Quickstart

Loading subject data

In [1]:

from bfm.subject.braintreebank import BrainTreebankSubject

subject_id, trial_id = 1, 1
subject = BrainTreebankSubject(subject_id, cache=False)
subject.load_neural_data(trial_id)

electrode_labels = subject.get_electrode_labels()
print(f"Subject {subject.subject_identifier} has {len(electrode_labels)} electrodes at sampling rate {subject.get_sampling_rate()} Hz")
print(f"Electrode labels: {electrode_labels}\n")

example_all_electrode_data = subject.get_all_electrode_data(trial_id, window_from=0, window_to=4096)
print(f"Example all electrode data shape: {example_all_electrode_data.shape}")
print(f"Example all electrode data: {example_all_electrode_data}\n")

from bfm.subject.braintreebank import BrainTreebankSubject subject_id, trial_id = 1, 1 subject = BrainTreebankSubject(subject_id, cache=False) subject.load_neural_data(trial_id) electrode_labels = subject.get_electrode_labels() print(f"Subject {subject.subject_identifier} has {len(electrode_labels)} electrodes at sampling rate {subject.get_sampling_rate()} Hz") print(f"Electrode labels: {electrode_labels}\n") example_all_electrode_data = subject.get_all_electrode_data(trial_id, window_from=0, window_to=4096) print(f"Example all electrode data shape: {example_all_electrode_data.shape}") print(f"Example all electrode data: {example_all_electrode_data}\n")

Subject btbank1 has 130 electrodes at sampling rate 2048 Hz
Electrode labels: ['F3aOFa2', 'F3aOFa3', 'F3aOFa4', 'F3aOFa7', 'F3aOFa8', 'F3aOFa9', 'F3aOFa10', 'F3aOFa11', 'F3aOFa12', 'F3aOFa13', 'F3aOFa14', 'F3aOFa15', 'F3aOFa16', 'F3bIaOFb1', 'F3bIaOFb2', 'F3bIaOFb3', 'F3bIaOFb4', 'F3bIaOFb5', 'F3bIaOFb6', 'F3bIaOFb7', 'F3bIaOFb8', 'F3bIaOFb9', 'F3bIaOFb10', 'F3bIaOFb11', 'F3bIaOFb12', 'F3bIaOFb13', 'F3bIaOFb14', 'F3bIaOFb15', 'F3bIaOFb16', 'F3cId1', 'F3cId2', 'F3cId3', 'F3cId4', 'F3cId5', 'F3cId6', 'F3cId7', 'F3cId8', 'F3cId9', 'F3cId10', 'T1aIb1', 'T1aIb2', 'T1aIb3', 'T1aIb4', 'T1aIb5', 'T1aIb6', 'T1aIb7', 'T1aIb8', 'T2aA1', 'T2aA2', 'T2aA3', 'T2aA4', 'T2aA5', 'T2aA6', 'T2aA7', 'T2aA8', 'T2aA9', 'T2aA10', 'T2aA11', 'T2aA12', 'T2bHa1', 'T2bHa3', 'T2bHa4', 'T2bHa5', 'T2bHa7', 'T2bHa8', 'T2bHa9', 'T2bHa10', 'T2bHa11', 'T2bHa12', 'T2bHa13', 'T2bHa14', 'T1bIc1', 'T1bIc2', 'T1bIc3', 'T1bIc4', 'T1bIc5', 'T1bIc6', 'T1bIc7', 'T1bIc8', 'F3dIe1', 'F3dIe2', 'F3dIe3', 'F3dIe4', 'F3dIe5', 'F3dIe6', 'F3dIe7', 'F3dIe8', 'F3dIe9', 'F3dIe10', 'F3dIe14', 'T3aHb6', 'T3aHb9', 'T3aHb10', 'T3aHb12', 'T1cIf1', 'T1cIf2', 'T1cIf3', 'T1cIf4', 'T1cIf5', 'T1cIf6', 'T1cIf7', 'T1cIf8', 'T1cIf10', 'T1cIf11', 'T1cIf12', 'T1cIf13', 'T1cIf14', 'T1cIf15', 'T1cIf16', 'T2c4', 'T2c5', 'T2c6', 'T2c7', 'T2c8', 'T3bOT1', 'T3bOT2', 'T3bOT3', 'T3bOT4', 'T3bOT5', 'T3bOT6', 'T3bOT8', 'T3bOT9', 'T3bOT10', 'T3bOT12', 'T2d1', 'T2d2', 'T2d3', 'T2d4', 'T2d5', 'T2d6']

