Vault

Statistical-Neuroimaging-Analysis

3 min read

Dataview:

list from [[]] and !outgoing([[]])

📗 -> Statistical Neuroimaging Analysis: An overview

🎤 Vocab

Neuroimaging = NI
Neuroimaging analysis (NIA): Used for recording the brain

  • Neurological disorders
  • Aging
  • BCI

❗ Information

Lexin Li
Prof at UC Berkeley - https://lexinli.biostat.berkeley.edu/

📷 -> Pics






📄 -> Methodology

  • Simple or full description

✒️ -> Talk scratch notes

  • NIA: Very hard, introduces many problems
    • Tensor data, ODE, point process
    • Brain networks, graphical models
    • Causal inference, multimodal analysis
  • Large NI studies, and public databases are becoming available

Imaging Modalities:

  • MRI< fMRI, PET, EEG, electrocorticography (ECoG), diffusion tensor imaging (DTI), neuronal spike trains…

Topics:

  • Imaging Tensor Analysis: Understanding individual brain regions
    • Look at MRI image of patient, tell if they have a disorder
  • Brain Connectivity Network Analysis: How do networks interact? How are they associated with outcomes?
  • Multimodal NIA: How different ttypes of images are associated with each other, and jointly affect the outcomes?
  • NI Causal Inference: Can we get “causal conclusions?”
  • New Imaging Modalities and Technologies: picture in phone…

🧪-> Case Studies

Case Study 1:

Amyloid-beta and Tau are closely associated in terms of spatial accumulations, and association patterns are believed to be affected with age

  • Potentially associated with Alzheimers too?

Looking to study association in concentration and location
Association between modalities: (protein 1), , protein 2, , demographic info (age)
Each modality can be high dimensional

Can rephrase to seeking linear combinations of X, Y and Z are the most contrastive (contrastive?)

Literature Review: CHECK PHONE FOR BETTER DESCRIPTION

  • Canonical Correlation Analysis
  • Sufficient Dimension Reduction
  • Liquid Association

They suggest the generalized liquid association (GLA) as the solution

Tensor Tucker Decompization problem?

ADNI Problem Studies:

  • n=81
  • modality X: amyloid-beta deposition …
    took a picture

Case Study 2:Kernel ODE?

Brain Effective Connectivity Analysis: Uncover the directional influence that one neural system exerts over another

  • Can be applied to: gene regulatory network analysis based on time-course gene expression data
    • Also diabetes study / artificial pancreas

ECoG study of brain during decision-making (DM)
Electrocorticography data: electrode x time data matrix
Patient performed gambling tasks with different levels of risk (high risk vs low rish)
Electrodes placed in the OFC: 61 electrodes over 3001 time points

Take results with caution

Experiment is done with 1 participant, over multiple trials. Generalization of results should be done with caution, only really applicable to the 1 participant, who was chosen because of neurosurgery anyway

Regulatory Functionals
Special Cases with assumptions, not the best approach but good to know

  • Linear ODE
  • Additive ODE
  • Neural ODE?

The data analysis (took a photo), show that different parts of OFC take priority in spreading activation

  • Posterior OFC tends to actively influence other nodes during low risk games, reward is simple and clear
  • Anterior OFC tends to actively influence other nodes during high risk games, where they involve more calculation

Case Study 3:

Skipped through, low on time
Took a picture of results

🔗 -> Related Word

Graph View

Backlinks