Investigating and Implementing Tractography Clustering on Diffusion-Weighted MR Images of the Brain by Using QuickBundles Method with Different Features


Milad Hospital <>



One of the common algorithms nowadays for the segmentation of Diffusion Tensor images is the clustering method, one of the techniques of which is the QuickBundles method. Its biggest capabilities are the analysis of images using different features and metrics so that many of these features and metrics can be introduced and distributed as desired. Of course, their implementation will require the use of open-source programming languages. In this study, some of the most important features of the QuickBundles method have been implemented and compared with each other using Python.
Materials and Methods:
Necessary information for Tractography was collected from 10 completely healthy patients according to the standard conditions for all patients. After data preprocessing, Python libraries were used to perform Tractography analysis. Four features namely center of mass, middle point, arc length, and end points between the directions vectors were analyzed by the QuickBundles method on all patients. Then, anatomically and clinically, all calculated streamlines were compared with each other.
The streamlines obtained in terms of computing time are variable according to the number of streamlines so that the implementation of the center of mass feature has the least computing time. In all four features, the route map in color has been appropriate in terms of anatomical comparison, and the radiologist has reported it as acceptable.
One of the most important features of the implementation of features in DTI data analysis is the appropriate speed of the QuickBundles method, especially when implemented in environments such as Python. It seems that the implementation of each feature can well express the expected results from the implementation of data analysis, such that the feature of the center of mass and the middle point to show the spatial position of the streamlines, the length of the arc to show the length of a streamline and the endpoints between the vectors to the amount of rotation of a streamline. They are very suitable for examining the mentioned features.