{\rtf1\ansi\ansicpg1252\cocoartf1561\cocoasubrtf400
{\fonttbl\f0\fswiss\fcharset0 Helvetica;\f1\fnil\fcharset0 Monaco;}
{\colortbl;\red255\green255\blue255;\red0\green0\blue0;\red255\green255\blue255;}
{\*\expandedcolortbl;;\cssrgb\c0\c0\c0;\cssrgb\c100000\c100000\c100000;}
\margl1440\margr1440\vieww22860\viewh17480\viewkind0
\pard\tx720\tx1440\tx2160\tx2880\tx3600\tx4320\tx5040\tx5760\tx6480\tx7200\tx7920\tx8640\pardirnatural\partightenfactor0
\f0\b\fs28 \cf0 Analysis pipeline for PD/ET comparison\
(last edited 3/25/2018)
\fs24 \
\b0 \
\b\fs28 Overview
\b0\fs24 \
All analysis done in MATLAB\
\
\b Dependencies/Code\
\pard\tx560\tx1120\tx1680\tx2240\tx2800\tx3360\tx3920\tx4480\tx5040\tx5600\tx6160\tx6720\pardirnatural\partightenfactor0
\b0 \cf2 \cb3 \CocoaLigature0 Code directory:
\f1 Nexus/Users/pwjones/code/SpeechPilotPD-ET
\f0\b \cf0 \cb1 \CocoaLigature1 \
\pard\tx720\tx1440\tx2160\tx2880\tx3600\tx4320\tx5040\tx5760\tx6480\tx7200\tx7920\tx8640\pardirnatural\partightenfactor0
\b0 \cf0 \
All code is versioned on GITHUB\
Main repository is https://github.com/Brain-Modulation-Lab/SpeechPilotPD-ET.git\
There possibly some dependent files in:\
https://github.com/Brain-Modulation-Lab/SpeechSpikeAnalysis.git\
https://github.com/pwjones/matlab_common.git \
I will work to remove dependencies on these other two projects so that all of the necessary code is self-contained.\
\b \
Setting up Instructions\
\b0 - You should clone the repository and use git to update it as you develop/change analyses so that everything stays synched and current in the online repository.\
- Important note on setting up paths: Many of the scripts read or write files in a way that I\'92ve set up to depend on paths that I\'92ve defined once in an m file setDirectories.m. The default lives in the main directory with all of the code, but other uses who may need different paths should make a copy that lives in their default matlab code directory that contains directory specifications that work for their systems. \
\
\
\b\fs28 Data Processing/Analysis Pipeline steps
\b0\fs24 \
\b 1)
\b0 Raw data was processed by Witek and Ahmad, ends up on Dyonisis in a file that looks like (paths are gonna be Mac formatted here but obviously may look different to you).\
\pard\tx560\tx1120\tx1680\tx2240\tx2800\tx3360\tx3920\tx4480\tx5040\tx5600\tx6160\tx6720\pardirnatural\partightenfactor0
\f1 \cf2 \cb3 \CocoaLigature0 /Volumes/Nexus/Electrophysiology_Data/DBS_Intraop_Recordings/DBS2001/Preprocessed\\ Data/DBS2001_Session1_LT.mat \
\f0 There is a single processed data file for each session, and these are the base of what the functions I\'92ve written work on. \
All data processing pipeline scripts start with these data files and refer to their original locations on the lab server. \
\
\pard\tx560\tx1120\tx1680\tx2240\tx2800\tx3360\tx3920\tx4480\tx5040\tx5600\tx6160\tx6720\pardirnatural\partightenfactor0
\b \cf2 \cb3 2)
\b0 \cb3 The first step: Calculating necessary spectral filtering for each experiment and saving the result. \
\b \cb3 Script
\b0 \cb3 : The_behemoth2pwj.m
\b \cb3 \
\pard\tx560\tx1120\tx1680\tx2240\tx2800\tx3360\tx3920\tx4480\tx5040\tx5600\tx6160\tx6720\pardirnatural\partightenfactor0
\b0 \cf2 \cb3 Goes through each of the subjects in one of the cohorts and compiles the necessary data for further analysis into a Results structure that is a struct array with the length being the number of sessions for that subject group. Basically it does the spectral filtering on the ECoG signal for each subject. This is a fairly computationally intensive step and can take several hours to run on a cohort. These files are saved to ones that look like\
\f1 /Volumes/Nexus/Users/pwjones/code/DataFiles/Band_modulation_referenced_PD_v5.mat
\f0 (the most recent PD cohort file, 44.3GB)\
\pard\tx720\tx1440\tx2160\tx2880\tx3600\tx4320\tx5040\tx5760\tx6480\tx7200\tx7920\tx8640\pardirnatural\partightenfactor0
\cf2 \cb3 These files are pretty big (at least 10s of GB) and are slow to load into matlab so some further population analysis works on smaller population data subsets. Also, to handle they need one of the servers that have enough memory for them to be loaded into.\
\pard\tx560\tx1120\tx1680\tx2240\tx2800\tx3360\tx3920\tx4480\tx5040\tx5600\tx6160\tx6720\pardirnatural\partightenfactor0
\b \cf2 Output
\b0 \cb3 :
\i \cb3 Results
\i0 \cb3 structure array: fields Base, Cue, Onset, Session, trials. One element for each session. \
Cue/Onset - fields that contain either cue-aligned or speech onset aligned data. Contains fields for each spectral band (eg \'91beta1\'92), with fields
\i \cb3 z_Amp, tr, bs
\i0 \cb3 (time x channel/trial interleaved, all channels for trial 1, all channels for trial 2, and so on). \
\i \cb3 z_Amp
\i0 \cb3 - zscored data,
\i \cb3 tr
\i0 \cb3 - band passed signal,
\i \cb3 bs
\i0 \cb3 - baseline period for same trials.\
\
\b \cb3 3)
\b0 \cb3 Compile large structure into smaller population data structures that are easier to work with and quicker to load.\
\b \cb3 Script:
\b0 \cb3 compileGroupPopulationData.m. \
\b \cb3 Input:
\b0 \cb3 \'91Results\'92 data structure - loaded into the workspace\
Additional instructions: Specify at top of script whether running on ET or PD dataset. \
\b \cb3 Output:
\b0 \cb3 It will save file such as
\b \cb3 ET_populationAvgs_Loc2.mat .
\b0 \cb3 This file contains the \'91PopResults\'92 structure, which will have trial averaged mean traces, etc, instead of all of the single trial information. \
\
\b \cb3 4)
\b0 \cb3 Those last data files are read and plotted by a number of scripts, each performing a different set of analyses:\
\
\b \cb3 Analysis
\b0 \cb3 : Comparison of Signal strength in specific spectral bands between PD and ET subjects\
\b \cb3 Script:
\b0 \cb3 popPD_ET_BandComparison.m\
What it does: Compares the PD/ET signals at population level. Does the statistical comparison and generates plots.\
\b \cb3 Output
\b0 \cb3 : Figures for each spectral band, for each alignment (Cue aligned and Speech onset aligned). \
\b \cb3 \
\
}