OSCR

Multimodal multicentre investigation of diagnostic and prognostic markers in disorders of consciousness.

Code ↔ Paper

4 matches between paragraphs of the paper and lines of its authors' code, computed by the harvester (lexical-v1). Click a colored paragraph or line to see its counterpart.

The 4 matches
  1. [1] § Results › Feature importance differs in diagnostic and prognostic models ↔ modeling/01_efm_model-results.py, lines 108–165 · score 0.82 · symbolic mutual information, permutation entropy, Kolmogorov complexity, half brain, EEG RS, beta
  2. [2] § Results › Feature importance differs in diagnostic and prognostic models ↔ modeling/07_auc_ci_per_feature.py, lines 209–245 · score 0.77 · symbolic mutual information, permutation entropy, Kolmogorov complexity, EEG RS, beta, evoked
  3. [3] § Results › Feature importance differs in diagnostic and prognostic models ↔ fmri_markers/fmri_functions.py, lines 217–324 · score 0.64 · connectivity features, subcortical areas, subcortical regions, limbic, volume, network
  4. [4] § Results › Multimodal integration improves predictive accuracy ↔ modeling/06_efm_model_outputs_and_surrogates_plots_generalization.py, lines 60–113 · score 0.53 · Mann Whitney, Bonferroni corrected, Balanced accuracy, model, modalities

Paper

Loaded from Europe PMC by your browser, not stored by OSCR: doi.org · Europe PMC

The paper is loaded when this pane is shown.

The authors' code

Python · 1,276 lines · 58 KB · no license · 1 match

  1. #%%
  2. import seaborn as sns
  3. import pandas as pd
  4. import matplotlib.pyplot as plt
  5. import numpy as np
  6. import os
  7. import seaborn as sns
  9. import nibabel as nib
  10. from nilearn import datasets
  11. from nilearn.datasets import fetch_atlas_harvard_oxford
  12. from nilearn import plotting
  13. from pdf2image import convert_from_path
  16. # %%
  17. base_path = '/home/Documents/codes/permed'
  18. efm_figs_dir = os.path.join(base_path, '00_derivatives/model_outputs/EarlyFusionMultimodal/figs')
  20. surrogates_analysis = False
  22. clinical_prediction = 'diagnosis' # '6m1y2y-prognosis' # 'diagnosis'
  23. name_ext=f'single-mods_allmod-per-patient_norfe_rskf_{clinical_prediction}_RandomForestClassifier'
  24. if clinical_prediction == '6m1y2y-prognosis': name_ext = name_ext + '_no-lata'
  25. if surrogates_analysis: name_ext = f'{name_ext}_surrogates'
  27. metric = 'balanced_accuracy'
  29. df_efm= pd.read_csv(os.path.join(base_path,
  30. f'00_derivatives/model_outputs/EarlyFusionMultimodal/csvs/efm_{name_ext}_folds-500.csv'))
  32. df_efm_fi = pd.read_csv(os.path.join(base_path,
  33. f'00_derivatives/model_outputs/EarlyFusionMultimodal/csvs/efm_feat-importances_{name_ext}_folds-500.csv'))
  34. df_efm_fi.set_index('Unnamed: 0', inplace=True)
  36. # %%
  37. eeg_rs_features = list(np.load(os.path.join(base_path, '00_derivatives/intermediate/eeg_rs_features.npy'), allow_pickle=True))
  38. eeg_lg_features = list(np.load(os.path.join(base_path, '00_derivatives/intermediate/eeg_lg_features.npy'), allow_pickle=True))
