Encoding

Encode text from Rubin et al. (2017) into images.

Start with the necessary imports

from os.path import join

from nilearn import plotting

from gclda.model import Model
from gclda.decode import encode
from gclda.utils import get_resource_path

Out:

Bootstrapped meta-analyses are enabled.

Load model

model_file = join(get_resource_path(), 'models/Neurosynth2015Filtered2',
                  'model_200topics_2015Filtered2_10000iters.pklz')
model = Model.load(model_file)
model.display_model_summary()

Out:

--- Model Summary ---
 Current State:
         Current Iteration   = 10000
         Initialization Seed = 1
         Current Log-Likely  = -11268037.3695
 Model Hyper-Parameters:
         Symmetric = True
         n_topics  = 200
         n_regions = 2
         alpha     = 0.100
         beta      = 0.010
         gamma     = 0.010
         delta     = 1.000
         roi_size  = 50.000
         dobs      = 25
 Model Training-Data Information:
         Dataset Label                 = Neurosynth2015Filtered2
         Word-Tokens (n_word_tokens)   = 520492
         Peak-Tokens (n_peak_tokens)   = 400801
         Word-Types (n_word_labels)    = 6755
         Documents (n_docs)            = 11362
         Peak-Dimensions (n_peak_dims) = 3

First text

text = 'motor'
text_img, _ = encode(model, text)
fig = plotting.plot_stat_map(text_img, display_mode='z',
                             threshold=0.00001,
                             cut_coords=[-18, 4, 32, 60])

Second text

text = 'effort difficult demands'
text_img, _ = encode(model, text)
fig = plotting.plot_stat_map(text_img, display_mode='z',
                             threshold=0.00001,
                             cut_coords=[-30, -4, 26, 50])

Third text

text = 'painful stimulation during a language task'
text_img, _ = encode(model, text)
fig = plotting.plot_stat_map(text_img, display_mode='z',
                             threshold=0.00001,
                             cut_coords=[-2, 22, 44, 66])

Total running time of the script: ( 0 minutes 59.833 seconds)