parse_spm_mvt_file.py

# -*- coding: utf-8 -*-
"""
A tool that searches and parses rp_* spm files computed from the re-alignment.
Data are saved in a csv file which contains:
    - 3 translations
    - 3 rotations
    - 3 quadratic translations
    - 3 cubic translations
    - 3 translations shifted 1 TR before
    - 3 translations shifted 1 TR after

@author: Mehdi Rahim
"""

import os, glob
import numpy as np
import pandas as pd

BASE_DIR = os.path.join('/', 'shfj', 'Ppsypim', 'CAIMAN', 'Subjects')
BASE_DIR = os.path.join('movement_files_caiman')
DST_BASE_DIR = os.path.join('movement_files_caiman')

def parse_spm_mvt_file(filename):
    """ returns a DataFrame df which contains the movement regressors
    """
    df = pd.read_csv(filename, sep='  ',
                     names=['t_x', 't_y', 't_z', 'r_x', 'r_y', 'r_z'])

    # Quadratic
    df['t_x_2'] = df['t_x']*df['t_x']
    df['t_y_2'] = df['t_y']*df['t_y']
    df['t_z_2'] = df['t_z']*df['t_z']

    # Cubic
    df['t_x_3'] = df['t_x']*df['t_x']*df['t_x']
    df['t_y_3'] = df['t_y']*df['t_y']*df['t_y']
    df['t_z_3'] = df['t_z']*df['t_z']*df['t_z']

    # Shift +1
    df['t_x_s1'] = np.roll(df['t_x'], 1, axis=0)
    df['t_y_s1'] = np.roll(df['t_y'], 1, axis=0)
    df['t_z_s1'] = np.roll(df['t_z'], 1, axis=0)

    # Shift -1
    df['t_x_s2'] = np.roll(df['t_x'], -1, axis=0)
    df['t_y_s2'] = np.roll(df['t_y'], -1, axis=0)
    df['t_z_s2'] = np.roll(df['t_z'], -1, axis=0)

    return df

##############################################################################
""" Parsing all the subjects and saving the regressors in a csv file
"""



rp_files = glob.glob(os.path.join(BASE_DIR, '*.txt'))

for rp_file in rp_files:
    _, filename = os.path.split(rp_file)
    subject = filename.split('_')[1][1:]
    print subject
    d = parse_spm_mvt_file(rp_file)
    d.to_csv(os.path.join(DST_BASE_DIR,
                          '_'.join([subject, 'reg.csv'])), sep=',')