Decoding the Principal Gradient

Gradec provides .

# Download HCP S1200 principal gradient
from gradec.fetcher import _fetch_gradients

gradients = _fetch_gradients(cortical=True)

1D Segmentation

Select and plot the principal gradient

from surfplot.utils import add_fslr_medial_wall

from gradec.plot import plot_surf_maps

principal_gradient = gradients[:, 0]

full_vertices = 64984
hemi_vertices = full_vertices // 2

prin_grad = add_fslr_medial_wall(principal_gradient)  # Add medial wall for plotting
prin_grad_lh, prin_grad_rh = prin_grad[:hemi_vertices], prin_grad[hemi_vertices:full_vertices]

fig = plot_surf_maps(prin_grad_lh, prin_grad_rh)
fig.show()

KMeans-based segmentation

from gradec.segmentation import KMeansSegmentation

segmentation = KMeansSegmentation(n_segments=2)
segmentation.fit(principal_gradient)
grad_maps = segmentation.transform()

1D Decoding

from gradec.decode import LDADecoder
from gradec.fetcher import _fetch_classification, _fetch_features, _fetch_frequencies
from gradec.plot import plot_cloud
from gradec.utils import _decoding_filter

SPACE, DENSITY = "fsLR", "32k"
DSET, MODEL = "neuroquery", "lda"

# Fit the decoder
decode = LDADecoder(space=SPACE, density=DENSITY, calc_pvals=False)
decode.fit(DSET)

# Load features for visualization
features = _fetch_features(DSET, MODEL)
frequencies = _fetch_frequencies(DSET, MODEL)
classification, class_lst = _fetch_classification(DSET, MODEL)

Downloading data from https://osf.io/k4xza/download ...
 ...done. (1 seconds, 0 min)
Downloading data from https://osf.io/u68w7/download ...
 ...done. (5 seconds, 0 min)
Downloading data from https://osf.io/u68w7/download ...
 ...done. (3 seconds, 0 min)
Downloading data from https://osf.io/ygpxf/download ...
 ...done. (4 seconds, 0 min)
Downloading data from https://osf.io/mtwvc/download ...
 ...done. (3 seconds, 0 min)

Decode map

corrs_df = decode.transform(grad_maps, method="correlation")

0%|          | 0/2 [00:00<?, ?it/s]
0%|          | 0/2 [00:00<?, ?it/s]

Visualize results

for map_i, (grad_map, segment) in enumerate(zip(grad_maps, corrs_df.columns)):
    seg_df = corrs_df[[segment]]
    # Filter results
    filtered_df, filtered_features, filtered_frequencies = _decoding_filter(
        seg_df,
        features,
        classification,
        freq_by_topic=frequencies,
        class_by_topic=class_lst,
    )
    filtered_df.columns = ["r"]
    corrs = filtered_df["r"].to_numpy()

    # Gradient map plot
    grad_map = add_fslr_medial_wall(grad_map)
    grad_map_lh, grad_map_rh = (
        grad_map[:hemi_vertices],
        grad_map[hemi_vertices:full_vertices],
    )
    fig = plot_surf_maps(
        grad_map_lh,
        grad_map_rh,
        color_range=(0, 1),
        cmap="YlOrRd",
        title=f"KMeans, segment {map_i+1}",
    )
    fig.show()

    # Word cloud plot
    cloud_fig = plot_cloud(
        corrs,
        filtered_features,
        MODEL,
        frequencies=filtered_frequencies,
    )
    cloud_fig.show()

• 
• 
• 
•