Automatic extraction of the cingulum bundle in diffusion tensor tract-specific analysis: feasibility study in Parkinson's disease with and without dementia.

PURPOSE

Tract-specific analysis (TSA) measures diffusion parameters along a specific fiber that has been extracted by fiber tracking using manual regions of interest (ROIs), but TSA is limited by its requirement for manual operation, poor reproducibility, and high time consumption. We aimed to develop a fully automated extraction method for the cingulum bundle (CB) and to apply the method to TSA in neurobehavioral disorders such as Parkinson's disease (PD).

MATERIALS AND METHODS

We introduce the voxel classification (VC) and auto diffusion tensor fiber-tracking (AFT) methods of extraction. The VC method directly extracts the CB, skipping the fiber-tracking step, whereas the AFT method uses fiber tracking from automatically selected ROIs. We compared the results of VC and AFT to those obtained by manual diffusion tensor fiber tracking (MFT) performed by 3 operators. We quantified the Jaccard similarity index among the 3 methods in data from 20 subjects (10 normal controls [NC] and 10 patients with Parkinson's disease dementia [PDD]). We used all 3 extraction methods (VC, AFT, and MFT) to calculate the fractional anisotropy (FA) values of the anterior and posterior CB for 15 NC subjects, 15 with PD, and 15 with PDD.

RESULTS

The Jaccard index between results of AFT and MFT, 0.72, was similar to the inter-operator Jaccard index of MFT. However, the Jaccard indices between VC and MFT and between VC and AFT were lower. Consequently, the VC method classified among 3 different groups (NC, PD, and PDD), whereas the others classified only 2 different groups (NC, PD or PDD).

CONCLUSION

For TSA in Parkinson's disease, the VC method can be more useful than the AFT and MFT methods for extracting the CB. In addition, the results of patient data analysis suggest that a reduction of FA in the posterior CB may represent a useful biological index for monitoring PD and PDD.