mutual information. Each classifier is dynamically chosen to improve overall image registration results. The motion estimation (ME) classifier is used to comprehend and recover motion-blurred input images, while the chamfer matching (CM) classifier is used for calculating and improving registration alignment. Next, the mutual information (MI) classifier is selected to improve registration robustness for greater clarity and visibility. Experimental results reveal that using our approach and the combination of the three classifiers yields a robust and accurate registration with respect to motion-blurred and misaligned images for intersubject and intrasubject study.
社団法人日本生体医工学会の論文
2003-12-10