作者: Mehmet Z Unlu , Andrzej Król , Alphonso Magri , James A Mandel , Wei Lee
DOI: 10.1016/J.COMPBIOMED.2009.10.010
关键词: Computer vision 、 Artificial intelligence 、 Initialization 、 Mesh generation 、 Finite element method 、 Image resolution 、 Fiducial marker 、 Similarity (geometry) 、 Image registration 、 Medicine 、 Displacement (vector)
摘要: Abstract We have developed and tested a new simple computerized finite element method (FEM) approach to MR-to-PET nonrigid breast-image registration. The requires five–nine fiducial skin markers (FSMs) visible in MRI PET that need be located the same spots on breast two flanks during both scans. Patients similarly positioned prone This is accomplished by means of low gamma-ray attenuation coil replica used as support scan. demonstrate that, under such conditions, observed FSM displacement vectors between MR images, distributed piecewise linearly over volume, produce deformed FEM mesh reasonably approximates deformation tissue method, which does not require biomechanical model, robust fast. Contrary other approaches utilizing voxel intensity-based similarity measures or surface matching, our works for matching with pure molecular images (i.e. SPECT only). Our good initialization would trapped local minima registration process. All processing including FSMs detection generation can fully automated. acquired 15 subjects. procedure yielded quality an average target error below 4 mm well spatial resolution 6–7 mm). Based results obtained subjects studied date, we conclude this very fast well-performing Therefore, it promising clinical practice. easily applied CT any type soft-tissue their counterparts using SPECT.