Meta-analysis of Voxel-Based Neuroimaging Studies using Seed-based d Mapping with Permutation of Subject Images (SDM-PSI)
November 27th, 2019
We detail how to conduct a meta-analysis of voxel-based neuroimaging studies using Seed-based d Mapping with Permutation of Subject Images (SDM-PSI).
Seed-based d mapping is a method for putting the meta-analyzing neuroimaging studies that may seem contradictory. For example, the results from studies of gray matter abnormalities in OCD. Previous meta-analytic methods have commonly assessed whether studies support more statistical peaks in one reason or in another, while SDM-PSI assesses whether there is an effect in a particular reason.
Researchers have mostly used SDM to investigate the neural substrates of psychological functions and neuropsychiatric disorders, which may include fMRI, VBM, DTI, PET or SBM studies. The use of SDM-PSI so far is quite straightforward. However, the most important steps for conducting a correct meta-analysis are designing an accurate meta-analysis plan and carefully collecting the data.
To perform an exhaustive search, open the database of interest and enter various keywords that will allow searching for any study that might meet the inclusion criteria. Then, record the number of studies retrieved and the number of studies excluded for each reason. Consider creating a PRISMA flow diagram with these numbers.
After reading the manuscripts of interest to find the specific data to extract, to convert Z-values and P-values into T-values, click Convert peaks in the SDM-PSI software. Then use positive T-values for peaks of increase and negative T-values for peaks of decrease, using the table for guidance for how to decide the sign of the T-values. In two sample studies, you use positive T-values when MRI signal is higher in patients than in controls.
And use negative T-values when MRI signal is lower in patients than in controls. To introduce the data into SDM-PSI, open SDM-PSI graphic user interface and close the About splash window. Click Change meta-analysis to select a new empty directory for the meta-analysis.
Click SDM table editor to input general information from the studies and open a text editor to create a text file for each study to type the coordinates and T-values of the peaks. And name the file, the software program used and the stereotactic space. For preprocessing of the data, click the preprocessing button, select the modality of the studies at the list box labeled Modality and press OK.Then wait several minutes while the permutation of subject images program calculated the maps of the lower and upper bounds of the potential effect sizes.
To perform the main analysis, click the Mean button and press OK.After the SDM-PSI has conducted the multiple imputation and meta-analysis, click Threshold, select the uncorrected P-values of the main analysis and click OK.SDM-PSI will automatically open both MRIcron to visualize the results and a web page with a detailed report of the results. Click family-wise error correction to select the main analysis in the list box and click OK.The SDM-PSI program will conduct the permutation test over several hours or days. At the end of the rest, click Threshold, select the TFCE correction of the main analysis and click OK.The program will automatically open both MRIcron to visualize the results and a web page with a detailed report of the analysis.
For heterogeneity, publication bias and grading, click extract, select a peak from the main analysis and press OK.The program will automatically open a web page with statistics of this P.Record the heterogeneity I-squared statistic. Next click the Bias Test button, select a peak from the main analysis and click OK.The program will automatically open a web page with a funnel plot and the results of a test for small-study effect and a test for excess significance. Then click the Evidence Grading button from the top tool box.
Select the main analysis from the list box and click OK to view the results. As observed in these representative maps, patients with OCD have a statistically significantly smaller gray matter volume in the dorsal anterior cingulate medial frontal cortex. The cluster is moderately small and mainly located at Brodmann area 32 with the peak of the cluster at MNI two, 32, 32 with a Z-value of 4.97 and a family-wise error rate corrected P-value of 0.01.
In this data analysis, the low I-squared statistic indicates very small heterogeneity and the funnel plot does not show asymmetries. Key aspects of a successful meta-analysis are the creation of clear inclusion and exclusion criteria, the careful collection of data and not limiting the results to P-values. SDM-PSI incorporates the general linear model for which users might conduct meta-regressions or meta-compressions of groups of studies and or at compare it to the analysis.
Many SDM meta-analysis have had a considerable influence on their specific fields. For example, several meta-analysis have revealed interesting effects of stimulants on the brain abnormalities observed in ADHD.
Overview
This article details the process of conducting a meta-analysis of voxel-based neuroimaging studies using Seed-based d Mapping with Permutation of Subject Images (SDM-PSI). The method aims to reconcile results from various studies that may appear contradictory, particularly concerning gray matter abnormalities in conditions like OCD.
Key Study Components
Area of Science
- Neuroscience
- Neuroimaging
- Meta-analysis
Background
- Seed-based d mapping (SDM) addresses discrepancies in neuroimaging studies.
- SDM has been applied to various psychological functions and neuropsychiatric disorders.
- Unlike previous methods, SDM-PSI evaluates the presence of effects rather than just statistical peaks.
- Clear inclusion criteria and careful data collection are crucial for successful meta-analysis.
Purpose of Study
- To demonstrate the steps involved in conducting a meta-analysis using SDM-PSI.
- To provide insights into the preprocessing and analysis of neuroimaging data.
- To highlight the impact of neuroimaging studies on understanding neuropsychiatric disorders.
Methods Used
- The primary platform utilized is SDM-PSI software for meta-analysis of neuroimaging data.
- The study focuses on neuroimaging studies relating to OCD, including various imaging modalities like fMRI and PET.
- Essential methodological steps involve designing a meta-analysis plan, data collection, and converting statistical values using the SDM-PSI interface.
