Sparse-Prior-Based Projection Distance Optimization Method for Joint CT-MRI Reconstruction

Simple item page
dc.contributor.authorCui, Xuelinen
dc.contributor.authorMili, Lamine M.en
dc.contributor.authorYu, Hengyongen
dc.contributor.departmentElectrical and Computer Engineeringen
dc.date.accessioned2019-05-15T14:57:48Zen
dc.date.available2019-05-15T14:57:48Zen
dc.date.issued2017en
dc.description.abstractMultimodal imaging techniques have received a great deal of attention, since their inceptions for achieving an enhanced imaging performance. In this paper, a novel joint reconstruction framework for computed tomography (CT) and magnetic resonance imaging (MRI) is implemented and evaluated. The CT and MRI data sets are synchronously acquired and registered from a hybrid CT-MRI platform. Because the image data sets are highly undersampled, the conventional methods (e.g., analytic reconstructions) are unable to generate decent results. To overcome this drawback, we employ the compressed sensing (CS) sparse priors from an application of discrete gradient transform. On the other hand, to utilize multimodal imaging information, the concept of projection distance is introduced to penalize the large divergence between images from different modalities. During the optimization process, CT and MRI images are alternately updated using the latest information from current iteration. The method exploits the structural similarities between the CT and MRI images to achieve better reconstruction quality. The entire framework is accelerated via the parallel processing techniques implemented on a nVidia M5000M Graph Processing Unit. This results in a significant decrease of the computational time (from hours to minutes). The performance of the proposed approach is demonstrated on a pair of undersampled projections CT and MRI body images. For comparison, the CT and MRI images are also reconstructed by an analytic method, and iterative methods with no exploration of structural similarity, known as independent reconstructions. Results show that the proposed joint reconstruction provides a better image quality than both analytic methods and independent reconstruction by revealing the main features of the true images. It is concluded that the structural similarities and correlations residing in images from different modalities are useful to mutually promote the quality of joint image reconstruction.en
dc.description.notesThe work of H. Yu was supported by the NSF CAREER Award CBET under Grant 1540898.en
dc.description.sponsorshipNSF CAREER Award CBET [1540898]en
dc.format.mimetypeapplication/pdfen
dc.identifier.doihttps://doi.org/10.1109/ACCESS.2017.2754327en
dc.identifier.eissn2169-3536en
dc.identifier.urihttp://hdl.handle.net/10919/89527en
dc.identifier.volume5en
dc.language.isoenen
dc.publisherIEEEen
dc.rightsCreative Commons Attribution 4.0 Internationalen
dc.rights.urihttp://creativecommons.org/licenses/by/4.0/en
dc.subjectMultimodal imagingen
dc.subjectimage reconstructionen
dc.subjectcompressed sensingen
dc.titleSparse-Prior-Based Projection Distance Optimization Method for Joint CT-MRI Reconstructionen
dc.title.serialIEEE Accessen
dc.typeArticle - Refereeden
dc.type.dcmitypeTexten

Files

Original bundle
Now showing 1 - 1 of 1
Thumbnail Image
Name:
08046006.pdf
Size:
11.01 MB
Format:
Adobe Portable Document Format
Description:
Download

Collections

Scholarly Works, Electrical and Computer Engineering