/Brainhack2020

Primary LanguagePython

LNCD PET tat2

Data

  • data/wide.csv is a wide (547 column) format CSV with a row per participant visit. The columns include individual measures for various surveys (behavioral) and brain (MR and PET scan) data.

  • data/mtr_striatum_258_QC.csv - MTR ROI mean and voxel coverage count ("NZmean") and QC (manual quality check binarized 1=pass, 0=fail)

  • data/conn_adj*.csv row per visit CSV files. A file per MR task type: rest1, task background, rest2. Columns are the widened (2d to 1d) upper triangle of each visits fisher-z transformed correlation adjacency matrix. This is the mean BOLD time series with each of 23 ROIS correlated to one another ( 23 choose 2 = 253). Fixation from 6 consecutive task aquistions are extracted and combined to make the "background" set.

    • also see data/conn_*.pkl for pre-made (and untested) pyradigm data structure: reloaded = RegrDataset('data/conn_rest1.pkl')

images

per subject tat2 3d nii.gz images are on google dirve: https://drive.google.com/drive/folders/1WuCqimxEBnrb8MkqXawTKLGZi7frYcy1?usp=sharing

wide.csv

prefix desc
tat2* time average T2* (proxy for iron?)
frogET* anti saccade task performance (esp correct: ncor, corlat, corsd for reward and neutral)
RT18.Score Risk taking measure (https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3160867/)
UPPSP* impulsivity (http://www.impulsivity.org/measurement/UPPS_P )
ysrasr.* DSM-oriented scales (http://fcon_1000.projects.nitrc.org/indi/enhanced/assessments/asr.html)

Neuropredict

python3 -m pip install neuropredict pyradigm
np_regress -y data/conn_bkgrd_subset.pkl --impute_strategy most_frequent -e randomforestregressor

Code

pet_data.py quick attempt at building pyradigm data structures

from pyradigm import RegressionDataset as RegrDataset
from pet_data import PET
pet = PET('data/wide.csv')
# see pet.widedf.shape: 384 visits with 6,987 measures

##  get sesid, age, and all the uppsp measures

# pyradigm
upps = RegrDataset()
upps.description = "Urgency, Premeditation (lack of), Perseverance (lack of), Sensation Seeking, Positive Urgency, Impulsive Behavior Scale"
pet.add_subset_to(upps, '^uppsp_') 
# Urgency, Premeditation (lack of), Perseverance (lack of), Sensation Seeking, Positive Urgency, Impulsive Behavior Scale 
# 337 samplets, 6 features

# or as dataframe
upps_df = pet.col_subset('^uppsp_')
upps_df.columns # ['sesid', 'age', 'upps_pre', 'upps_pers', 'upps_ss', 'upps_pu', 'upps_tot', 'upps_negurg']

# also include other confound columns (works same for add_subest_to and col_subset)
upps_df = pet.col_subset('^sex|^uppsp_', 'uppsp_upps')
upps_df.columns # ['sesid', 'age', 'sex', 'pre', ...]

Glossary

  • lunaid unique identifier for a participant. Repeats across visits
  • visitnum longitudinal study. We expect 3 distinct visits. Not every participant finishes.
  • sessionid/sesid session identifier unique for visit year. Behavioral and scan may happen on separate days. These will still have the same sesid.
  • vdate MR visit date, behavedate date of behavioral battery (RT18, UPPS, YSR/ASR), dtbzdate may have returned if the scanner portion did not complete. this is the date MR was resumed. few have separate vdate and dtbzdate.
  • ROI brain region of interest. Anatomical or functionally defined region composed of many voxels. Voxel values are often averaged within this region.
  • MR
  • taT2 time averaged T2* - average of all the BOLD signal time course.
  • R2' R2 prime
  • MTR magnetic transfer ratio
  • BOLD
  • ASY/YSR adult/youth self report. 2 sets of questions, depending on age of participant.
  • PET
  • DTBZ PET radio tracker administered second. vessicular monoamine transpoter/DA
  • Rac Raclopride PET radio tracker administered first