Automated Segmentation of Cortical Grey Matter from T1-Weighted MRI Images
January 7th, 2019
This protocol describes the process of applying seven different automated segmentation tools to structural T1-weighted MRI scans to delineate grey matter regions that can be used for the quantification of grey matter volume.
This method can answer key questions in the neuroimaging and neurology fields. Such as, whether there are group differences in cortical volume for non-clinical versus clinical populations. The main advantage to this technique is that it enables researchers to use the best tool for their data.
To begin, first open the SPM software by opening a MATLAB command window and typing SPM into the command line. Then, to perform unified segmentation, select PET VBM to open the structural MRI toolbox. Now, open the batch editor to perform segmentation on multiple scans at once.
Select SPM, Spatial, and Segment, then Data, select Files, and choose the T1-Weighted scans as Input. Next, click on Output Files, Grey Matter, and ensure that Native Space is selected, and repeat this for White Matter. If the CSF segmentation is not required, leave this set to None.
If the scans have already been Bias Corrected, change this option to Don't Save Corrected. Then use Clean up any partitions and test all three options prior to running the full analysis. Now, leave the other settings set to defaults and click on the green flag to run the segmentation.
The MATLAB window will say Done when segmentation is finished. Finally, perform Visual Quality Control on the resulting Grey Matter NIfTI file. To perform the New Segment option in SPM 8, first select PET VBM before opening the batch editor.
Then select SPM, Tools, New Segment, and select the NIfTI format T1 image files to be used. Set the Native Tissue Type option to Native Space, and turn off Tissue Classes that are not required. Also turn off Warped Tissue, then click on the green flag to run the segmentation and perform Visual Quality Control when completed.
To perform segmentation in SPM 12 again press PET VBM and open the batch editor. Then select SPM, Spatial Segment, and Data Volumes. Then select Native Space Tissue Type and turn off unneeded Tissue Classes.
Set Warped Tissue to None, and click the green flag. Once the segmentation has completed, again be sure to perform Visual Quality Control as detailed in the next section of this protocol. Visual Quality Control can be performed using FSLeyes.
Begin by opening a terminal window, then open FSLeyes by typing FSLeyes in the terminal. Then select File, Add from File, and select the original T1 and the segmented regions to view them. Once FSLeyes opens, use the opacity toggle to allow visualization of the underlying T1 image.
Also change the color of the segmentation overlay as needed via the color dropdown tab in the top pane. Now, scroll through every slice in the brain and check each one for regions of under or overestimation in the region being inspected. Visual Quality Control is an essential step for this procedure.
By comparing your segmented regions to the original T1 scan, you can ensure that your regions are of a high quality, and that your conclusions are biologically accurate. To perform Visual Quality Control of FreeSurfer data using Freeview, open a terminal window and change the directory to the subject folder that contains the processed FreeSurfer output. Then type the command seen on on the screen here to view the volumetric grey matter region overlaid on the T1.Again, scroll through every slice in the brain and check for regions of under or overestimation for the brain region being inspected.
Here, we see an example of a failed segmentation displayed on a T1 scan. This segmentation should be reprocessed and excluded from analysis if it cannot be improved. This figure shows examples of the performance of different tools on the temporal lobe on a T1 scan.
Examples of a good regional delineation are seen here, while examples of poor regional delineation, showing spillage in the left and right temporal lobes, are shown here. This figure shows examples of the performance of different tools on the occipital lobe on a T1 scan. Here we see the T1 scan with an example of a good regional delineation, while here is an example of a poor regional delineation, showing spillage into the medial dura.
Here we see an example of a grey matter region spilled into the dura, highlighted by the blue region. This figure shows an example of a grey matter region that has excluded regions of the cortex from segmentation, best shown in the axial view. While attempting this procedure, it's important to remember to test different tools in your data, and to perform Visual Quality Control on the process scans.
After it's development, this technique paved the way for neuroimaging researchers to study changes in brain volume over time without requiring invasive tests.
View the full transcript and gain access to thousands of scientific videos
Overview
This protocol outlines a method for applying seven different automated segmentation tools to structural T1-weighted MRI scans. The goal is to delineate grey matter regions, facilitating the quantification of grey matter volume, which can aid in understanding group differences in cortical volume between non-clinical and clinical populations.
Key Study Components
Area of Science
- Neuroimaging
- Neurology
- Brain segmentation analysis
Background
- Investigates cortical volume differences across populations.
- Utilizes automated segmentation techniques for efficiency.
- Emphasizes the importance of Visual Quality Control for accuracy.
- Demonstrates segmentation performance of various tools.
Purpose of Study
- To provide a reliable method for quantifying grey matter volume.
- To facilitate non-invasive monitoring of brain volume changes.
- To compare segmentation outcomes of multiple tools for precision.
Methods Used
- SPM software in MATLAB is used for segmentation tasks.