  39. anat_features = list(np.load(os.path.join(base_path, '00_derivatives/intermediate/anat_features.npy'), allow_pickle=True))[1:]
  40. func_rs_features = list(np.load(os.path.join(base_path, '00_derivatives/intermediate/func_rs_features.npy'), allow_pickle=True))[1:]
  41. dti_features = list(np.load(os.path.join(base_path, '00_derivatives/intermediate/dti_features.npy'), allow_pickle=True))
  42. pet_features = list(np.load(os.path.join(base_path, '00_derivatives/intermediate/pet_features.npy'), allow_pickle=True))
  43. # %%
  44. df_efm
  45. # in df_efm create a column with the difference between the accuracy and balanced accuracy and plot the violin plot
  46. df_efm['diff_acc_bal_acc'] = df_efm['accuracy'] - df_efm['balanced_accuracy']
  47. df_efm['diff_acc_bal_acc'].describe()
  49. # %%
  51. sns.set_theme()
  52. sns.set_context("paper")
  53. sns.set_style("white")
  55. metric = 'auc' # 'auc' 'balanced_accuracy'
  59. # %%
  61. # Separate the feature importances per modality
  63. # From df_efm_fi keep only the rows that have /rs/ in the index name, and eeg_rs in the column name
  64. fi_eeg_rs = df_efm_fi[df_efm_fi.index.str.contains('/rs/')]
  65. fi_eeg_rs = fi_eeg_rs[fi_eeg_rs.columns[fi_eeg_rs.columns.str.contains('eeg_rs')]]
  67. fi_eeg_lg = df_efm_fi[df_efm_fi.index.str.contains('/lg/')]
  68. fi_eeg_lg = fi_eeg_lg[fi_eeg_lg.columns[fi_eeg_lg.columns.str.contains('eeg_lg')]]
  70. fi_anat = df_efm_fi[df_efm_fi.index.str.contains('anat_')]
  71. fi_anat = fi_anat[fi_anat.columns[fi_anat.columns.str.contains('anat')]]
  73. fi_func = df_efm_fi[df_efm_fi.index.str.contains('func-')]
  74. fi_func = fi_func[fi_func.columns[fi_func.columns.str.contains('func')]]
  76. fi_dti = df_efm_fi[df_efm_fi.index.str.contains('maskJHU')]
  77. fi_dti = fi_dti.append(df_efm_fi[df_efm_fi.index.str.contains('fa_global')])
  78. fi_dti = fi_dti.append(df_efm_fi[df_efm_fi.index.str.contains('md_global')])
  79. fi_dti = fi_dti[fi_dti.columns[fi_dti.columns.str.contains('dti')]]
  81. fi_pet = df_efm_fi[df_efm_fi.index.str.contains('pet_')]
  82. fi_pet = fi_pet[fi_pet.columns[fi_pet.columns.str.contains('pet')]]
  83. # %%
  85. fi_eeg_rs_mean_std = fi_eeg_rs.copy().apply(lambda x: pd.Series({"mean":x.mean(),"std":x.std()}), axis = 1)
  86. fi_eeg_rs_mean_std_sorted = fi_eeg_rs_mean_std.sort_values(by='mean', ascending=False, inplace=False)
  89. fi_eeg_lg_mean_std = fi_eeg_lg.copy().apply(lambda x: pd.Series({"mean":x.mean(),"std":x.std()}), axis = 1)
  90. fi_eeg_lg_mean_std_sorted = fi_eeg_lg_mean_std.sort_values(by='mean', ascending=False, inplace=False)
  92. fi_anat_mean_std = fi_anat.copy().apply(lambda x: pd.Series({"mean":x.mean(),"std":x.std()}), axis = 1)
  93. fi_anat_mean_std_sorted = fi_anat_mean_std.sort_values(by='mean', ascending=False, inplace=False)
  95. fi_func_mean_std = fi_func.copy().apply(lambda x: pd.Series({"mean":x.mean(),"std":x.std()}), axis = 1)
  96. fi_func_mean_std_sorted = fi_func_mean_std.sort_values(by='mean', ascending=False, inplace=False)
