A Study On Registration Of Diffusion Tensor Images

In recent years,modern medical imaging devices have been developing quickly, and new imaging technique has kept coming forth.A new Magnetic Resonance imaging(MRI) method called diffusion tensor imaging(DTI) has emerged,which utilizes the diffusion tensor of each voxel calculated from multiple direction diffusion weighted image(DWI)(at least 6 directions).As a new technology which can reflect the direction of molecule diffusion,DTI can show the information of the structure of tissues and the exchanges of water molecule with each tissue in pathology.It has great advantages on showing the distribution of white matter fiber pathway and its three-dimensional structures because of the distinct anisotropy of the diffusion in water molecule of the brain white matter.Without any other methods can measure the character of the live white matter presently,DTI has great significance for the study of the anatomization of the brain and the diagnosis of the diseases of white matter.As a non-invasive and very important tool for study of the internal structure of the brain,DTI has been widely used in the research of neurophysiology,neurosurgery,and diagnosis of brain cancer, etc.All those function are fully dependent on accurate image registration. Given two image sets acquired from different patient or the same patient but at different times or with different devices,image registration is the process of finding the geometric transformation that aligns one image to another.As is well known, image registration is a sophisticated image processing technology and algorithms about image registration are also getting more perfect.A lot of work has been done about registration of traditional 2-D and 3-D images,whose values are scalars magnitudes.However,only few papers have been yet published about tensor registration(registration of diffusion tensor images is a problem that has received much less attention.).Unlike conventional images,DTI not only measures the intensity at each voxel,but also the orientation.The main difference between scalar and tensor registration is that in the latter case,not only a spatial transformation must be computed,but also another one must be applied to tensor values in every image voxel.This is the main reason why those scalar registration algorithms can not be directly applied to tensor images.This paper introduces the basic theory and concept of DTI and its anisotropy measures.It reviews the registration categories.The principal and methods of tensor reorientation is described in detail.It also analyzes the similarities and optimizations of DT images.Registration of diffusion tensor images is a totally new field,and we can not apply methods of scalar registration directly.Our works are mainly described as follows:1,Acquisition of DICOM images,calculating diffusion tensor images and their anisotropy images,such as FA images and Trace images.2,A fast registration method is proposed,which is circumvents tensor reorientation by registering scalar images derived from DT images,thus discarding the orientation component of the data.Although every pixel of diffusion tensor images is a Second-order tensor,their anisotropy images are scalar.Therefore transformation and reorientation of diffusion tensor images can be achieved by registration of those anisotropy images.This method is simple,rapid and easy to achieve,but fails to take full advantage of the abundant direction information and parts of which is lost.It is suitable for registration application which requires for higher computing speed and lower level of accuracy.3,A novel registration method which performs tensor reorientation by incorporating tensor reorientation in its cost function iteratively is proposed.Taking full account of the characteristic and physical meaning of tensors,information about tensor orientation is merged into objection function and then tensor registration can be achieved simultaneously with tensor reorientation.The final optimization seeks for the global hybrid optimization method,avoiding local extrema successfully.Results shows that this method is a stable one and can keep the principle direction of tensor consistent with fiber orientation effectively.It's really a practical method for tensor registration.