A Bayesian approach to fiber orientation estimation guided by volumetric tract segmentation. (December 2016)

• Record Type: Journal Article
• Authors: Ye, Chuyang; Prince, Jerry L.
• Is Part Of: Computerized medical imaging and graphics. Volume 54(2016)
• Page Start: 35
• Page End: 47
• Publication Date: 2016-12
• Subjects: DTI -- Fiber orientation estimation -- Volumetric tract segmentation; Diagnostic imaging -- Periodicals; Imaging systems in medicine -- Periodicals; Diagnosis, Radioscopic -- Data processing -- Periodicals; Diagnostic Imaging -- Periodicals; Imagerie pour le diagnostic -- Périodiques; Diagnostic imaging; Periodicals; Electronic journals; Electronic journals; 616.0754
• DOI: 10.1016/j.compmedimag.2016.09.003
• Languages: English
• ISSNs: 0895-6111

Abstract

Abstract : Highlights: We use anatomical knowledge from tract segmentation to aid fiber tracking. The fiber orientation (FO) estimation is formulated in an MAP framework. FOs are estimated by iteratively solving Euler–Lagrange equations. The method can reduce FO estimation errors and resolve crossing fibers. The method was applied for a connectivity study to show its scientific benefit. Abstract: Diffusion magnetic resonance imaging (dMRI) provides information about the microstructure of white matter in the human brain. From dMRI, streamlining tractography is often used to reconstruct computational representations of white matter tracts from which differences in structural connectivity can be explored. In the fiber tracking process, anatomical information can help reduce tracking errors caused by crossing fibers and image noise. In this paper, we propose a Bayesian method for estimating fiber orientations (FOs) guided by anatomical tract information using diffusion tensor imaging (DTI), which is a standard clinical and research dMRI protocol. The proposed method is named Fiber Orientation Reconstruction guided by Tract Segmentation (FORTS). A first step segments and labels the white matter tracts volumetrically, including explicit representations of crossing regions. A second step estimates the FOs using the diffusion information and the anatomical knowledge from segmented white matter tracts. A single FO is estimated in the noncrossing regions while two FOs are estimated in the crossing regions. A third step carries out streamlining tractography that uses information from both the segmented tracts and the estimated FOs. Experiments performed on a digital crossing phantom, a physical phantom, and brain DTI of 18 healthy subjects show that FORTS is able to use the anatomical information to produce FOs with better accuracy and to reduce anatomically incorrect streamlines. In particular, on the brain DTI data, we studied the connectivity of anatomically defined tracts to cortical areas, which is not straightforwardly achievable using only volumetric tract segmentation. These connectivity results demonstrate the potential application of FORTS to scientific studies.