Study And Application Of Non-rigid Registration Algorithms In Diffusion-weighted Imaging Sequence Of The Prostate

Prostate cancer(PCa)is one of the most common tumors in the male genitourinary system.The incidence and mortality of PCa are third in the western countries.With the popularization of specific antigen detection,the increasing westernization of diet and living habits in China,the incidence of PCa has been significantly increased in the last thirty years.The prognosis and treatment of PCa mainly depend on early diagnosis and preoperative staging.Early diagnosis and early treatment can effectively reduce the mortality of PCa.Diffusion weighted image sequence(DWI)and apparent diffusion coefficient(ADC)map have the advantage of non-invasive detection of the characteristics of water dispersion limitation in living tissues,and the advantages of high sensitivity and specificity.Therefore,it is regarded as one of the best diagnostic methods for PCa.In the acquiring of DWI sequence,prostate movement caused by body movement and other factors in the same slice will cause the change of position and shape.For the DWI sequences that are misalignment due to movement,the clinical MRI scanner workstation usually does not do post-processing,but estimates it directly and calculates the ADC value for PCa diagnosis.But this misalignment results in overestimation of ADC values or error values.The overestimation of ADC value would underestimate the degree of tissue water disffusion limitation and influence the PCa staging,and the error value would cause the ADC map to be blurred and affect the location of the PCa.As a post-processing technology,image registration can improve the diagnostic efficiency of DWI and ADC by effectively solving this problem.It has important application value.In this paper,the research of non-rigid registration of prostate DWI sequence and its application in PCa quantitative diagnosis have laid the foundation for PCa automatic diagnosis system based on MRI.The summary of this paper as follow:(1)Affine registration algorithm based on three restriction criterions for feature point pair matching.In order to eliminate the drift caused by movement or breathing quickly and accurately,an affine registration algorithm based on feature point pair matching which determine image transformation relationship is proposed.In the light of three restriction criterions,local information entropy,similarity measure and criterion of constant distance scale,match the feature points of fixed and floating images.Then,using these matched feature points,affine translation parameters of two images were computed by matrix least square method.The experimental results show that the algorithm has a higher matching rate and registration accuracy.(2)B spline free-form deformation(FFD)registration algorithm based on injective constrainted.In order to correct distortion and deformation of soft tissue which be extruded by other organs during DWI acquiring,and meet the requirements of clinical application,which require restricted diffusion texture features generated by the suspected lesion tissue should be retained,proposed a FFD algorithm based on multi resolution constraints.By adding the injective constraint in the cost function,the algorithm limits the influence range of the deformation field.The experimental results show that the algorithm has the characteristic of maintaining the diffusion texture features,and there is no mass artifact.(3)An automatic hybrid registration algorithm is proposed to solve the problem of movement and distortion in the DWI sequence of the prostate.This algorithm includes two parts,the global affine registration and the local FFD registration,which can correct the image distortion of different scales.The experimental results show that the method can effectively correct the movement and distortion of the prostate.The algorithm has high robustness,which can keep the details of the dispersion texture of the water molecules in the prostate cancerous tissue and avoid the undesired deformation.(4)The diagnostic value of ADC in the differential diagnosis of benign and malignant tumors in each anatomic partition of the prostate was analyzed.The impact of image registration on the effectiveness of the DWI diagnosis for PCa is evaluated.By calculating and analyzing the ADC values of normal tissues and cancerous tissues in 19 cases of PCa samples,with the pathological results cross referencing,we concluded that ADC value can be used for PCa identification,location and staging.The hybrid registration method can get higher reliability ADC value,improve the visibility of PCa in DWI and ADC diagrams,further clarify the relationship between ADC value and Gleason classification,and improve DWI and ADC diagrams to identify the sensitivity and specificity of PCa and benign lesions.