In this study, five analysis pipelines were constructed to extract lesion-related features from brain MRI. The first is a preprocessing pipeline for the noise correction and spatial registration, and the second is a volumetry pipeline for calculating volumes of anatomical regions of the brain. The third is the WMH pipeline that extracts White Matter Hyperintensity (WMH) from FLAIR image, the fourth is the T2 HSL pipeline that extracts T2 high signal lesions, and the fifth is the Myelination Index pipeline that extracts myelination levels of each white matter tracts
The preprocessing process begins with converting DICOM data downloaded from the PACS server into a file in NIFTI format, which is mainly used in brain image analysis. The patient's MRI converted to the NIFTI format undergoes a process of correcting magnetic field nonuniformity. Thereafter, T2 weighted MRI and FLAIR MRI are spatially registered to the patient's T1 weighted MRI. And the brain parenchyma is extracted from the T1 MRI and registered on the MNI template. In the figure 1, the flow of the preprocessing, script files for the analysis, and the outputs are described.
The volumetry pipeline begins with the inverse registration of standard anatomical regions from AAL Atlas onto the patient's brain MRI by using the patient's T1 weighted MRI, standard T1 MRI, and AAL Atlas as inputs. Through the segmentation of gray matter(GM), white matter(WM), and CSF in the patient's T1 MRI , the volume of each tissue is obtained, and the volume of 268 anatomical regions that have registered inversely is also obtained. In the figure 2, the analysis flow in the volumetry pipeline, script files for the processing, and the outputs are described.
The WMH pipeline begins with the segmentation of the WMH in the preprocessed FLAIR image using the LST algorithm. The standard WM region is inversely registered on the patient's FLAIR image to obtain the individual WM region. And, the WMH extracted with the LST algorithm is masked to extract the WMH volume in each individual WM area. In the figure 3, the processing flow in the WMH pipeline, script files for the analysis, and the outputs are described.
The T2 HSL pipeline consists of almost the same steps as the WMH pipeline, but there is a difference in that it inversely transforms the standard anatomical region rather than the standard white matter region into the patient FLAIR imaging space. Then, the abnormal T2 high signals in FLAIR image extracted by the LST algorithm is masked with the patient's anatomical area to obtain the volume of T2 high signal lesion in each area. In the figure 4, the processing flow in the T2 HSL pipeline, script files for the analysis, and the outputs are described.
Previous studies have shown that ratio image from T1 and T2 weighted MRI is effective in mapping axonal myelination (Marco Ganzetti et al., 2014, Glasser, M.F., and Van Essen, D.C., 2011). Based on these researchies, a pipeline was constructed as follows to extract the myelination level of each patient. The analysis begins with the input of the patient's T1 and the spatially registered T2 MRI. After inverse transformation of the standard white matter tracts to the patient's T1 MRI, the T1/T2 ratio image is generated from the T1 and T2 weighed MRI to extract the average value of the T1/T2 ratio values in the patient's white matter paths. In the figure 5, the processing flow in the Myelination Index pipeline, script files for the analysis, and the outputs are described.
* Several atlas files and their informations are required to process brain MRI images in the aforementioned procesing pipelines. You can download these atlas files and the matlab m files that generate them through the Utils link.