  98. fi_dti_mean_std = fi_dti.copy().apply(lambda x: pd.Series({"mean":x.mean(),"std":x.std()}), axis = 1)
  99. fi_dti_mean_std_sorted = fi_dti_mean_std.sort_values(by='mean', ascending=False, inplace=False)
  101. fi_pet_mean_std = fi_pet.copy().apply(lambda x: pd.Series({"mean":x.mean(),"std":x.std()}), axis = 1)
  102. fi_pet_mean_std_sorted = fi_pet_mean_std.sort_values(by='mean', ascending=False, inplace=False)
  105. # Summarizing feat importances per cortical networks and subcortical
  106. # or by conceptual families etc
  108. sns.set_context("talk")
  111. plot_feature_groups = True
  113. if plot_feature_groups:
  114. nets = ['Cont', 'Default', 'DorsAttn', 'Limbic', 'SalVentAttn', 'SomMot', 'Vis']
  116. ###### DISTRIBUTIONS
  118. # aMRI
  119. anat = fi_anat_mean_std['mean'].to_frame()
  120. # Add the 'region_groups' column
  121. anat['region_groups'] = anat.index.to_series().apply(
  122. lambda x: next((net for net in nets if net in x), 'Subcortical' if any(sub in x for sub in ['stem', 'left', 'right']) else 'Other')
  123. )
  125. # pet
  126. pet = fi_pet_mean_std['mean'].to_frame()
  127. # Add the 'region_groups' column
  128. pet['region_groups'] = pet.index.to_series().apply(
  129. lambda x: next((net for net in nets if net in x), 'Subcortical' if any(sub in x for sub in ['Stem', 'Left', 'Right']) else 'Other')
  130. )
  131. # add the value 'Half Brain' in the column region_groups for the features that contain 'high_mi' or 'low_mi'
  132. pet.loc[pet.index.str.contains('high_mi|low_mi'), 'region_groups'] = 'Half Brain'
  135. # func
  136. func = fi_func_mean_std['mean'].to_frame()
  137. def assign_region_group(x):
  138. if any(sub in x for sub in ['subcortical', 'Left', 'Right', 'Stem']):
  139. return 'Subcortical'
  140. elif 'sub-to-cort' in x:
  141. return 'Subcortical to Cortical'
  142. else:
  143. return next((net for net in nets if net in x), 'Other')
  145. func['region_groups'] = func.index.to_series().apply(assign_region_group)
  147. # EEG RS
  148. # 3 divisions (code in the 021_efm_model-results.py)
  149. # 5 divisions
  150. eeg_rs = fi_eeg_rs_mean_std['mean'].to_frame()
  151. def assign_region_group(x):
  152. if 'PowerSpectralDensity' in x:
  153. if any(sub in x for sub in ['delta', 'theta', 'alpha']):
  154. return 'Spectral: Low'
  155. elif any(sub in x for sub in ['beta', 'gamma']):
  156. return 'Spectral: High'
  157. else:
  158. return 'Spectral: Other'
  159. elif 'SymbolicMutualInformation' in x:
  160. return 'Connectivity'
  161. elif any(sub in x for sub in ['KolmogorovComplexity', 'PermutationEntropy']):
  162. return 'Information theory'
  163. else:
  164. return 'Other'
  165. eeg_rs['region_groups'] = eeg_rs.index.to_series().apply(assign_region_group)
  167. # EEG LG
  168. eeg_lg = fi_eeg_lg_mean_std['mean'].to_frame()
  169. def assign_region_group(x):
  170. if 'PowerSpectralDensity' in x:
  171. if any(sub in x for sub in ['delta', 'theta', 'alpha']):
  172. return 'Spectral: Low'