- T1-weighted MRI scans serve as the primary data source.
- Multiple segmentation techniques are tested for efficiency.
- Visual Quality Control protocols are applied to validate results.
Main Results
- Found variability in segmentation accuracy across different tools.
- Emphasized the necessity of performing Visual Quality Control.
- Noted specific instances of successful and unsuccessful regional delineation.
- Concluded that careful testing of tools influences quality outcomes.
Conclusions
- Demonstrates a comprehensive approach to grey matter volume quantification.
- Highlights the importance of tool selection and quality control.
- Provides groundwork for future research on brain volume changes without invasive methods.
Frequently Asked Questions
Executive Industry Relevance
In neuroimaging research, variability in segmentation tool performance can introduce bias in volumetric measurements, affecting the reliability of group comparisons in cortical volume studies. This protocol enables researchers to evaluate and select the most accurate automated segmentation method for their specific dataset, improving measurement consistency and reducing technical noise. By standardizing quality control through visual validation, the approach supports robust, reproducible quantification of grey matter volume, which is critical for target validation and mechanistic de-risking in CNS drug discovery programs.
Strategic Applications in Biopharma R&D
Early Discovery & Target Validation
- Scientific Value: Enables interrogation of therapeutic hypotheses by providing accurate cortical volume measurements to assess group differences between disease models and controls.
- Operational Value: Supports biological de-risking through standardized segmentation workflows that reduce variability in neuroimaging endpoints.
- Predictive Value: Enhances confidence in target engagement studies by ensuring volumetric readouts are derived from validated, high-fidelity tissue delineation.
Screening & Assay Development
- Assay Readiness: Prepares validated biological systems (brain imaging datasets) for downstream screening by establishing reproducible grey matter segmentation protocols.
- Quantitative Output: Generates standardized volumetric measurements that enable reliable compound effect screening in preclinical CNS models.
- Platform Reuse: Encourages cross-study applicability of segmentation tools, allowing consistent assay performance across multiple therapeutic areas and target classes.
Translational & Preclinical Research
- Disease Relevance: Supports translational biomarker alignment by providing accurate grey matter volume readouts that correlate with neurodegenerative or neuropsychological phenotypes.
- Preclinical Continuity: Facilitates longitudinal tracking of brain volume changes in disease models, enabling risk-adjusted advancement decisions based on mechanistic and phenotypic outcomes.
- Mechanistic De-risking: Reduces false positives in target validation by ensuring observed volumetric changes reflect true biological effects rather than segmentation artifacts.
Pipeline & Workflow Integration
The method fits within the discovery continuum from early target validation through preclinical evaluation, where accurate neuroimaging phenotyping informs lead identification and go/no-go decisions. It enables consistent assay development by providing a framework for comparing segmentation tools and selecting optimal pipelines for specific datasets. The quantitative volumetric outputs support analytics-driven comparisons across treatment groups, time points, or genetic models. When applied longitudinally, the method contributes to translational research by linking imaging biomarkers to preclinical disease progression. Enterprise reuse is supported through standardized protocols that can be adopted across imaging cores and external collaborators.
- Discovery Biology: Supports hypothesis testing and pathway clarification by delivering reliable cortical volume data to assess target modulation effects in disease-relevant systems.
- Screening: Enhances assay readiness and reproducibility through standardized segmentation and visual quality control, ensuring consistent input for compound screening campaigns.
- Analytics: Provides volumetric readouts and quality-controlled outputs that enable statistical comparison of experimental conditions and treatment effects.
- Translational Research: Connects imaging biomarkers to preclinical continuity by validating grey matter segmentation as a surrogate for neurodegeneration or neuroprotection.
- Enterprise Reuse: Establishes a reusable neuroimaging capability that reduces redundant method development and promotes cross-project standardization.
Operational & Enterprise Impact
- Scientific Value: Increases predictive confidence in target validation by minimizing technical variability in neuroimaging endpoints.
- Operational Value: Improves reproducibility and scalability through standardized tool evaluation and visual quality control procedures.
- Strategic Value: Enhances go/no-go decision-making by reducing late-stage attrition risk from unreliable biological readouts.
- Portfolio Impact: Enables risk-adjusted prioritization of CNS programs based on biologically accurate, volumetrically validated target engagement data.
Implementation Considerations
- Requires expertise in neuroimaging analysis, MRI processing, and familiarity with segmentation toolkits such as SPM, FreeSurfer, and FSL.
- Depends on access to MATLAB, SPM, FSLeyes, or FreeView for image processing and visualization.
- Necessitates cross-team standardization of segmentation protocols and quality control criteria to ensure consistency across sites or vendors.
- Involves adaptation considerations when applying the protocol to different species, developmental stages, or pathological conditions affecting brain morphology.
- Includes practical limitations such as time investment for tool comparison and the need for manual visual inspection, which may limit throughput in large-scale studies.
