Current Topics in

Machine Learning for Neuroimaging

PSYC 121 / PSYC 221 | 2022 Fall

Course Information

  • Location : Turing Auditorium, Polya Building
  • Zoom : Available in Canvas
  • Time : 1:30pm - 2:50pm, M/W
  • Contact : psyc221-aut2223-staff@lists.stanford.edu
  • Office Hours : M 3-4pm Psychiatry Building Rm 3370, W 3-4pm Gates Building Rm 300
  • Description : The discovery of biological markers in medical applications is a fast-growing field. For this purpose, different experimental and neuroscientific procedures are incorporated to detect biological signatures and improve diagnosis or treatment of complex brain disorders. Neuroimaging is a discipline that studies the structure and function of the nervous system by means of imaging technology. In the recent years, machine and deep learning methods have revolutionized neuroimaging studies by enabling the development of imaging signatures of brain function and structure which can be detected at an individual level, and hence aid in developing personalized treatments. In this course, we explore the methodological gaps in analyzing high-dimensional, longitudinal, and heterogeneous clinical neuroscientific data and study novel, robust, scalable, and interpretable machine learning models for this purpose.
  • Instructors

     
     
     

    Course Assistants

     
     

    Course Advisor

     

    Syllabus

    Week Date Lecturer Topics Materials and Assignments
    Week 1 9/26 Mon Kilian Pohl
    Ehsan Adeli
    Qingyu Zhao
    Introduction
    • Course outline
    • Brain anatomy
    • Neuroimaging
    • Statistical modeling
    • Machine learning
    [Slides]
    9/28 Wed Kilian Pohl Basic operations and processing of neuroimages
    • Structural/diffusion/functional MRI processing
    • Quality control
    • Data curation for ML
    • Public datasets & course projects
    [Slides]
    Week 2 10/3 Mon Greg Zaharchuk
    (Stanford Radioloy)
    Understanding the brain through imaging
    10/5 Wed Qingyu Zhao Basics for statistical analysis
    • Hypothesis testing and group-wise analysis
    • Common statistical models for neuroimaging studies
    • What are confounders/confounding effects?
    • Statistical analysis of ML experiments
    [Slides TBD]
    [Assignment TBD] Due on 10/24
    Week 3 10/10 Mon Ehsan Adeli Basics for machine learning
    • Un/self/weak/full supervision
    • Evaluation & metrics of ML models
    • Traditional ML models
    • End-to-end deep learning (CNN, Transformers, generative models)
    • Example tasks and applications (classification/segmentation/registration)
    [Slides TBD]
    [Assignment TBD] Due on 10/26
    10/12 Wed Holger Roth
    (NVIDIA)
    Exploratory topics in medical and neuroimaging (tailored to help students finalize course projects)
    • Federated learning
    • Active learning
    • AutoML
    Week 4 10/17 Mon Magda Paschali
    Wei Peng
    ML/Stats programming with Python
    • Intro into Jupyter notebook
    • Hypothesis testing
    • Traditional ML models
    • Traditional ML based on image features
    • CNN on 3D MRIs
    [Slides TBD]
    10/19 Wed Ehsan Adeli Applications to structural MRI
    • ML based on ROI-wise features
    • Deep learning on 3D MRI volumes
    • MRI processing (segmentation/registration/denoising/etc)
    [Slides TBD]
    Week 5 10/24 Mon James Duncan
    (Yale University)
    Neuroimage Analysis in Autism: from Model-based Estimation to Data-driven learning
    10/26 Wed Midterm exam
    Project proposal
    Week 6 10/31 Mon Ehsan Adeli Exploratory topics on structural MRI
    • Generative models
    • Controlling confounding effects in deep learning
    [Slides TBD]
    11/2 Wed Ender Konukoglu
    (ETH Zürich)
    Applications to structural MRI
    • Historical view of how ML evolved for structural neuroimaging, covering univariate, multivariate ML, RF and DL models
    • What can current DL models (not) do for structural neuroimaging?
    • Challenges ahead
    Week 7 11/7 Mon Qingyu Zhao Applications to functional MRI
    • Correlation-based analysis
    • Independent component analysis
    • Auto-regressive models
    • Deep learning on 4D BOLD signals
    [Slides TBD]
    11/9 Wed Thomas Wolfers
    (University of Tübingen)
    From Estimating Activation Locality to Conceptualizing Disorder: Machine Learning for Brain Imaging Psychiatry
    Week 8 11/14 Mon Qingyu Zhao Applications to diffusion MRI
    • ML based on diffusivity measures
    • ML-based ODF estimation and tractography
    • Tensor-based convolution
    [Slides TBD]
    11/16 Wed Lauren O'donnell
    (Harvard University)
    Machine learning in diffusion MRI tractography
    Week 9 Thanksgiving No Classes
    Week 10 11/28 Mon Martin Styner
    (University of North Carolina at Chapel Hill)
    Machine Learning for predictive, longitudinal studies of infant MRI data
    • Image imputation/generation for missing infant MRI data in longitudinal studies
    • Age and contrast agnostic image segmentation for infant MRI
    • Classification/learning from cortical surface data for prediction in early brain development
    11/30 Wed Islem Rekik
    (Imperial College London)
    Exploratory topics
    Week 11 12/5 Mon Final exam
    Project presentation

    Resources

    Reading List

    Public Datasets

  • Human Connectome Project (HCP)
  • Alzheimer's Disease Neuroimaging Initiative: ADNI
  • Parkinson's Progression Markers Initiative (PPMI)
  • ABIDE - International Neuroimaging Data-sharing Initiative
  • OASIS Brains - Open Access Series of Imaging Studies
  • Multimodal Brain Tumor Segmentation Challenge 2020 (BraTS)