  173. elif any(sub in x for sub in ['beta', 'gamma']):
  174. return 'Spectral: High'
  175. else:
  176. return 'Spectral: Other'
  177. elif 'SymbolicMutualInformation' in x:
  178. return 'Connectivity'
  179. elif any(sub in x for sub in ['KolmogorovComplexity', 'PermutationEntropy']):
  180. return 'Information theory'
  181. elif any(sub in x for sub in ['TimeLockedTopography', 'TimeLockedContrast', 'ContingentNegativeVariation', 'WindowDecoding']):
  182. return 'Evoked'
  183. else:
  184. return 'Other'
  185. eeg_lg['region_groups'] = eeg_lg.index.to_series().apply(assign_region_group)
  188. # DTI
  189. dti = fi_dti_mean_std['mean'].to_frame()
  191. region_dict = {
  192. "maskJHU_FA_1_MCP": "Brainstem",
  193. "maskJHU_FA_2_PCT": "Brainstem",
  194. "maskJHU_FA_3_gCC": "Commissural fibers",
  195. "maskJHU_FA_4_bCC": "Commissural fibers",
  196. "maskJHU_FA_5_sCC": "Commissural fibers",
  197. "maskJHU_FA_7_CST_R": "Brainstem",
  198. "maskJHU_FA_8_CST_L": "Brainstem",
  199. "maskJHU_FA_9_mLEM_R": "Brainstem",
  200. "maskJHU_FA_10_mLEM_L": "Brainstem",
  201. "maskJHU_FA_11_ICP_R": "Brainstem",
  202. "maskJHU_FA_12_ICP_L": "Brainstem",
  203. "maskJHU_FA_13_SCP_R": "Brainstem",
  204. "maskJHU_FA_14_SCP_L": "Brainstem",
  205. "maskJHU_FA_15_CP_R": "Projection fibers",
  206. "maskJHU_FA_16_CP_L": "Projection fibers",
  207. "maskJHU_FA_17_ALIC_R": "Projection fibers",
  208. "maskJHU_FA_18_ALIC_L": "Projection fibers",
  209. "maskJHU_FA_19_PLIC_R": "Projection fibers",
  210. "maskJHU_FA_20_PLIC_L": "Projection fibers",
  211. "maskJHU_FA_21_RLIC_R": "Projection fibers",
  212. "maskJHU_FA_22_RLIC_L": "Projection fibers",
  213. "maskJHU_FA_23_ACR_R": "Projection fibers",
  214. "maskJHU_FA_24_ACR_L": "Projection fibers",
  215. "maskJHU_FA_25_SCR_R": "Projection fibers",
  216. "maskJHU_FA_26_SCR_L": "Projection fibers",
  217. "maskJHU_FA_27_PCR_R": "Projection fibers",
  218. "maskJHU_FA_28_PCR_L": "Projection fibers",
  219. "maskJHU_FA_29_PTR_R": "Projection fibers",
  220. "maskJHU_FA_30_PTR_L": "Projection fibers",
  221. "maskJHU_FA_31_ILF_IFOF_R": "Associative fibers",
  222. "maskJHU_FA_32_ILF_IFOF_L": "Associative fibers",
  223. "maskJHU_FA_33_EC_R": "Associative fibers",
  224. "maskJHU_FA_34_EC_L": "Associative fibers",
  225. "maskJHU_FA_35_Cing_R": "Associative fibers",
  226. "maskJHU_FA_36_Cing_L": "Associative fibers",
  227. "maskJHU_FA_37_Cing_h_R": "Associative fibers",
  228. "maskJHU_FA_38_Cing_h_L": "Associative fibers",
  229. "maskJHU_FA_41_SLF_R": "Associative fibers",
  230. "maskJHU_FA_42_SLF_L": "Associative fibers",
  231. "maskJHU_FA_43_SFOF_R": "Associative fibers",
  232. "maskJHU_FA_44_SFOF_L": "Associative fibers",
  233. "maskJHU_FA_45_UNC_R": "Associative fibers",
  234. "maskJHU_FA_46_UNC_L": "Associative fibers",
  235. "maskJHU_MD_1_MCP": "Brainstem",
  236. "maskJHU_MD_2_PCT": "Brainstem",
  237. "maskJHU_MD_3_gCC": "Commissural fibers",
  238. "maskJHU_MD_4_bCC": "Commissural fibers",
  239. "maskJHU_MD_5_sCC": "Commissural fibers",
