Brain-hacking 101

Author: Ariel Rokem, The University of Washington eScience Institute

Hack 4: interact with the data

To get a sense of your data, one of the best things you can do is to interactively explore the patterns in your data. While building full-fledged interactive applications that do more than one thing is rather hard, it is possible to build small interactive data-exploration tools, that do just one thing, with only a few lines of code. Here, we'll show how to do that using IPython's interactive widget system. We will demonstrate this below



In [1]:

    
import numpy as np
import nibabel as nib
import matplotlib as mpl
import matplotlib.pyplot as plt
%matplotlib inline
mpl.style.use('bmh')



In [2]:

    
img = nib.load('./data/run1.nii.gz')
data = img.get_data()



In [3]:

    
import scipy.signal as sps



In [4]:

    
tsnr = np.mean(data, -1) / np.std(data, -1)



In [8]:

    
def plot_tsnr(x=data.shape[0]/2, y=data.shape[1]/2, z=data.shape[2]/2):
    fig, axes = plt.subplots(2, 2)
    ax = axes[0, 0]
    ax.axis('off')
    ax.matshow(tsnr[:, :, z], cmap=mpl.cm.hot)
    ax = axes[0, 1]
    ax.axis('off')
    ax.matshow(np.rot90(tsnr[:, y, :]), cmap=mpl.cm.hot)
    ax = axes[1, 0]
    ax.axis('off')
    ax.matshow(np.rot90(tsnr[x, :, :]), cmap=mpl.cm.hot)
    ax = axes[1, 1]
    ax.plot(data[x, y, z])
    ax.set_xlabel('Time')
    ax.set_ylabel('FMRI signal  (a.u.)')
    fig.set_size_inches(10, 10)
    return fig



In [9]:

    
import IPython.html.widgets as wdg
import IPython.display as display



In [10]:

    
pb_widget = wdg.interactive(plot_tsnr, 
                            x=wdg.IntSliderWidget(min=1, max=data.shape[0], value=data.shape[0]//2),
                            y=wdg.IntSliderWidget(min=1, max=data.shape[1], value=data.shape[1]//2),
                            z=wdg.IntSliderWidget(min=1, max=data.shape[2], value=data.shape[2]//2)
                            )
display.display(pb_widget)



In [ ]: