-
Notifications
You must be signed in to change notification settings - Fork 6
Home
Welcome to the project-iota wiki!
Our Roadmap: Project Group Iota
Data:
(1) We have 4 groups of participants: schizophrenia patients, patients' siblings, normal people, and normal people's siblings.
(2) We have 4 tasks: R, 1B, 2B, 3B.
(3) We have 4 ROIs: FP, CO, CER, DMN.
Original Paper Objectives:
- Within-network connectivity:
(a) For each participant, find connectivity within each ROI on different tasks.
2.Between-network connectivity:connectivity:
(a) For each participant, find connectivity between each ROI on different tasks.
Our Project Objectives: 1.Simplified assumptions:
(a) Since we do not have the knowledge about partition the brain into ROIs, we will simply focus on the entire brain (We can ask Matthew for help for this part).
(b) Since we do not have the knowledge of comparing different subjects who have different brain shapes, we will simply visually compare brain images and approximate.
- We will examine two subjects, one subject with schizophrenia and one healthy subject.
General to do list:
-
Get convolved time course and save as convolved.txt.
-
Pick one subjects and go full length analysis. Or pick one Schizophrenic participant and one healthy participant.
-
Make design matrix using convolved.txt and run regression on BOLD signals to get the estimated betas. (Note: the betas are the scaling of the BOLD signals in response to the stimulus.) We can do t-tests on the betas (However, keep in mind the assumptions we use.) and see which of the betas are significant.
-
Find correlations between average voxels in each ROI (Before we figure out ROIs, we can try to find correlation between each voxels as from lecture 12, Correlation per voxel, in 2D). Measure mean signal. Make a histogram of the values of correlations (values from 0 to 1), and then determine a cutoff threshold to determine which of the correlations are significant.
-
Use Graphical Lasso to visualize covariance and precision of data.
-
Make convolved.txt using notes from lecture Day 13 (reference: http://www.jarrodmillman.com/rcsds/lectures/convolution_background.html).
-
Make design matrix using convolved.txt and run regression on 2D BOLD signals. Question: what are we regressing on? -- We are regressing the BOLD signals with the convolved time course. If the betas are large, then the BOLD signal is strong in response to the stimulus. (reference: see introduction of this about convolution: http://practical-neuroimaging.github.io/on_convolution.html)
-
Make progress slides for 11/12 progress presentation.
-
Go to office hours on Friday morning 11/6 for project questions.
-
Project-related questions will be asked on Gitter.
####Task Assignment:
-Zeyu is responsible for making weekly plans, and assign tasks. Will be helping with other teammates with code writing or reviewing as well as slides making.
-Jie and Qingyuan are responsible for writing code for convolved time course. If there are questions, ask Zeyu or Yunchuan. -Yunchuan is responsible for making the progress presentation report.