  240. "maskJHU_MD_7_CST_R": "Brainstem",
  241. "maskJHU_MD_8_CST_L": "Brainstem",
  242. "maskJHU_MD_9_mLEM_R": "Brainstem",
  243. "maskJHU_MD_10_mLEM_L": "Brainstem",
  244. "maskJHU_MD_11_ICP_R": "Brainstem",
  245. "maskJHU_MD_12_ICP_L": "Brainstem",
  246. "maskJHU_MD_13_SCP_R": "Brainstem",
  247. "maskJHU_MD_14_SCP_L": "Brainstem",
  248. "maskJHU_MD_15_CP_R": "Projection fibers",
  249. "maskJHU_MD_16_CP_L": "Projection fibers",
  250. "maskJHU_MD_17_ALIC_R": "Projection fibers",
  251. "maskJHU_MD_18_ALIC_L": "Projection fibers",
  252. "maskJHU_MD_19_PLIC_R": "Projection fibers",
  253. "maskJHU_MD_20_PLIC_L": "Projection fibers",
  254. "maskJHU_MD_21_RLIC_R": "Projection fibers",
  255. "maskJHU_MD_22_RLIC_L": "Projection fibers",
  256. "maskJHU_MD_23_ACR_R": "Projection fibers",
  257. "maskJHU_MD_24_ACR_L": "Projection fibers",
  258. "maskJHU_MD_25_SCR_R": "Projection fibers",
  259. "maskJHU_MD_26_SCR_L": "Projection fibers",
  260. "maskJHU_MD_27_PCR_R": "Projection fibers",
  261. "maskJHU_MD_28_PCR_L": "Projection fibers",
  262. "maskJHU_MD_29_PTR_R": "Projection fibers",
  263. "maskJHU_MD_30_PTR_L": "Projection fibers",
  264. "maskJHU_MD_31_ILF_IFOF_R": "Associative fibers",
  265. "maskJHU_MD_32_ILF_IFOF_L": "Associative fibers",
  266. "maskJHU_MD_33_EC_R": "Associative fibers",
  267. "maskJHU_MD_34_EC_L": "Associative fibers",
  268. "maskJHU_MD_35_Cing_R": "Associative fibers",
  269. "maskJHU_MD_36_Cing_L": "Associative fibers",
  270. "maskJHU_MD_37_Cing_h_R": "Associative fibers",
  271. "maskJHU_MD_38_Cing_h_L": "Associative fibers",
  272. "maskJHU_MD_41_SLF_R": "Associative fibers",
  273. "maskJHU_MD_42_SLF_L": "Associative fibers",
  274. "maskJHU_MD_43_SFOF_R": "Associative fibers",
  275. "maskJHU_MD_44_SFOF_L": "Associative fibers",
  276. "maskJHU_MD_45_UNC_R": "Associative fibers",
  277. "maskJHU_MD_46_UNC_L": "Associative fibers",
  278. "fa_global": "Global",
  279. "md_global": "Global"
  280. }
  282. region_dict_split = {
  283. "maskJHU_FA_1_MCP": "FA Brainstem",
  284. "maskJHU_FA_2_PCT": "FA Brainstem",
  285. "maskJHU_FA_3_gCC": "FA Commissural fibers",
  286. "maskJHU_FA_4_bCC": "FA Commissural fibers",
  287. "maskJHU_FA_5_sCC": "FA Commissural fibers",
  288. "maskJHU_FA_7_CST_R": "FA Brainstem",
  289. "maskJHU_FA_8_CST_L": "FA Brainstem",
  290. "maskJHU_FA_9_mLEM_R": "FA Brainstem",
  291. "maskJHU_FA_10_mLEM_L": "FA Brainstem",
  292. "maskJHU_FA_11_ICP_R": "FA Brainstem",
  293. "maskJHU_FA_12_ICP_L": "FA Brainstem",
  294. "maskJHU_FA_13_SCP_R": "FA Brainstem",
  295. "maskJHU_FA_14_SCP_L": "FA Brainstem",
  296. "maskJHU_FA_15_CP_R": "FA Projection fibers",
  297. "maskJHU_FA_16_CP_L": "FA Projection fibers",
  298. "maskJHU_FA_17_ALIC_R": "FA Projection fibers",
  299. "maskJHU_FA_18_ALIC_L": "FA Projection fibers",
  300. "maskJHU_FA_19_PLIC_R": "FA Projection fibers",
  301. "maskJHU_FA_20_PLIC_L": "FA Projection fibers",
  302. "maskJHU_FA_21_RLIC_R": "FA Projection fibers",
  303. "maskJHU_FA_22_RLIC_L": "FA Projection fibers",
