Mevislab fMRI Data Processing learning 1

Task: Preprocessing of fMRI data on the Mevislab platform

Difficulties: Unfamiliar to the Mevislab platform, the platform itself about the fMRI content of less (I do not have to do = =); Language barriers, a lot of professional vocabulary can not translate, or translation is not over ...

Specific process:

Introduction of functional Core magnetic data

Automatic parameter Extraction

Level time correction

Head motion correction (rigid body transformation)

Image Registration

Go to Baseline drift

Band-Pass Filtering

Time delay (temporarily not known)

Progress:

0. Right-click Show Help for the module (Modules) You do not know. [packages/mevislab/standard/documentation/publish/modulereference/genre.html can display core by class in the installation directory modules]

1. About data import, requirements for the DICOM format import, can be folder identification What, try a lot of import methods, finally determine Directdicomimport, select folder, can identify file information. In this case, the imported file is mosaic arranged, in order to later normal processing, it needs to be a slice in the z out, using the Demosaic module, the parameter is [X,y,z,t,c,u] the image set. Here x, Y, z is the spatial coordinates of the image, T is the time coordinate, that is, how many dicom files, how many tr, C=color, refers to the image of the color display method, the MRI data for the grayscale display (Gary). You seem to be user, but I try to import multiple subjects, you don't change, you probably understand the error, and you need to confirm it.

2. Parameter extraction is simple, directly through info implementation, if the need to implement GUI later, can directly refer to the parameters;

3. Time-level correction (optional), has not yet found the relevant module, can be implemented by their own module, in the generated CPP project using the loop of the T in loop to achieve time-level correction, in addition, in the info information also has the display of time information, After confirmation is not the level of time information. It is also necessary to continue learning the content of the SPM mid-level time correction.

4. Head motion correction, using rigid-body transformation model to achieve. The specific module is merit and the transformation type is rigid. When the head motion correction is performed, select the T=1 image as the template image, use Subimage to select a sub-image from the imported dicom file, Subimage can re-select six coordinates, here I only select the T, In the SPM manual, it seems that the first frame is used as the basis for the head motion correction, then the first frame is selected with Subimage. The MERIT can be used to transform the image into a rigid body and get the transformation matrix (merit/output/transformation Matrix). But the problem is that merit can only transform an image into a template image transformation matrix, and fmri-bold data is absolutely more than a pair, how to achieve multi-image correction, I still did not think of the appropriate method, in my existing understanding in the Mevislab, module and module links can only transfer data, Instead of data flow, that is, I can't implement a circular reference to certain links outside of the module, should I re-develop the module based on merit? Or is there another way to do it, but I haven't found it yet?

In addition, the Registrationmanual module can display the registration effect through overlay, which is used as inspection tool during the debugging process.

5. Image registration, the same use of merit module, the transformation type is rigid. The template image is T1, the registration operation, the specific problem and the same as the head motion correction. In the registration process, the structure image needs to be reconstructed, and the resolution and coordinate information can be adjusted by Resample3d. [In the SPM is the first completion of the segmentation and registration, if not split whether there is an impact, need to split it? ]

6. Go to Baseline drift & bandpass filter processing to implement filtering directly using Fftbandpassfilter. The core algorithm is realized by fft→ Plus filter template →ifft.

For the improvement of baseline drift, do not know how to test the effect, and how to improve, but also need to learn.

7. Time Delay Correlation characteristics

8. Normalization to the standard brain template (optional) and head motion correction and image registration is the same, all belong to the space transformation, the transformation form is different, need to use the whole transformation affine transform affine realization, template image for the SPM provided standard brain template, the use of merit, transformation type of affine.

Other accumulation:

1. Forum Study: http://forum.mevis.fraunhofer.de/index.php

2. Modules that may be used

Imagepropertyconvert	Adjust image parameters, including resolution, size, etc.
Soview2doverlay	Used to compare pictures
Imagespector	Separates an image along a coordinate, but up to 6 outputs ...
Reshape	X*y*z*c*t*u remains unchanged
Sogvrvolumerender	Render rendering for 3D, 4D
Minmaxscan	resizing, which may be used during the transformation
Affinetransformation3d	Perform 3D affine transformations
Fiter class	Filtering and smoothing, etc.

　

Mevislab fMRI Data Processing learning 1