Introduction to feature extraction

import vbi
import numpy as np
import matplotlib.pyplot as plt
from vbi import report_cfg
from vbi import extract_features_df, extract_features_list
from vbi import get_features_by_domain, get_features_by_given_names

D = vbi.LoadSample(nn=88)
ts = D.get_bold()
print(ts.shape)

plt.figure(figsize=(12,2.5))
plt.plot(np.arange(ts.shape[1])*0.5, ts.T, lw=1, alpha=0.5);
plt.margins(0, 0.1)
plt.xlabel("Time (s)")
plt.ylabel("BOLD signal");

(88, 208)

cfg = get_features_by_domain(domain="connectivity")
cfg = get_features_by_given_names(cfg, ['fc_stat'])
cfg = vbi.update_cfg(cfg, 'fc_stat', {"features":["sum"], "fc_function":"corrcoef"})
report_cfg(cfg)

Selected features:
------------------
■ Domain: connectivity
 ▢ Function:  fc_stat
   ▫ description:  Extracts features from functional connectivity (FC).
   ▫ function   :  vbi.feature_extraction.features.fc_stat
   ▫ parameters :  {'k': 0, 'masks': None, 'positive': False, 'verbose': False, 'eigenvalues': True, 'pca_num_components': 3, 'fc_function': 'corrcoef', 'quantiles': [0.05, 0.25, 0.5, 0.75, 0.95], 'features': ['sum']}
   ▫ tag        :  all
   ▫ use        :  yes

data = extract_features_list([ts], 0.5, cfg)
data

  0%|          | 0/1 [00:00<?, ?it/s]
100%|██████████| 1/1 [00:00<00:00, 157.36it/s]

([array([-5.5172849e-02,  1.8978356e-01,  3.6331561e-01,  5.4156488e-01,
          7.6214993e-01, -8.8817842e-15,  1.0000000e+00,  1.3969272e+03,
          2.9416184e+03], dtype=float32)],
 ['fc_full_quantile_0.05',
  'fc_full_quantile_0.25',
  'fc_full_quantile_0.5',
  'fc_full_quantile_0.75',
  'fc_full_quantile_0.95',
  'fc_full_pca_sum',
  'fc_full_eig_sum',
  'fc_full_ut_sum',
  'fc_full_sum'])

data = extract_features_df([ts, ts], 0.5, cfg, n_workers=2)
data

  0%|          | 0/2 [00:00<?, ?it/s]
100%|██████████| 2/2 [00:00<00:00, 172.80it/s]

	fc_full_quantile_0.05	fc_full_quantile_0.25	fc_full_quantile_0.5	fc_full_quantile_0.75	fc_full_quantile_0.95	fc_full_pca_sum	fc_full_eig_sum	fc_full_ut_sum	fc_full_sum
0	-0.055173	0.189784	0.363316	0.541565	0.76215	-8.881784e-15	1.0	1396.927246	2941.618408
1	-0.055173	0.189784	0.363316	0.541565	0.76215	-8.881784e-15	1.0	1396.927246	2941.618408

Add custom features

from vbi import add_features_from_json
import my_features

cfg = add_features_from_json("my_features.json", my_features)
report_cfg(cfg)

Selected features:
------------------
■ Domain: statistical
 ▢ Function:  mean
   ▫ description:
   ▫ function   :  mean
   ▫ parameters :  {'arg1': 1.0, 'arg2': 2.0}
   ▫ tag        :  all
   ▫ use        :  yes
 ▢ Function:  std
   ▫ description:
   ▫ function   :  std
   ▫ parameters :  {'arg1': 1.0, 'arg2': 2.0}
   ▫ tag        :  all
   ▫ use        :  yes

data = extract_features_df([ts, ts], 0.5, cfg, n_workers=2)
data

  0%|          | 0/2 [00:00<?, ?it/s]
100%|██████████| 2/2 [00:00<00:00, 190.50it/s]

	m_0	m_1	m_2	m_3	m_4	m_5	m_6	m_7	m_8	m_9	...	s_78	s_79	s_80	s_81	s_82	s_83	s_84	s_85	s_86	s_87
0	0.017111	0.009509	0.029611	0.006994	0.025506	0.029548	0.029585	-0.01176	-0.039832	0.018644	...	3.790679	3.203614	2.421138	2.160106	4.565415	3.082228	3.249217	1.872746	3.495425	1.700081
1	0.017111	0.009509	0.029611	0.006994	0.025506	0.029548	0.029585	-0.01176	-0.039832	0.018644	...	3.790679	3.203614	2.421138	2.160106	4.565415	3.082228	3.249217	1.872746	3.495425	1.700081

2 rows × 176 columns