Tuesday, Jun 24: 9:00 AM - 1:00 PM
1528
Educational Course - Half Day (4 hours)
Brisbane Convention & Exhibition Centre
Room: P3 (Plaza Level)
Meta-analysis remains a cornerstone of evidence synthesis in neuroimaging, yet the exponential growth of published studies far exceeds the rate at which meta-analyses are conducted. With the advent of widespread data sharing, particularly through platforms like NeuroVault, researchers now have access to full statistical images—enabling the more powerful Image-Based Meta-Analysis (IBMA). These images offer a level of detail and resolution far beyond that of traditional Coordinate-Based Meta-Analyses (CBMAs), which are inherently limited by reliance on peak coordinates.
This course addresses the urgent need to train researchers in these emerging methodologies. Participants will learn to:
Perform reproducible CBMAs and IBMAs using modern tools and workflows.
Identify and curate relevant statistical images for IBMA, accounting for metadata sparsity.
Utilize Large Language Models (LLMs) to enrich metadata and enhance data selection for meta-analysis
Validate the quality and accuracy of metadata and analyses to ensure robust findings.
By mastering these skills, students will be prepared to conduct transformative meta-analyses that leverage openly shared data, advancing the field of neuroimaging and driving new discoveries. This course is not only timely but essential for bridging the gap between growing data availability and methodological expertise.
Identify the relevant filters for selecting appropriate images for IBMA
Execute a fully reproducible meta-analysis
Name the steps in the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA)
Understand how to validate LLM output
This course will be useful for researchers and clinicians with any of these three goals:
1) To read and understand a meta-analysis’ methods, limitations, and interpretation; 2) To use past meta-analyses to form new hypotheses and interpret results in novel research; and 3) To perform neuroimaging meta-analyses in an open and reproducible manner.
Presentations
Learning outcome
--------------------------
Understanding the role of meta-analyses in neuroimaging and the broad families of methods.
Points to cover
----------------------
Explain the limitations of single studies in neuroimaging, and the relation to the reproducibility crisis.
Explain the main uses of meta-analyses: synthesizing past results, developing new hypotheses, and interpreting novel results.
Outline the broad families of methods: coordinate- versus image-based methods, and manual versus automated methods.
Presenter
Angela Laird, Florida International University Miami, FL
United States
Timing
----------
20 min lecture
Learning outcome
-------------------------
Describe the strengths and weaknesses of different algorithms for CBMA
Points to cover
----------------------
Provide an overview of the different algorithms in CBMA
Describe automatic versus manual meta-analysis
Introduce Neurosynth Compose
Timing
----------
20 min lecture
Learning outcome
-------------------------
Describe how LLMs can be used to aid in the search/filtering of relevant data in Meta-analyses
Points to cover
----------------------
Describe the workflow to extract metadata
Identify the features that can be extracted from academic papers
Explain the considerations for extracting different features from papers
Presenter
James Kent, UT Austin Austin, TX
United States
Presenter
Angela Laird, Florida International University Miami, FL
United States
Presenter
Yifan Yu, Big data institute, University of Oxford Oxford, Oxfordshire
United Kingdom
Timing
----------
20 min lecture
Learning outcome
-------------------------
Explain the practical considerations curating images on Neurovault
Points to cover
----------------------
Provide an overview of the data on Neurovault
Identify the filters used to select appropriate images
How to reproducibly execute an IBMA
Presenter
Julio Peraza, Florida International University Miami, FL
United States