Andy's BrainCast

Andrew Jahn

Andy talks to the leading scientists and software developers in the field of neuroimaging, learning about the latest trends, interesting findings, and the future of brain science.

Episodes

  1. Bayesian Statistics and Individualized Neuroimaging | Amanda Mejia (Andy's BrainCast #005)

    19/02/2025

    Bayesian Statistics and Individualized Neuroimaging | Amanda Mejia (Andy's BrainCast #005)

    Today's guest is Amanda Mejia, an associate professor in the Department of Statistics at Indiana University. Dr. Mejia is the director of the StatMIND lab (https://www.statmindlab.com/), which develops statistical methods for the analysis of brain imaging data, including principled statistical and Bayesian approaches. The goal of these techniques is to increase statistical power and reliability, and also to advance the generalizability of fMRI studies to more diverse populations. Currently, she is collaborating with other researchers to apply these methods to a range of topics, including neurodegenerative diseases, neonatal development, and psilocybin therapy. Dr. Mejia will be hosting a half-day educational workshop at 2025's Organization for Human Brain Mapping Conference, focusing on incorporating spatial information into fMRI analysis: https://ww6.aievolution.com/hbm2501/Events/viewEv?ev=2297 She is also one of the lecturers for Robert Welsh's Advanced Statistical Methods in Neuroimaging and Genetics (https://medicine.utah.edu/psychiatry/advanced-statistics) Papers Discussed =============== Eklund et al. (2012) paper on pre-whitening: https://www.sciencedirect.com/science/article/pii/S1053811912003825?casa_token=15Pk6_RjtzsAAAAA:rPz-6S-KhJOeo_hp7cuU2qX5SIhL5YpT4S1hvdj9RPOIHnqrbQKaclUrFwpNxbo9S1iP-mcM_Q Eklund et al. (2016) paper on spatial autocorrelation and multiple comparisons correction in fMRI analyses: https://www.pnas.org/doi/pdf/10.1073/pnas.1602413113 Lindquist (2008): The Statistical Analysis of fMRI Data https://www.researchgate.net/profile/Martin-Lindquist/publication/45857794_The_Statistical_Analysis_of_fMRI_Data/links/0912f5099351f63829000000/The-Statistical-Analysis-of-fMRI-Data.pdf?_sg%5B0%5D=started_experiment_milestone&origin=journalDetail Lindquist & Mejia (2015): Zen and the Art of Multiple Comparisons: https://pmc.ncbi.nlm.nih.gov/articles/PMC4333023/ Table of Contents ============== 00:00 Introduction to StatMind Lab and Its Focus 02:54 Precision Medicine and Individualized Analysis 06:04 The Evolution of fMRI in Clinical Settings 09:01 Statistical Techniques in Neuroimaging 11:46 Bayesian Approaches in fMRI Analysis 15:02 Frequentist vs Bayesian: A Balanced Perspective 17:56 Resources for Understanding fMRI Data 21:07 Training the Next Generation of Researchers 23:03 Interdisciplinary Learning in Statistics 24:26 Innovative Teaching Tools for Data Science 27:18 Workshops and Educational Opportunities 29:54 Reproducibility in Research 33:01 Leveraging Large Datasets for Insights 40:05 Challenges of Longitudinal Data Analysis

    47 min

  2. 28/12/2024

    Constrained PCA and the Future of Task-Based fMRI | Todd Woodward (Andy's BrainCast #004)

    Today's guest is Todd Woodward (https://psychiatry.ubc.ca/todd-woodward/), Professor at the Department of Psychiatry at the University of British Columbia, and also a Research Scientist at the British Columbia Mental Health and Substance Use Services. We talk about his development of Constrained PCA, a method for improving the reliability of task-based fMRI results. "Constrained PCA and Various Applications", by Hunter & Takane (2002): https://journals.sagepub.com/doi/pdf/10.3102/10769986027002105 Table of Contents ============== 00:00 Introduction to CPCA and fMRI Analysis 05:33 The Evolution of CPCA in fMRI Research 11:37 Understanding Cognitive Modes and Networks 17:06 Task-Based vs. Resting State Networks 23:03 The Importance of Shape in Brain Activation Patterns 29:01 Classification Challenges in fMRI Analysis 32:34 Understanding Correlation and Overlap in Brain Networks 36:38 The Role of Patterns in fMRI Data 41:39 Reverse Inference and Its Implications 49:21 The Impact of HRF on Multivoxel Pattern Analysis 55:41 Future Directions in Task-Based fMRI Research

    1h 3m

  3. 02/12/2024

    Neuroimaging, Reproducibility, and Neurodesk | Steffen Bollmann (Andy's BrainCast #003)

