Comprehensive and quantitative study of rank‐4 order diffusion tensor imaging and positive definite rank‐4 order diffusion tensor imaging: A higher order tensor imaging study. Issue 1 (March 2014)

• Record Type: Journal Article
• Authors: Jiang, Shan; Zhang, Pengfei
• Is Part Of: International journal of imaging systems and technology. Volume 24:Issue 1(2014:Mar.)
• Page Start: 83
• Page End: 93
• Publication Date: 2014-03
• Subjects: Imaging systems -- Periodicals; Image processing -- Periodicals; 621.367
• DOI: 10.1002/ima.22082
• Languages: English
• ISSNs: 0899-9457

Abstract

<abstract abstract-type="main"> <title>ABSTRACT</title> <p>Rank‐4 order diffusion tensor imaging (DT4) has been shown to be able to capture complex local tissue structures, for example, crossing fibers, but it cannot ensure the positive‐definite property of diffusion. As a result, the positive‐definite rank‐4 order diffusion tensor imaging was presented to guarantee the positive‐definite property. However, a comprehensive comparison that concerns SNR of Rician noise, effect of the numbers of gradient directions and the ability of detecting IVOH between DT4 and PDT4 has never been performed. To overcome these limitations, a comprehensive comparison between the two methods is made. Results show that it can help to determine whether the applied noise is close to the clinical application by estimating corresponding SNRs. Compared with DT4, PDT4 can achieve better result in terms of fiber error, standard deviation (SD) of generalized trace (GT) and generalized anisotropy (GA). PDT4 can increase the values of GA and scaled entropy, especially in nonwhite matter areas. Thus this study comes to conclusions as follows. The relationship between SNR and SD of Rician noise should be considered. PDT4 can achieve higher accuracy and better stability, but the advantage will become unobvious with increasing of sample directions and SNR. PDT4 will detect fake IVOH in nonwhite matter areas. The works in this article can provide an optimized plan for making sampling scheme and choosing proper imaging methods.</p> </abstract>