2011 Summer Project Week DTIPrep
Diffusion Weighted Imaging and Diffusion Tensor Imaging Quality Control_DTIPrep
3D view of gradients before and after Quality Control procedures
Motivations: Automated Diffusion Weighted Imaging Quality Control
Diffusion Tensor Imaging (DTI) has become an important MRI procedure to investigate the integrity of white matter in brain in vivo. DTI is estimated from a series of acquired Diffusion Weighted Imaging (DWI) volumes. Unfortunately, the DWIs suffer from a lots of artifacts because of the inherent low SNR and long scanning time of multiple directional encoding. Thus, it necessitates the development of Quality Control (QC) as the automated tool for DWIs. We have developed a tool called DTIPrep which is fully open source. In the DTIPrep framework, the following checking procedures are applied within a pipelining structure and described by a protocoling file. These pipeline procedures include: 1) image info checking, 2) diffusion info checking, 3) slice-wise intensity checking, 4) interlace-wise intensity checking, 5) averaging baselines, 6) eddy-motion correction, 7) gradient-wise checking, 8) computing DTI measurements and saving. As our current study, we are focusing on extending DTIPrep to detect artifacts through DTI data. According to our experiments, there are a number of artifacts which can not be detected from the DWIs using only voxel-wise knowledge. For example, residual artifact presence after DWI QC which shows a reddish appearance for DTIs.
Key Investigators
- UIowa: Hans Johnson, Mark Scully, Joy Matsui
- UNC: Martin Styner, Clement Vachet, Mahshid Farzinfar
Objective
1. Introducing new automated QC tool called DTIPrep that includes the pipeline checking and correcting procedures against the DWI acquired protocol and generating QCed result.
2. Integrating the visual checking considering into the automated QC procedures consistently and synchronously.
3. Developing DTIPrep by introducing new measurement of DTIs properties to detect residual artifacts presence after DWI QC.
Approach, Plan
1. Checking size, origin, spacing of image, checking b-values, measurement frames and diffusion sensitizing directions. Applying statistical analysis on computed intensity-based correlations between successive slices and furthermore against interleaved parts of each gradient. Applying rigid and affine registration to correct and detect the motion artifacts. Generating xml QC report file. More details are found in [1].
2. Describing each gradient as "included" or "excluded" after not only applying QC but also visual checking analysis and defining further QC when some gradients get "included" in the visual checking step.
3. Detecting and correcting artifacts based on new entropy-based measurement of Principal Directions (PD) distribution within whole image.
Progress
* All the pipeline procedures of DWI QC have been implemented successfully. DTIPrep generates QC report in xml format and with facilities of reloading QC result and checking mismatching between the image and corresponding reloaded QC report. "-" In this week, we focused on the part which helps user to be able change the protocols features in the pipleline and generates proper QCed report.
* Doing visual checking after QC and then asking user whether further QC is applied for updated gradients or the QCed result is just saved. "-" In this week, we did testing an checking for saving QCed result after doing visual checking.
* Training and testing the new entropy-based measurement of PD histogram within gray-matter, white-matter areas and the entire image. "-" We were applying the new measurement on the new data from Tutorial on the whole image, while matter and gray-matter regions.
References
[1] Z Liu, Y Wang, G Gerig, S Gouttard, R Tao, T Fletcher, M Styner: Quality control of diffusion weighted images. Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, vol: 7628, pg: 17, 2010.
