Automatic Extraction and Matching of Neonatal Cerebral Vasculature from MRA-TOF Images
Malamateniou C, Xue H, Hajnal J, Rueckert D, Allsop J, Srinivasan L
Hammersmith Hospital, Imperial College London
An automated method to extract and match neonatal cerebral vasculature from MRA-Time of Flight (MRA-TOF) images is presented. Vessel extraction is first performed to find vessel centerlines from 3D MRA-TOF images and the extracted segments are connected iteratively to compose a vessel tree. To study vasculature development from serial MRA examinations of neonates, account must be taken of rapid brain growth, accompanied by extensive vascular development, which changes vessel geometry so that voxel-by-voxel comparison is unreliable. Accordingly a vessel matching algorithm is developed to determine topological correspondence between different branches. The methods were tested on 4 infants each scanned twice.