Research On Dynamic MR Image Reconstruction Problem Based On Weighted Nuclear Norm And Total Variation

With the rapid development of computer technology,dynamic MR image reconstruction algorithms are playing an increasingly important role in the diagnosis and treatment of clinical lesions.However,in practice,dynamic MR scanning is a time-consuming process due to the limitations of the MRI scanner hardware and human body conditions.Therefore,this thesis investigates how to improve the quality and efficiency of dynamic MRI image reconstruction,mainly as follows:In order to improve the image reconstruction quality,a new model based on weighted nuclear norm and total variation dynamic MRI image reconstruction is proposed.In the proposed model,the image is divided into a low-rank component and a sparse component,where the low-rank component uses a weighted nuclear norm constraint instead of the traditional nuclear norm constraint,and different weights are assigned to different singular values,thus making the design of soft thresholds more reasonable.The sparse component uses the total variation regularization constraint to further improve the reconstruction quality of dynamic MR images.In terms of numerical solution,the model proposed in this thesis contains a non-smooth regular term,which makes it difficult to solve directly.The model is difficult to solve directly.Therefore,this thesis adopts the primal dual method to solve the model.Specifically,the Legendre-Fenchel transformation is used to transform the proposed model.In particular,the model is transformed into a more tractable eigenpair problem using the Legendre-Fenchel transformation,and then multiple subproblems are solved.In numerical experiments,the proposed model is applied to the Pincat image data set,an in vivo myocardial perfusion dynamic MR image data set and an in vivo breast dynamic contrast-enhanced MRI data set,and compared with six classical reconstruction models.The experimental results showed that this model can effectively improve the reconstruction quality of dynamic MR images and better preserve the detail information and edge features of the images.In terms of quantitative evaluation results and subjective visual effects,our method achieves high reconstruction metrics and visual advantages.In particular,in terms of reconstruction time,the proposed model significantly reduces the reconstruction time and improves the reconstruction speed while maintaining the quality of the reconstructed images.