    Today's guest is Steffen Bollmann (https://about.uq.edu.au/experts/11402), Senior Research Fellow at the School of Electrical Engineering and Computer Science and the Queensland Digital Health Centre at The University of Queensland, Brisbane, Australia. We discuss how Neurodesk was developed, how it is currently used, and what its prospects are for making neuroimaging data analysis more accessible and reproducible, along with considerations of Jupyter notebooks and what technical training is required of today's researcher. Neurodesk Website: https://www.neurodesk.org/ Steffen providing a real-time demonstration of how to download and analyze a dataset, and how to register the analysis with the Open Science Framework: https://youtu.be/RTy5iVHHGO8?t=4030 Table of Contents ============== 0:00 Preview 0:34 How did Neurodesk Start? 6:35 Technical Architecture and Functionality of Neurodesk 13:19 Resource Management and Scalability Challenges 18:24 Funding and Future Prospects of Neurodesk 28:46 Security and Clinical Data Anonymity 33:00 Version Control and Software Compatibility 42:49 Jupyter Notebooks and Reproducibility 49:49 The Importance of Diverse Skill Sets 55:27 Community Involvement in Neurodesk Development

    58 min

  4. 16/10/2024

    Resting-State Analysis, Effect Size, and Power Analysis | Stephanie Noble (Andy's BrainCast #002)

    Today we talk with Dr. Stephanie Noble (https://neuroprismlab.github.io/), Assistant Professor at Northeastern University, in particular about resting-state analysis, her lab's development of effect size and power analysis tools, and the field moving away from localizationism. Links discussed during the interview: BioImage Suite: https://bioimagesuiteweb.github.io/webapp/ BrainEffex Effect Size Estimator: https://neuroprismlab.shinyapps.io/effect_size_shiny/ The CONN Toolbox: https://web.conn-toolbox.org/ Jonathan Power's Reading List: https://www.jonathanpower.net/reading-series.html Table of Contents ============== 0:00 Introduction 1:25 Reliability and Functional Connectivity 5:37 Budgeting for Resting-State Scans 8:35 How to learn Multivariate Methods 14:35 The Software Stephanie Uses for Her Analyses 16:34 Have Resting-State Scan Lengths Changed Recently? 22:00 Recommendations for Dealing with High-Moving Subjects 26:58 Clinical Applications of Resting-State Analysis 29:42 Individual Differences and Resting-State Results 34:53 Effect Size and Power Toolboxes Developed by Stephanie's Lab 42:12 Localizationism and Anti-Localizationism 48:52 How Would You Train Someone in Resting-State Analysis? 54:09 Trends in the Field about Revisiting the Basics

    57 min

  5. 25/08/2024

    Effect Magnitude, P-Values, and Statistical Power | Gang Chen (Andy's BrainCast #001)

    In this episode, we talk with Gang Chen, a mathematical statistician at the Scientific and Statistical Computing Core (SSCC) of the National Institutes of Health. Gang is one of the developers of the fMRI analysis package Analysis of Functional NeuroImages (AFNI: https://afni.nimh.nih.gov/). He has written about several topics related to statistics in neuroimaging, in particular his latest blog post about thresholding for statistical significance and what effects this may have on reproducibility: https://discuss.afni.nimh.nih.gov/t/enhancing-result-reporting-in-neuroimaging/7453 We talk about this and other topics, including statistical power, reproducibility, and whether the phrase "statistically significant" should be retired in favor of something else. Links to other papers discussed: -The Neuroimaging Analysis Replication and Prediction Study (NARPS, Botvinik-Nezer et al., 2020: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7771346/ -Vul et. al (2009), Puzzlingly High Correlations: https://tinyurl.com/yc26nuur -Eklun et. al (2016), Cluster Failure: https://www.pnas.org/doi/pdf/10.1073/pnas.1602413113 -Chen et. al (2017), Effect Sizes and P-Values: https://tinyurl.com/5eb7d6r3 -Taylor et. al (2023), Highlighting vs. Hiding Results: https://www.sciencedirect.com/science/article/pii/S1053811923002896 -Allen et. al (2012), Reporting Neuroimaging Results: https://www.cell.com/neuron/fulltext/S0896-6273(12)00428-X Table of Contents ============== 0:00 Introduction for Gang Chen 1:39 Effect Sizes (Magnitudes) and P-Values: What is the Difference? 10:48 Difficulty in Adopting New Reporting Standards: Beta Values and Percent Signal Change 13:13 Reviewing Previous Statistical "Crises" in Neuroimaging: Vul, Eklund, Bennet, and the NARPS Paper 21:43 Highlight vs. Hiding Results: Taylor et al. (2023) 25:46 The Problem with Dichotomizing Results: Reproducibility and P-Hacking 29:52 Statistical Power: Issues and Recommendations 38:38 Open-Access Datasets and Statistical Power 40:03 How Widespread is Requiring Power Analyses for Grants? 42:18 Meta-Analyses: Potential Problems with Reporting Peak T-Statistics 45:05 Pre-Registered Reports and Reproducibility 47:46 Should we Consider Replacing the Phrase, "Statistically Significant"? 54:45 Papers about Hierarchical Models 56:20 Concluding Remarks

    58 min
  • 5 Episodes

About

Andy talks to the leading scientists and software developers in the field of neuroimaging, learning about the latest trends, interesting findings, and the future of brain science.

Information

  • Creator
    Andrew Jahn
  • Years Active
    2024 - 2025
  • Episodes
    5