Non-Rigid Registration Using A Rigidity Constraint Based On The Grey Information

With the rapid progress and development of high technologies,the medical imaging technologies have provided many imaging equipments,which are more and more assisting the modern medicine.The computer tomography(CT) and magmatic resonance(MR) have achieved high performance in medical imaging.At the same time,the functional magmatic resonance imaging(fMRI),positron emission tomography(PET),single photon emission CT(SPECT),etc.,these functional,metabolic imaging technologies are employed more and more in medicine.These imaging techniques are providing high-quality anatomy and function information for clinical diagnose and other computer-assistant operation and treatment.They also help people realize the importance of health,which further improve the medical treatment and heath of people.Howerver,non-rigid registration,which is widely used in the field of medical imaging,usually models all tissue as non-rigid objects and does not take the rigidity of different tissue types into account.If the rigid objects like bones are present in the registration,these algorithms will not necessarily preserve the rigidity of these objects.Preservation of the rigidity is not only useful for tissue that is rigidity by nature, it can also be very useful in other cases.If we register the two images that have the structures changed in size,it will be useful to retain the difference of the structures such as the tumor growth between follow-up images.Traditional non-rigid registration matches two images to the same size,and no difference will be observed.Tumor growth will be concealed.Keeping the tumor objects rigid can prevent this.In this paper,we proposed a novel method to keep required objects rigid while performing registration.This is achieved by modifying the Free-form Deformations(FFD) based on B-Spline to get a rigid constraint.After that,the method was employed in the vertebra registration.We use the properties of the vertebra and employ this method in the vertebra registration.The main works are as described as below:1.Different medical image non-rigid registration with rigid constraint algorithms are discussed and compared.The non-rigid registration with rigid constraint can be classified into two categories:the metric modification based method and the transform modification based method.2.After the discussion,the novel non-rigid registration with rigid constraint was proposed.The B-spline control points can be divided into rigid controls and free-form deformation controls.All the rigid control points can be transformed together and the displacement of these controls points can be got from the rigid transformation.If the control points set was extended 2 control points simply,then the area of rigid parts would expand too much,the non-uniform method was also employed into this method.3.the vertebra properties and the non-rigid registration with rigid constraint were used to process the multi-stage vertebra registration.the vertebra bone centers were got from the vertebra pre-segmentation.And the approximation two times curves was use to simulate the center lines of the bones,which was used to set the initial values of the non-rigid soft tissue.And the rigid registration of the every bone use to set the initial values of the rigid control points.After that,the non-rigid registration with rigid constraint was used the got the last registration result.