  304. "maskJHU_FA_23_ACR_R": "FA Projection fibers",
  305. "maskJHU_FA_24_ACR_L": "FA Projection fibers",
  306. "maskJHU_FA_25_SCR_R": "FA Projection fibers",
  307. "maskJHU_FA_26_SCR_L": "FA Projection fibers",
  308. "maskJHU_FA_27_PCR_R": "FA Projection fibers",
  309. "maskJHU_FA_28_PCR_L": "FA Projection fibers",
  310. "maskJHU_FA_29_PTR_R": "FA Projection fibers",
  311. "maskJHU_FA_30_PTR_L": "FA Projection fibers",
  312. "maskJHU_FA_31_ILF_IFOF_R": "FA Associative fibers",
  313. "maskJHU_FA_32_ILF_IFOF_L": "FA Associative fibers",
  314. "maskJHU_FA_33_EC_R": "FA Associative fibers",
  315. "maskJHU_FA_34_EC_L": "FA Associative fibers",
  316. "maskJHU_FA_35_Cing_R": "FA Associative fibers",
  317. "maskJHU_FA_36_Cing_L": "FA Associative fibers",
  318. "maskJHU_FA_37_Cing_h_R": "FA Associative fibers",
  319. "maskJHU_FA_38_Cing_h_L": "FA Associative fibers",
  320. "maskJHU_FA_41_SLF_R": "FA Associative fibers",
  321. "maskJHU_FA_42_SLF_L": "FA Associative fibers",
  322. "maskJHU_FA_43_SFOF_R": "FA Associative fibers",
  323. "maskJHU_FA_44_SFOF_L": "FA Associative fibers",
  324. "maskJHU_FA_45_UNC_R": "FA Associative fibers",
  325. "maskJHU_FA_46_UNC_L": "FA Associative fibers",
  326. "maskJHU_MD_1_MCP": "MD Brainstem",
  327. "maskJHU_MD_2_PCT": "MD Brainstem",
  328. "maskJHU_MD_3_gCC": "MD Commissural fibers",
  329. "maskJHU_MD_4_bCC": "MD Commissural fibers",
  330. "maskJHU_MD_5_sCC": "MD Commissural fibers",
  331. "maskJHU_MD_7_CST_R": "MD Brainstem",
  332. "maskJHU_MD_8_CST_L": "MD Brainstem",
  333. "maskJHU_MD_9_mLEM_R": "MD Brainstem",
  334. "maskJHU_MD_10_mLEM_L": "MD Brainstem",
  335. "maskJHU_MD_11_ICP_R": "MD Brainstem",
  336. "maskJHU_MD_12_ICP_L": "MD Brainstem",
  337. "maskJHU_MD_13_SCP_R": "MD Brainstem",
  338. "maskJHU_MD_14_SCP_L": "MD Brainstem",
  339. "maskJHU_MD_15_CP_R": "MD Projection fibers",
  340. "maskJHU_MD_16_CP_L": "MD Projection fibers",
  341. "maskJHU_MD_17_ALIC_R": "MD Projection fibers",
  342. "maskJHU_MD_18_ALIC_L": "MD Projection fibers",
  343. "maskJHU_MD_19_PLIC_R": "MD Projection fibers",
  344. "maskJHU_MD_20_PLIC_L": "MD Projection fibers",
  345. "maskJHU_MD_21_RLIC_R": "MD Projection fibers",
  346. "maskJHU_MD_22_RLIC_L": "MD Projection fibers",
  347. "maskJHU_MD_23_ACR_R": "MD Projection fibers",
  348. "maskJHU_MD_24_ACR_L": "MD Projection fibers",
  349. "maskJHU_MD_25_SCR_R": "MD Projection fibers",
  350. "maskJHU_MD_26_SCR_L": "MD Projection fibers",
  351. "maskJHU_MD_27_PCR_R": "MD Projection fibers",
  352. "maskJHU_MD_28_PCR_L": "MD Projection fibers",
  353. "maskJHU_MD_29_PTR_R": "MD Projection fibers",
  354. "maskJHU_MD_30_PTR_L": "MD Projection fibers",
  355. "maskJHU_MD_31_ILF_IFOF_R": "MD Associative fibers",
  356. "maskJHU_MD_32_ILF_IFOF_L": "MD Associative fibers",
  357. "maskJHU_MD_33_EC_R": "MD Associative fibers",
  358. "maskJHU_MD_34_EC_L": "MD Associative fibers",
  359. "maskJHU_MD_35_Cing_R": "MD Associative fibers",
  360. "maskJHU_MD_36_Cing_L": "MD Associative fibers",