- Critical steps such as data preprocessing and execution of the main analysis are highlighted.
Main Results
- Patients with OCD exhibited significant gray matter volume reductions in specific brain areas.
- Statistical analyses revealed important findings regarding heterogeneity and potential biases in existing studies.
- Results illustrated the practical application of SDM-PSI in uncovering neuroimaging effects and spatial distributions.
- Key insights into how SDM-PSI methods can be leveraged for future neuroimaging meta-analyses.
Conclusions
- This study demonstrates the utility of SDM-PSI for conducting comprehensive meta-analyses of neuroimaging studies.
- Findings underscore the significance of careful methodological approaches in revealing inconsistencies across studies.
- The implications of the results enhance the understanding of neuroanatomical variations in psychiatric disorders.
Frequently Asked Questions
Executive Industry Relevance
In neuropharmacology and CNS drug development, reconciling contradictory neuroimaging findings is critical for target validation and mechanistic de-risking. SDM-PSI advances beyond binary significance testing by quantifying effect strength and assessing bias, enabling more confident go/no-go decisions in early discovery. This supports portfolio triage by providing continuous evidence grading rather than binary outcomes.
Strategic Applications in Biopharma R&D
Early Discovery & Target Validation
- Scientific Value: Interrogates therapeutic hypotheses by assessing whether observed neuroimaging effects are non-null rather than merely counting significant peaks.
- Operational Value: Enables pathway clarification through effect size estimation and bias detection, reducing false target assumptions.
- Predictive Value: Supports predictive confidence by grading evidence strength across significance levels and data quantity, informing target prioritization.
Screening & Assay Development
- Assay Readiness: Prepares validated neuroimaging biomarkers for downstream screening by establishing reproducible effect maps with quantified uncertainty.
- Quantitative Outputs: Generates continuous effect size estimates (d-values) and heterogeneity metrics (I-squared) that enable reliable compound screening decisions.
- Platform Reuse: Standardized workflow supports cross-modality application (fMRI, VBM, DTI, PET) for consistent biomarker validation across discovery campaigns.
Translational & Preclinical Research
- Translational Continuity: Bridges discovery and preclinical validation by providing disease-relevant effect maps with bias assessment, supporting mechanistic de-risking.
- Risk-Adjusted Advancement: Evidence grading criteria (incorporating significance levels, data amount, small-study effect, excess significance) inform preclinical go/no-go decisions.
- Biomarker Alignment: Enables alignment of imaging biomarkers with phenotypic outcomes through spatially specific, bias-corrected effect maps.
Pipeline & Workflow Integration
SDM-PSI integrates into the discovery continuum from target validation through lead identification to preclinical work by providing quantitative, bias-aware neuroimaging evidence that supports hypothesis testing and mechanistic de-risking.
- Discovery Biology: Supports hypothesis testing by assessing non-null effects and estimating effect sizes, moving beyond peak counting to biological effect quantification.
- Screening: Delivers assay-ready, reproducible effect maps with quantitative outputs (T-values, d-values, P-values) enabling reliable compound evaluation across modalities.
- Analytics: Provides statistical outputs including heterogeneity (I-squared), bias tests (funnel plot, excess significance), and graded evidence levels for cross-functional interpretation.
- Translational Research: Connects to preclinical continuity through disease-relevant effect maps (e.g., OCD gray matter volume) with quantified uncertainty and bias assessment.
- Enterprise Reuse: Establishes a reusable neuroimaging meta-analysis platform applicable across psychiatric indications and imaging modalities, reducing redundant method development.
Operational & Enterprise Impact
- Scientific Value: Predictive confidence through effect size estimation, target validation via non-null effect assessment, reduction of mechanistic ambiguity by quantifying biological effects.
- Operational Value: Standardization via predefined preprocessing and analysis pipelines, reproducibility through permutation-based inference, scalability across studies and modalities.
- Strategic Value: Better go/no-go decisions via evidence grading, capital efficiency by reducing false target pursuit, reduced late-stage biological risk through early bias detection.
- Portfolio Impact: Risk-adjusted prioritization using graded evidence (significance levels, data amount, bias detection), advancement decisions informed by continuous effect metrics.
Implementation Considerations
- Expertise in neuroimaging meta-analysis, statistical interpretation of permutation tests, and understanding of bias assessment methods (funnel plots, excess significance tests).
- SDM-PSI software infrastructure, capacity for permutation testing (hours/days per analysis), and visualization tools (MRIcron) for result interpretation.
- Cross-team standardization of inclusion/exclusion criteria, data extraction protocols (converting Z/P to T-values), and stereotactic space reporting (MNI/Talairach).
- Adaptation considerations across neuroimaging modalities (fMRI, VBM, DTI, PET, SBM) requiring modality-specific preprocessing in SDM-PSI.
- Practical limitations include dependence on primary study quality, potential bias from unpublished null results, and computational demands of permutation testing for large datasets.
Biopharma FAQs
Chapters in this video
0:04
Title
0:59
Exhaustive Search and Data Collection
2:03
Seed-Based d Mapping with Permutation of Subject Images (SDM-PSI) Data Introduction
2:42
Pre-Processing and Main Analysis
4:08
Heterogeneity, Publication Bias, and Grading
4:53
Results: Representative Smaller Gray Matter Volume Analyses
5:38
Conclusion
