Visualizing the Correspondence of Feature Point Mapping between DICOM Images before and after Surgery

Abstract

We extract feature point mapping between preoperative and postoperative Digital Imaging and Communications in Medicine (DICOM) images from magnetic resonance imaging (MRI) or from computer tomography (CT). The aim is to quantitatively investigate brain shift during intraoperative surgery. First, using 124 two-dimensional images constituting DICOM, a large number of 2D feature points are extracted as uniformly as possible inside the brain. Next, we extract one pair from the 124 preoperative images and the 124 postoperative images and construct map correspondences of similar feature points with a range of DICOM gray values. If the Euclidean distance between the two feature points in the 2D images is too large, the pair of feature points is deleted to prevent mis-mapping; brain shifts are usually less than 2-3 cm. Finally, we find image pairs with the highest number of mappings from DICOM images before and after surgery (two-dimensional stacked three-dimensional images), and generate graph representing correspondences between image pairs with the highest number. Finally, we visualize 3D correspondences between DICOM images before and after surgery.