  361. "maskJHU_MD_37_Cing_h_R": "MD Associative fibers",
  362. "maskJHU_MD_38_Cing_h_L": "MD Associative fibers",
  363. "maskJHU_MD_41_SLF_R": "MD Associative fibers",
  364. "maskJHU_MD_42_SLF_L": "MD Associative fibers",
  365. "maskJHU_MD_43_SFOF_R": "MD Associative fibers",
  366. "maskJHU_MD_44_SFOF_L": "MD Associative fibers",
  367. "maskJHU_MD_45_UNC_R": "MD Associative fibers",
  368. "maskJHU_MD_46_UNC_L": "MD Associative fibers",
  369. "fa_global": "FA Global",
  370. "md_global": "MD Global"
  371. }
  373. dti['region_groups'] = dti.index.to_series().apply(lambda x: region_dict_split[x])
  376. #
  377. def plot_ordered_boxplot(df, modality_name, clin_pred, fig_name, arrows_plot=False, arrows_info=None,
  378. x_limits=None, x_ticks=None, x_tick_labels=None):
  379. """
  380. Plot a boxplot with the regions ordered by the mean of the feature importance.
  382. Parameters
  383. ----------
  384. df : DataFrame
  385. DataFrame with the feature importances.
  386. modality_name : str
  387. Name of the modality.
  388. clin_pred : str
  389. Clinical prediction.
  390. fig_name : str
  391. Name of the figure.
  392. arrows_plot : bool, optional
  393. If True, plot arrows in the plot. The default is False.
  394. arrows_info : list
  395. List of tuples with the number of arrows, color and location of the arrows in the plot.
  396. x_limits : tuple, optional
  397. Limits for the x-axis.
  398. x_ticks : list, optional
  399. Ticks for the x-axis.
  400. x_tick_labels : list, optional
  401. Labels for the x-axis ticks.
  402. """
  404. # Calculate the order
  405. order = df.groupby('region_groups')['mean'].mean().sort_values(ascending=False).index
  407. sns.boxplot(x='mean', y='region_groups', data=df, order=order, orient="h", color=".8")
  409. plt.xlabel("Mean Feature Importance")
  410. if clin_pred == 'diagnosis':
  411. plt.ylabel("Groups of Features")
  412. plt.title(f'Diagnosis: {modality_name}')
  413. else:
  414. plt.ylabel('')
  415. plt.title(f'Prognosis: {modality_name}')
  417. sns.despine(top=True, right=True)
  419. if x_limits:
  420. plt.xlim(x_limits)
  421. if x_ticks:
  422. plt.xticks(x_ticks)
  423. if x_tick_labels:
  424. plt.gca().set_xticklabels(x_tick_labels)
  426. if arrows_plot:
  427. # Add arrows to y-labels
  428. for i, (n_arrows, sign, location) in enumerate(arrows_info):
  429. color = 'green' if sign == '↑' else 'red'
  430. plt.text(location[0], location[1], str(n_arrows)+' x'+sign if n_arrows != 0 else '',
  431. color=color,
  432. va='center',
  433. ha='center',
  434. fontsize=12
  435. )
  437. plt.savefig(base_path+f'/00_derivatives/feature_importances_groups_{clin_pred}/{fig_name}.pdf',
  438. bbox_inches='tight')
  439. plt.show()
  444. if clinical_prediction == 'diagnosis':
  445. arrows_plot = False
  446. elif clinical_prediction == '6m1y2y-prognosis':
  447. arrows_plot = False # TODO changed
  449. fig_name = f'{clinical_prediction}_mean-groups'
  450. if arrows_plot:
  451. fig_name = fig_name + '_arrows-plot'
