How to Measure Cortical Folding from MR Images: a Step-by-Step Tutorial to Compute Local Gyrification Index

Measuring gyrification (cortical folding) at any age represents a window into early brain development. Hence, we previously developed an algorithm to measure local gyrification at thousands of points over the hemisphere¹. In this paper, we detail the computation of this local gyrification index.

The overall goal of this procedure is to quantify cortical folding or ification with an exquisite resolution based on cerebral MRI. This is accomplished by first creating accurate three dimensional representations of the cortical surfaces using the free surface software. The second step is to construct an outer surface tightly warping the pile surface using a morphological closing algorithm.

Next hundreds of overlapping circular regions are created over the outer surface and are matched to their corresponding patch of cortical area. Then the ratio of both corresponding regions is calculated at each point, resulting in individual maps of local ification, which can be subsequently used for statistical comparisons between different diagnostic groups. Ultimately, measurement of cortical structure in grownup children, adolescents, or adults can uncover abnormal cortical development that occurred in the first months of life.

The main advantage of this method compared to existing techniques such as curvature based or cell morph geometry, is that it provides an easily interpretable measurement of the ification, which is unrestrained by the circle walls and also offers exquisite resolution. This method can help us answer some key question for, for the development of psychiatric condition or neurodevelopmental disorder. In other words, we can understand how very early neurodevelopmental events cortical folding can later on impact the development of pathogenesis in adulthood.

We first had the idea of this technique when we realized that the ification index that was used in comparative neuroanatomy since decades could be a very powerful indicator of the brain development if we could overcome the issues in the way it was implemented. The method will be demonstrated to you by Mari, she, who is a senior researcher in our group, as well as Mari Ra, who is a senior researcher in Jean Philip Tran's Lab. This first part of the protocol uses the standard free surfer pipeline.

The commands detailed here describe one way to achieve the cortical surface reconstructions, but equivalent commands may also be used to import the raw MRI DICOM into free surfer. Begin by creating the folder architecture. Next, go to the MRI folder of your subject and convert the raw MRI in the free surfer format.

Then visualize the converted volume. Verify the image quality by checking that the orientation is correct, the contrast is sufficient and the image is not moved. Next, using free surfer create the three dimensional cortical mesh models.

This runs for several hours In order to cope with the problem of Barrett. Cellie free surfer first creates a unitary white matter volume, which is used later as a starting point for the great white surface. First, the reconstruction process optimizes the gray white surface according to the local radiant of intensity.

Then this surface is further expanded to the gray CSF interface to create the pile mesh model. The reconstruction process produces two 3D mesh models composed of about 150, 000 points per hemisphere. The white model is of the gray white interface, and the pile model is of the gray CSF interface.

All the surfaces and volumes are in their native space, allowing measurement without deformation. Now, check for the accuracy of these reconstructed surfaces by overlaying the two surfaces. The example shows the alignment for the subject distributed with the free surfer package.

Ensure that the white surface accurately follows the gray white interface and that the vessels and membrane are not included in the pile surface to manually correct the result of the reconstruction process. Consult the tutorials on the free surfer wiki. When you are satisfied with the surfaces compute the local ification index or LGI.

This command usually runs for about three hours for the two hemispheres of one study participant. Depending on the power of your workstation, the LGI process begins with the creation of an outer surface using morphological closing operation. Then about 800 overlapping circular regions of interest are created on the outer surface.

For each one of these regions, a corresponding region of interest is defined on the pile surface. The whole computation ends up with the creation of an individual map containing one LGI value for each point of the cortical surface. This results in about 150, 000 values per hemisphere at the end of the LGI computation.

Both hemispheres of each subject can be checked in free surfer as correct. LGI values are typically comprised between one and five. The minimum threshold can be set to one for a rapid check.

The purpose of a statistical group comparison is to quantify the effect of a group at each vertex over the cortical surface phase while controlling for the effect of gender and age. Begin with creating a study specific template, giving all your subjects and input. This command creates a subject named average.

Next, to use the classical commands to compare LGI results between groups, create a text file containing the description of the subjects involved in your study. This is also known as the free surfer group descriptor file. Then re-sample the LGI data in this space of the average subject for each hemisphere.

Smooth the data on the cortical surface to reduce the signal to noise. Now compute the group comparison at the level of each vertex. This requires a contrast text file to be created.

For example, in the case of FSG D text, the contrast text file should contain the values one minus one zero to compute the difference between controls and patients while controlling for age and gender. Finally, run the comparison. Then visualize the results on the average subject using TK surfer by loading the newly generated data file.

The Vertex wise analysis create a statistical map showing color coded P values overlay at each point. This example shows a significantly reduced ification, mostly in the medial aspects of the brain. Using the option configure overlay, you can further modify the P threshold as well as correct for multiple comparisons using false discovery rate.

Q deck is a graphical user interface implemented in free surfer as previously described with the command line. You will also need a study specific template to use of Q deck with local ification index. First pres smooth, the LGI data QE requires the construction of a data table.

That includes the description of the different groups and other confounding variables such as age. If the LGI is not available in the list of dependent variables in Q deck, then add the following line to the Q deck RC file located in your home directory. Using Q deck, you'll get exactly the same results as with the command line, and you will further be able to correct for multiple comparisons using Montecarlo simulations.

Even though the local ification index was developed for a Vertex wise analysis, averaging the data at a courser level may be instructive. For exploratory analyses mean LGI values can be extracted for the 34 gyro regions of the cortical parcelization scheme integrated in free surfer. Those measurements can then be compared between the different groups using classical statistics software.

So parcel wise analysis can be an attractive way to limit the number of statistical compar reasons. However, the LGI at each point already quantifies the stratification in the surrounding regions of interest. As a result, the average LGI in the given regions of interest can also reflect to some extent the ification in the neighboring regions of interest.

For solutions to problems that may be encountered during the free surfer or LGI processing, consult the archives of the free surfer mailing list. Along with this procedure, other methods like cortical thickness or T tractography can be combined. This will allow you to see how early changes in ification will associate with further cortical development or white matter tracks development.

Watching this video, you should have a good understanding on how to create three dimensional representation of the cortical surface, how to create individual maps of ification, and how to compare them between groups to detect statistically significant deteriorations of the cortical folding.