Example all electrode data shape: torch.Size([130, 4096])
Example all electrode data: tensor([[-24.7234, -25.7867, -25.5209,  ..., -61.4097, -64.3339, -71.5117],
        [ 22.0649,  19.6724,  21.2674,  ..., -26.3184, -31.1036, -33.7620],
        [135.3139, 137.4407, 140.0991,  ...,  47.0542,  45.4591,  38.8130],
        ...,
        [  9.3045,  13.5580,  15.4189,  ..., -68.3216, -62.4730, -67.5240],
        [ -4.5193,   0.7975,   1.3292,  ..., -68.0557, -64.0681, -70.1825],
        [-27.1160, -22.3308, -19.9382,  ..., -33.2303, -31.3694, -35.6229]])

In [ ]:

electrode_locations = {electrode_label: subject.get_electrode_metadata(electrode_label)['DesikanKilliany'] for electrode_label in subject.get_electrode_labels()}
amygdala_electrode_locations = {label: location for label, location in electrode_locations.items() if 'amygdala' in location.lower()}
amygdala_electrode_locations

electrode_locations = {electrode_label: subject.get_electrode_metadata(electrode_label)['DesikanKilliany'] for electrode_label in subject.get_electrode_labels()} amygdala_electrode_locations = {label: location for label, location in electrode_locations.items() if 'amygdala' in location.lower()} amygdala_electrode_locations

Out[ ]:

{'T2aA1': 'Right-Amygdala',
 'T2aA2': 'Right-Amygdala',
 'T2aA3': 'Right-Amygdala',
 'T2aA5': 'Right-Amygdala'}

In [3]:

Get the neural data index for a given subject, trial, and movie time

Get the neural data index for a given subject, trial, and movie time

Out[3]:

{'T2aA1': 'Right-Amygdala',
 'T2aA2': 'Right-Amygdala',
 'T2aA3': 'Right-Amygdala',
 'T2aA5': 'Right-Amygdala'}

In [4]:

# Get the neural data index for a given subject, trial, and movie time

from bfm.evaluation.neuroprobe.config import ROOT_DIR, SAMPLING_RATE
import os
import pandas as pd
import numpy as np

def obtain_neural_data_index(sub_id, trial_id, movie_times):
    # Data frames column IDs
    start_col, end_col = 'start', 'end'
    trig_time_col, trig_idx_col, est_idx_col, est_end_idx_col = 'movie_time', 'index', 'est_idx', 'est_end_idx'

    # Path to trigger times csv file
    trigger_times_file = os.path.join(ROOT_DIR, f'subject_timings/sub_{sub_id}_trial{trial_id:03}_timings.csv')

    trigs_df = pd.read_csv(trigger_times_file)
    display(trigs_df.head())
    print(trigs_df['type'].unique())

    last_t = trigs_df[trig_time_col].iloc[-1]
    assert np.all(movie_times < last_t), "Movie times must be less than the last trigger time"
    
    # Vectorized nearest trigger finding
    start_indices = np.searchsorted(trigs_df[trig_time_col].values, movie_times)
    start_indices = np.maximum(start_indices, 0) # handle the edge case where movie starts right at the word
    
    # Vectorized sample index calculation
    return np.round(
        trigs_df.loc[start_indices, trig_idx_col].values + 
        (movie_times - trigs_df.loc[start_indices, trig_time_col].values) * SAMPLING_RATE
    ).astype(int)

# subject 10, trial 0, movie "Cars 2"
print("For subject 10, trial 0, movie 'Cars 2'")
print(obtain_neural_data_index(10, 0, np.array([0, 1000])))

# subject 7, trial 0, movie "Cars 2"
print("For subject 7, trial 0, movie 'Cars 2'")
print(obtain_neural_data_index(7, 0, np.array([0, 1000])))

