Diffusion Spectrum Imaging (DSI)
DSI is one of the model-free techniques of diffusion-weighted imaging which samples the q-space signals at the Cartesian grid points. By performing Fourier transform on the q-space signals, the ensemble diffusion propagators can be estimated on a voxel-by-voxel basis. Therefore, this technique can resolve the directions of the intra-voxel crossing fiber tracts, and then allows more accurate estimation of axonal trajectories.

According to the directional information provided by diffusion-weighted imaging, tractogrphy is a technique used to reconstruct the trajectories of axonal tracts in the white matter. Our lab developed the software, DSI Studio, to present the tracts, which can greatly facilitate the visualization of the 3D trajectories. In addition, DSI Studio implemented the common reconstruction algorithms and the popular diffusional indexes.

Half-sphere sampling scheme

The sampling scheme is a rapid DSI technique. Under the assumption of antipodal symmetry about the origin of q-space, only half of the grid points are necessary to be acquired for a DSI acquisition. Currently, 102 q-space grid points located in the same hemisphere with a maximum b-value of 4000 s/mm2  is routinely used, resulting in an acquisition time of about 16 minutes.


LDDMM-DSI  is a registration method for DSI datasets under the LDDMM framework. This method made use of the fact that a DSI dataset is 6D, and generalized the original 2D/3D LDDMM algorithm to the 6D case with some modifications made for the DSI datasets. In this manner, the conventional reorientation problem that arises from transforming diffusion-weighted datasets was avoided by making the DSI datasets capable of being freely deformed in the q-space. The algorithm treated the data-matching task as a variational problem and sought optimal velocity fields from which the generated transformations were diffeomorphic and the transformation curve was a geodesic.


NTU-DSI-122 is a DSI template constructed in the standard ICBM-152 space from 122 healthy adults. This template was built through incorporating the macroscopic anatomical information using high-resolution T1-weighted images and the microscopic structural information obtained from DSI datasets, rendering it to achieve a high anatomical matching to the ICBM-152 space. Therefore, this template can serve as a representative DSI dataset for a healthy adult population, and will be of potential value for brain research and clinical applications. This template is released in the original DWI format, so the users have the most freedom to perform their own advanced processing algorithms on NTU-DSI-122. The template can be downloaded here.


We create the most detailed map of the white matter tract bundles based
on the high-quality NTU-DSI-122 template. White matter tract bundles in the human brain were reconstructed using the multiple-ROI methods combined the stream-line algorithm. The 76 white matter tract bundles are categorized into association, projection, and commissural tract systems according to anatomical definitions. Coordinates of each tract bundles are defined on the Montreal Neuroscience Institute (MNI) space. The DSI atlas can be used for public education to understand the anatomy of the white matter. Moreover, it can be applied to implement a template-based approach that enables a high-throughput automated analysis (TBAA) of the microstructural integrity of the tracts on the atlas.
(To link the online tractatlas, please click here. Please install Unity Web Player first ( and use IE browserto view the tractatlas)

Tract-based automatic analysis (TBAA)

Tract-based automatic analysis (TBAA) is an automatic and high-throughput technique for analyzing the information of white matter tracts. Via integrating several MRI techniques established in ABMRI lab, TBAA converts complex analyses for white matter tracts into a one-button calculation. The concept of TBAA is to build up a study specific template (SST) from numerous imaging datasets. The SST was registered with an established DSI template, NTU-DSI-122, and the pre-defined 76 white matter tracts on the NTU-DSI-122 were than transformed into the SST resulting 76 transformed tracts. The 76 transformed tracts were than transformed into each dataset by applying the deformations between the dataset and SST resulting 76 individual tracts for each subject. After getting the 76 individual tracts, the information of each fiber tract could be extracted and standardized whole brain information, called 2D connectogram, could be obtained for each subject. Not only for DSI datasets, TBAA technique could be also used for other diffusion MR imaging including diffusion tensor imaging, Q-ball imaging etc. This technique could provide a very easy, convenient, and reliable brain analysis for clinical purposes and scientific researches.

We have performed DSI to different clinical applications, such as schizophrenia, autism spectrum disorder, attention deficit hyperactivity disorder, mild cognitive impairment, Alzheimer's disease, temporal lobe epilepsy, normal aging. We aimed to find white matter lesions as biomarkers in these disorders via DSI tractography analysis.

Recently, we developed an imaging-based brain age predictive model with
machine learning approach. It provides a promising way to assess an individual's brain age relative to healthy populations, potentially offering clinically relevant biomarkers of neurodegenerative diseases which often manifest accelerated aging process.