Brain Infarct Segmentation and Registration on MRI or CT for Lesion-symptom Mapping

Provided here is a practical tutorial for an open-access, standardized image processing pipeline for the purpose of lesion-symptom mapping. A step-by-step walkthrough is provided for each processing step, from manual infarct segmentation on CT/MRI to subsequent registration to standard space, along with practical recommendations and illustrations with exemplary cases.

Lesion-symptom mapping is a powerful tool to localize brain functions by studying patients with brain lesions. An important step involves image processing, including lesion segmentation and registration to standard space. This protocol describes a unified framework that works with all structural imaging modalities, providing a consistent output of lesions in standard space for use in lesion-symptom mapping studies.

The lesions must be transformed to standard space to allow for comparisons across subjects. Each patient's brain must be spatially aligned to correct for differences in brain size and shape. While using this method, it helps to see how each step should be performed, rather than following written instructions alone.

It also requires a good understanding of brain images, which makes a visual demonstration useful. This protocol describes how to perform the processing steps required for lesion-symptom mapping. Once the lesion maps are registered to standard space, lesion-symptom mapping can be performed on large groups of patients.

When all the lesion maps are transformed to standard space, each voxel represents the same brain region across subjects. This allows for statistical analyses to be performed, for example, to test if the presence of a lesion in a particular voxel is associated with a cognitive deficit. Begin by collecting brain CT or MRI scans of patients with ischemic stroke.

Most scanners have the scans as DICOM files that can be copied to a hard disk or server. Collect the clinical variables in a data file by making separate rows for each case and columns for each clinical variable. For infarct segmentation, include at least the date of stroke and the date of imaging or a variable that indicates the time interval between stroke and imaging.

To convert the DICOM images to uncompressed NIfTI files, using the DICOM-to-NIfTI tool, type the command seen on-screen here in the command prompt, using the folder path to the DICOM files. An example of the command with the folder paths inserted can be seen here. This command will run the executable, convert the DICOM images in the selected folder, and save the NIfTI files in this folder.

Finally, organize the NIfTI files in a convenient folder structure, with a subfolder for each case, before beginning infarct segmentation. First, assure that the scan was performed at least 24 hours after stroke symptom onset. Otherwise, the acute infarct will not be visible on CT or will only be partially visible, and the scan cannot be used for segmentation.

Open the native CT using ITK-SNAP software by selecting File, then Open Main Image from the dropdown menu. Then click Browse, and select the file to open the scan. Now identify the infarct based on imaging characteristics.

First, note that infarcts have a low signal compared to normal brain tissue. In the acute stage, large infarcts can cause mass effect, resulting in displacement of surrounding tissues, compression of ventricles, midline shift, and obliteration of sulci. There can be hemorrhagic transformation, which is visible as regions with high signal within the infarct.

In the chronic stage, the infarct will consist of a hypodense, cavitated center, with a similar density as the cerebrospinal fluid, and a less hypodense rim, which represents damaged brain tissue. In case of large infarcts, there can be ex vacuo enlargement of adjacent sulci or ventricles. Segment the infarcted brain tissue using the Paintbrush Mode from the main toolbar, using left-click to draw and right-click to erase.

Alternatively, use the Polygon Mode to place anchor points at the borders of the lesion, and hold the left mouse button while moving the mouse over the borders of the lesion. Once all points are connected, click accept to fill the delineated area. After finishing the segmentation, save it as a binary NIfTI file in the same folder as the scan by clicking Segmentation and Save Segmentation Image from the dropdown menu, then save the segmentation by giving it the exact same name as the segmented scan, with the extension of lesion.

First, assure that the DWI scan was performed within seven days of stroke onset. Infarcts are visible on DWI within several hours after stroke onset, and their visibility on DWI gradually decreases after approximately seven days. Next, identify and annotate the infarcted brain tissue based on the high signal on DWI and low signal on the ADC map.

Do not mistake a high diffusion signal near interfaces between air and either tissue or bone, which are a commonly observed artifact on DWI. For FLAIR imaging, first check that the scan was performed 48 hours after stroke symptom onset. In the hyperacute stage, the infarct is usually not visible, or the exact boundaries are unclear.

Next, open the FLAIR image in ITK-SNAP, as well as the T1-weighted scan in a separate window for reference, if available. In the acute stage, the infarct is visible as a somewhat homogeneous hyperintense lesion, with or without swelling and mass effect. If available, use the DWI to differentiate between the acute infarct and chronic lesions, such as white matter hyperintensities.

To register images, first download RegLSM, and use this tool to process CT scans or any kind of MRI sequence. The registration procedure can be seen in the figure shown here. Next, open MATLAB Version 2015a or higher, set the current folder to RegLSM, then enable SPM by typing addpath followed by the folder name for SPM.

Next, type RegLSM to open the graphic interface. To perform the registration for a single case, select Test Mode in the Registration dropdown menu. Then use the Open Image button to select the segmented scan, the annotation, and optionally the T1, and select the desired registration scheme.

Alternatively, Batch Mode can be used to register all cases in the selected folder. Ensure that RegLSM saved the resulting registration parameters, the registered scans, and the registered lesion map in the automatically generated subfolders. Now, review the registration results by selecting Check Results in the RegLSM interface, and browse to the main folder with these results.

Scroll through the registered scan, and use the crosshair to check the alignment to the MNI152 template, paying attention to recognizable anatomical landmarks. Be sure to mark all failed registrations in a separate column in the previously created data file for subsequent manual correction. For any lesion maps that do need correction, open the MNI152 T1 template in ITK-SNAP, then select Open Segmentation from the Segmentation menu, and choose the registered lesion map, which will overlay onto the template.

Also, open the registration results in a separate window for reference. Correct the registered lesion map in ITK-SNAP for any type of misalignment, using the brush function to add voxels with left-click or remove voxels with right-click. Finally, save the corrected lesion map as a NIfTI file in the same folder as the uncorrected lesion map, then save the segmentation by giving it the exact same name as the uncorrected lesion map, with the extension of corrected.

Here, we see CT scans for a single patient. The initial scan cannot be used for segmentation because the infarct is not yet visible, even though the CT perfusion maps show ischemia. The CT on day six shows swelling of the infarcted brain tissue with slight midline shift and hemorrhagic transformation.

The scan after four months shows brain tissue loss with ex vacuo enlargement of ventricles and nearby sulci. The result of registration of the CT on day six to the MNI152 template is seen here. The registration algorithm insufficiently compensated for the midline shift and compression of the left ventricle, which required a manual correction.

After correction, the registered lesion map is an accurate representation of the infarct in native space and can be used for lesion-symptom mapping. Here, we see comparison of the registered DWI sequence and the corresponding registered infarct with the MNI152 template. Note the slight error at the head of the right caudate nucleus.

This required a manual correction of the segmentation in standard space. Finally, this figure shows the results of the registration of the MRI FLAIR sequence on day three, which is adequate and requires no manual correction. This protocol covers the process of preparing imaging data for lesion-symptom mapping, including practical guidelines on how to perform accurate infarct segmentation and registration to a brain template.

The lesion maps can be used in lesion-symptom mapping studies to localize a specific cognitive function or to use lesion location to predict cognitive outcome in stroke patients. This protocol accelerates and harmonizes the workflow for lesion-symptom mapping studies, enabling researchers to perform large multicenter studies including hundreds to thousands of subjects.