# Get the neural data index for a given subject, trial, and movie time from bfm.evaluation.neuroprobe.config import ROOT_DIR, SAMPLING_RATE import os import pandas as pd import numpy as np def obtain_neural_data_index(sub_id, trial_id, movie_times): # Data frames column IDs start_col, end_col = 'start', 'end' trig_time_col, trig_idx_col, est_idx_col, est_end_idx_col = 'movie_time', 'index', 'est_idx', 'est_end_idx' # Path to trigger times csv file trigger_times_file = os.path.join(ROOT_DIR, f'subject_timings/sub_{sub_id}_trial{trial_id:03}_timings.csv') trigs_df = pd.read_csv(trigger_times_file) display(trigs_df.head()) print(trigs_df['type'].unique()) last_t = trigs_df[trig_time_col].iloc[-1] assert np.all(movie_times < last_t), "Movie times must be less than the last trigger time" # Vectorized nearest trigger finding start_indices = np.searchsorted(trigs_df[trig_time_col].values, movie_times) start_indices = np.maximum(start_indices, 0) # handle the edge case where movie starts right at the word # Vectorized sample index calculation return np.round( trigs_df.loc[start_indices, trig_idx_col].values + (movie_times - trigs_df.loc[start_indices, trig_time_col].values) * SAMPLING_RATE ).astype(int) # subject 10, trial 0, movie "Cars 2" print("For subject 10, trial 0, movie 'Cars 2'") print(obtain_neural_data_index(10, 0, np.array([0, 1000]))) # subject 7, trial 0, movie "Cars 2" print("For subject 7, trial 0, movie 'Cars 2'") print(obtain_neural_data_index(7, 0, np.array([0, 1000])))

For subject 10, trial 0, movie 'Cars 2'

	type	movie_time	start_time	end_time	trig_type	index	diff
0	beginning	0.000000	211276.153251	211276.397999	manual	148150	622
1	trigger	0.083417	211276.456513	211276.473401	high_real	148772	339
2	trigger	0.208542	211276.621184	211276.638270	high_real	149111	174
3	trigger	0.291959	211276.706721	211276.723284	high_real	149285	272
4	trigger	0.417084	211276.839682	211276.855888	high_real	149557	137

['beginning' 'trigger' 'pause' 'unpause' 'end']
[ 148150 2335453]
For subject 7, trial 0, movie 'Cars 2'

	type	movie_time	start_time	end_time	trig_type	index	diff
0	beginning	0.000000	162108.308298	162108.554011	manual	28933	623
1	trigger	0.083417	162108.611594	162108.629048	high_real	29556	478
2	trigger	0.208542	162108.845048	162108.861856	high_real	30034	168
3	trigger	0.291959	162108.926974	162108.943775	high_real	30202	275
4	trigger	0.417084	162109.061252	162109.076978	high_real	30477	172

['beginning' 'trigger' 'pause' 'unpause' 'end']
[  28933 2136814]

In [5]:

def obtain_movie_start_end(sub_id, trial_id):
    # Data frames column IDs
    start_col, end_col = 'start', 'end'
    trig_time_col, trig_idx_col, est_idx_col, est_end_idx_col = 'movie_time', 'index', 'est_idx', 'est_end_idx'

    # Path to trigger times csv file
    trigger_times_file = os.path.join(ROOT_DIR, f'subject_timings/sub_{sub_id}_trial{trial_id:03}_timings.csv')

    trigs_df = pd.read_csv(trigger_times_file)
    display(trigs_df.head())

    start_row = trigs_df[trigs_df['type'] == 'beginning']
    end_row = trigs_df[trigs_df['type'] == 'end']

    display(trigs_df[trigs_df['type'].isin(['end', 'beginning', 'pause', 'unpause'])])

    start_row_movie_time = start_row[trig_time_col].iloc[-1]
    end_row_movie_time = end_row[trig_time_col].iloc[-1]

    return start_row_movie_time, end_row_movie_time

obtain_movie_start_end(6, 0), 

def obtain_movie_start_end(sub_id, trial_id): # Data frames column IDs start_col, end_col = 'start', 'end' trig_time_col, trig_idx_col, est_idx_col, est_end_idx_col = 'movie_time', 'index', 'est_idx', 'est_end_idx' # Path to trigger times csv file trigger_times_file = os.path.join(ROOT_DIR, f'subject_timings/sub_{sub_id}_trial{trial_id:03}_timings.csv') trigs_df = pd.read_csv(trigger_times_file) display(trigs_df.head()) start_row = trigs_df[trigs_df['type'] == 'beginning'] end_row = trigs_df[trigs_df['type'] == 'end'] display(trigs_df[trigs_df['type'].isin(['end', 'beginning', 'pause', 'unpause'])]) start_row_movie_time = start_row[trig_time_col].iloc[-1] end_row_movie_time = end_row[trig_time_col].iloc[-1] return start_row_movie_time, end_row_movie_time obtain_movie_start_end(6, 0),

	type	movie_time	start_time	end_time	trig_type	index	diff
0	beginning	0.000000	799931.832360	799932.077568	manual	750390	594
1	trigger	0.083417	799932.121551	799932.138288	high_real	750984	514
2	trigger	0.208542	799932.372601	799932.389387	high_real	751498	170
3	trigger	0.291959	799932.456058	799932.472823	high_real	751668	241
4	trigger	0.417084	799932.574002	799932.590690	high_real	751909	167

