How to analyze diffusion MRI data

Introduction

     This short tutorial is a summary of the review paper I coauthored [1]. Started from the simplest approach to more recent connectome analysis, I am listing a step-by-step workflow in the following sections. I would also recommend checking out the citation page to see how other studies have used DSI Studio to do their analysis.

Region-based analysis


     The simplest way to analyze diffusion data is to assign a region in the brain (either manual drawing or from an atlas) and get statistics on its diffusion indices. 

     The following are the steps for ROI-based analysis in the subject space.


     2. SRC file quality control to exclude problematic data.

     3. Reconstruct each SRC file using a reconstruction method (GQI or QSDR) to get a FIB file. 

         I would recommend using QSDR because it will reconstruct data in the standard space and make further analysis much more efficient. Both GQI and QSDR reconstruction will also output additional DTI metrics such as FA, AR,RD. 

     4. Load the fib file. There are two ways to assign a region. You can manually draw a regionload a region from a file, or load the region from an atlas

         Loading region from an atlas will be the most efficient way.         

     5. [Optional]  If you would like to analysis other measures (e.g. DKI, NODDI, or PET), you would need to have those measures saved in the NIFTI format, and insert them using [Slices][Insert T1W/T2W images]. DSI Studio will register newly added image to the diffusion data set. The added metrics will be analyzed in the followings steps.

     6. Once you assign a region or a set of regions, you can get the average value of the diffusion indices (e.g., FA) and newly added measures from the main menu [Region][Statistics]. Click on "show details" button to see the results. You may load the results in Excel to do further analysis. Each subject should have a diffusion index value calculated from a region.

     7. On other tool or software (Excel) calculate the correlation of the indices with the study variables.
     
     Another way to do region-based analysis is to reconstruct/normalize all data in the MNI space and extract region statistics from all subjects in one shot. Here are the steps

     1. Construct a connectometry database from all subjects. For each "Index of interest" in Step C2, you would need to create one database (e.g. SDF is an analogue of QA in the MNI space, dti_fa is DTI FA, ...etc)
    
     2. Open the database file in Step T3, and follow the above mentioned steps 3 to 6. Please note that this db.fib.gz file is already in the MNI space, and the regions should be MNI regions.

Track-specific analysis


     Example studies: https://www.nature.com/articles/nn.3870

     Tractography based analysis uses diffusion fiber tracking to define the trajectories of fiber pathways, and the diffusion indices along the pathways can be averaged and analyzed. The advantage of tractography based analysis is that it is fiber specific, whereas region based analysis does not specify which fiber pathways to be analyzed. The following list is the steps for tractography-based analysis.


     2. SRC file quality control to exclude problematic data.

     3. Reconstruct each SRC file using a reconstruction method to get a FIB file. You may use DTI reconstruction, but it is not able to resolve fiber crossing. Other options include GQI reconstruction or QSDR (reconstructs data in the MNI space).

     4. Load the fib file and do fiber tracking You may need to use a combination of SEED/ROI/ROA to obtain the exact fiber pathways you are interested in. You may also need to use track editing to get the exact segments or branches of the fiber pathways.

     5. [Optional]  If you would like to analysis other measures (e.g. DKI, NODDI, or PET), you would need to have those measures saved in the NIFTI format, and insert them using [Slices][Insert T1W/T2W images]. DSI Studio will register newly added image to the diffusion data set. The added metrics will be analyzed in the followings steps.
     
     6. Use main menu [Tracts][Statistics] to get the average diffusion indices along the fiber pathways. Click on "show details" to get information and copy it to Excel for further analysis.

     Note: the anisotropy threshold is used as a mask to filter out background voxels, and its value will determine how DSI Studio samples along track anisotropy. To avoid this inconsistency issue, you may need to keep a consistent anisotropy threshold as you save the along track anisotropy.

     7. Alternatively, you can use track profile to get the distributed pattern of a diffusion index along the fiber pathways. 


      Another way to do region-based analysis is to reconstruct/normalize all data in the MNI space and extract tract statistics from all subjects in one shot. Here are the steps

     1. Construct a connectometry database from all subjects. For each "Index of interest" in Step C2, you would need to create one database (e.g. SDF is an analogue of QA in the MNI space, dti_fa is DTI FA, ...etc)
    
     2. Open the database file in Step T3, and follow the above mentioned steps 3 to 5. Please note that this db.fib.gz file is already in the MNI space, and the regions should be MNI regions.


Track Profile



     2. SRC file quality control to exclude problematic data.

     3. Reconstruct each SRC file using a reconstruction method to get a FIB file. You may use DTI reconstruction, but it is not able to resolve fiber crossing. Other options include GQI reconstruction or QSDR (reconstructs data in the MNI space).

     4. Load the fib file and do fiber tracking You may need to use a combination of SEED/ROI/ROA to obtain the exact fiber pathways you are interested in. You may also need to use track editing to get the exact segments or branches of the fiber pathways.

     5. Use main menu [Tracts][Tract Profile] to get the the track profile data 


Shape analysis






























     2. SRC file quality control to exclude problematic data.

     3. Run automatic fiber tracking and shape analysis using [Step B4: Automatic Fiber Tracking] to output track statistics

Connectivity matrix and graph theoretical analysis



     Connectivity matrix uses an atlas for partitioning the cortex into several regions and use a number of connecting tracks as the matrix entry. A connectivity matrix can be generated for each subject, and the matrices of a group of subjects can be compared with those from another group of subjects to study the differences.

     The following list is the steps for generating a connectivity matrix.


     2. SRC file quality control to exclude problematic data.

     3. Reconstruct each SRC file using QSDR reconstruction. 

     
     

Differential tractography





     


























     Differential tractography is a new type of tractography that compares repeat scans of the same individuals to capture neuronal injury reflected by a decrease of anisotropy.  The following is the steps 

  
     2. SRC file quality control to exclude problematic data.

     3. Follow the steps at Differential tractography for longitudinal data

     Individual connectometry aims to find the difference between the longitudinal study of a single subject. The method tracks the deviate pathways of an individual by comparing the subject with a normal population, an atlas, or the subject's previous scan. It is a powerful tool to map the tracks that are damaged. The steps to get the group or individual connectometry is described in the following documentation:


Correlational Tractography (Group Connectometry)

     
Example studies: (1)(2)

     Correlational tractography is a tractography modality that shows pathways correlated with a study variable (e.g. age). The method to derive correlational tractography and test its reliability is called "group connectometry." There are two types of connectometry analyses are available in DSI Studio: group connectometry and individual connectometry. 

     Group connectometry aims to find trajectories associated with a study variable in a group study. The method allows you to identify the exact segment/subcomponent/branches of tracks that are correlated to group difference or any study variable (age, sex, ...etc).

     The following are the steps for doing connectometry analysis.

     For group connectometry see:

     1. See Create a connectometry database to create a connectometry database



Reference


[1] Abhinav, K., et al., Advanced diffusion MRI fiber tracking in neurosurgical and neurodegenerative disorders and neuroanatomical studies: A review. Biochim Biophys Acta, 2014. 1842(11): p. 2286-2297.
[2] Yeh, F.C., et al., Deterministic diffusion fiber tracking improved by quantitative anisotropy. PLoS ONE, 2013. 8(11): p. e80713.
Comments