	type	movie_time	start_time	end_time	trig_type	index	diff
0	beginning	0.000000	799931.832360	799932.077568	manual	750390	594
84	pause	8.259646	799940.479197	799940.670997	high_real	768059	4114
85	end	8.299967	799942.488024	799942.757538	est	772173	23405
86	beginning	8.299967	799953.897067	799954.141700	manual	795578	1899
14922	pause	1493.119612	801439.495334	801439.687382	est	3838230	111439
14923	unpause	1493.119612	801493.906661	801494.124263	manual	3949669	572
24492	pause	2450.719612	802451.785065	802451.978110	est	5911498	80659
24493	end	2450.742409	802491.169149	802491.438261	est	5992157	-1

Out[5]:

((8.299966633, 2450.742409075),)

In [6]:

Out[6]:

1.3497222222222223

In [ ]:

import os
import pandas as pd
import numpy as np
from bfm.evaluation.neuroprobe.config import ROOT_DIR, SAMPLING_RATE


def obtain_neural_data_index(sub_id, trial_id, movie_times):
    # Data frames column IDs
    start_col, end_col = 'start', 'end'
    trig_time_col, trig_idx_col, est_idx_col, est_end_idx_col = 'movie_time', 'index', 'est_idx', 'est_end_idx'

    # Path to trigger times csv file
    trigger_times_file = os.path.join(ROOT_DIR, f'subject_timings/sub_{sub_id}_trial{trial_id:03}_timings.csv')

    trigs_df = pd.read_csv(trigger_times_file)
    #display(trigs_df.head())

    last_t = trigs_df[trig_time_col].iloc[-1]
    assert np.all(movie_times < last_t), "Movie times must be less than the last trigger time"
    
    # Vectorized nearest trigger finding
    start_indices = np.searchsorted(trigs_df[trig_time_col].values, movie_times)
    start_indices = np.maximum(start_indices, 0) # handle the edge case where movie starts right at the word
    
    # Vectorized sample index calculation
    return np.round(
        trigs_df.loc[start_indices, trig_idx_col].values + 
        (movie_times - trigs_df.loc[start_indices, trig_time_col].values) * SAMPLING_RATE
    ).astype(int)

# subject 10, trial 0, movie "Cars 2"
print("For subject 10, trial 0, movie 'Cars 2'")
print(obtain_neural_data_index(10, 0, np.array([0, 1000])))

# subject 7, trial 0, movie "Cars 2"
print("For subject 7, trial 0, movie 'Cars 2'")
print(obtain_neural_data_index(7, 0, np.array([0, 1000])))

import os import pandas as pd import numpy as np from bfm.evaluation.neuroprobe.config import ROOT_DIR, SAMPLING_RATE def obtain_neural_data_index(sub_id, trial_id, movie_times): # Data frames column IDs start_col, end_col = 'start', 'end' trig_time_col, trig_idx_col, est_idx_col, est_end_idx_col = 'movie_time', 'index', 'est_idx', 'est_end_idx' # Path to trigger times csv file trigger_times_file = os.path.join(ROOT_DIR, f'subject_timings/sub_{sub_id}_trial{trial_id:03}_timings.csv') trigs_df = pd.read_csv(trigger_times_file) #display(trigs_df.head()) last_t = trigs_df[trig_time_col].iloc[-1] assert np.all(movie_times < last_t), "Movie times must be less than the last trigger time" # Vectorized nearest trigger finding start_indices = np.searchsorted(trigs_df[trig_time_col].values, movie_times) start_indices = np.maximum(start_indices, 0) # handle the edge case where movie starts right at the word # Vectorized sample index calculation return np.round( trigs_df.loc[start_indices, trig_idx_col].values + (movie_times - trigs_df.loc[start_indices, trig_time_col].values) * SAMPLING_RATE ).astype(int) # subject 10, trial 0, movie "Cars 2" print("For subject 10, trial 0, movie 'Cars 2'") print(obtain_neural_data_index(10, 0, np.array([0, 1000]))) # subject 7, trial 0, movie "Cars 2" print("For subject 7, trial 0, movie 'Cars 2'") print(obtain_neural_data_index(7, 0, np.array([0, 1000])))

For subject 10, trial 0, movie 'Cars 2'
[ 148150 2335453]
For subject 7, trial 0, movie 'Cars 2'
[  28933 2136814]