  453. # Define the tick limits, values, and labels based on the clinical prediction
  454. if clinical_prediction == 'diagnosis':
  455. eeg_rs_x_limits = (0, 0.05)
  456. eeg_rs_x_ticks = [0, 0.01, 0.02, 0.03, 0.04, 0.05]
  457. eeg_rs_x_tick_labels = ['0', '0.01', '0.02', '0.03', '0.04', '0.05']
  459. pet_x_limits = (0, 0.05)
  460. pet_x_ticks = [0, 0.01, 0.02, 0.03, 0.04, 0.05]
  461. pet_x_tick_labels = ['0', '0.01', '0.02', '0.03', '0.04', '0.05']
  462. elif clinical_prediction == '6m1y2y-prognosis':
  463. amri_x_limits = (0, 0.05)
  464. amri_x_ticks = [0, 0.01, 0.02, 0.03, 0.04, 0.05]
  465. amri_x_tick_labels = ['0', '0.01', '0.02', '0.03', '0.04', '0.05']
  467. # Plot the ordered boxplots with the specified tick limits, values, and labels
  468. plot_ordered_boxplot(anat, 'aMRI', clinical_prediction,
  469. f'{fig_name}_anat',
  470. arrows_plot=arrows_plot,
  471. arrows_info=[(7, '↑', (-0.0058, 0)), # Subcortical
  472. (1, '↑', (-0.006, 1)), # SalVentAttn
  473. (2, '↑', (-0.001, 2)), # Vis
  474. (3, '↑', (-0.003, 3)), # Limbic
  475. (1, '↑', (-0.002, 4)), # Cont
  476. (5, '↓', (-0.004, 5)), # SomMot
  477. (3, '↓', (-0.0045, 6)), # DorsAttn
  478. (6, '↓', (-0.0037, 7)) # Default
  479. ],
  480. x_limits=amri_x_limits if clinical_prediction == '6m1y2y-prognosis' else None,
  481. x_ticks=amri_x_ticks if clinical_prediction == '6m1y2y-prognosis' else None,
  482. x_tick_labels=amri_x_tick_labels if clinical_prediction == '6m1y2y-prognosis' else None)
  484. plot_ordered_boxplot(pet, 'PET', clinical_prediction,
  485. f'{fig_name}_pet',
  486. arrows_plot=arrows_plot,
  487. arrows_info=[(7, '↑', (-0.0075, 0)), # Subcortical
  488. (0, '', (-0.0008, 1)), # SomMot
  489. (4, '↑', (-0.004, 2)), # Cont
  490. (2, '↑', (-0.0065, 3)), # DorsAttn
  491. (1, '↓', (-0.0078, 4)), # SalVentAttn
  492. (3, '↑', (-0.005, 5)), # Limbic
  493. (2, '↓', (-0.0055, 6)), # Default
  494. (5, '↓', (-0.003, 7)), # Vis
  495. (8, '↓', (-0.007, 8)), # Half Brain
  496. ],
  497. x_limits=pet_x_limits if clinical_prediction == 'diagnosis' else None,
  498. x_ticks=pet_x_ticks if clinical_prediction == 'diagnosis' else None,
  499. x_tick_labels=pet_x_tick_labels if clinical_prediction == 'diagnosis' else None)
  501. plot_ordered_boxplot(func, 'fMRI', clinical_prediction,
  502. f'{fig_name}_func',
  503. arrows_plot=arrows_plot,
  504. arrows_info=[(2, '↑', (-0.0093, 0)), # SomMot
  505. (0, '', (-0.0075, 1)), # Subcortical
  506. (2, '↓', (-0.0182, 2)), # Subcortical to Cortical
  507. (3, '↑', (-0.0083, 3)), # Limbic
  508. (4, '↑', (-0.007, 4)), # Cont
  509. (0, '', (-0.0055, 5)), # Default
  510. (2, '↓', (-0.0115, 6)), # SalVentAttn
  511. (0, '', (-0.0065, 7)), # DorsAttn
  512. (5, '↓', (-0.006, 8)), # Vis
  513. ])
  515. plot_ordered_boxplot(eeg_rs, 'EEG RS', clinical_prediction,
  516. f'{fig_name}_eeg_rs',
  517. arrows_plot=arrows_plot,
  518. arrows_info=[(1, '↑', (-0.0001, 0)), # Spectral: High
  519. (1, '↓', (0.0003, 1)), # Spectral: Low
  520. (1, '↑', (0.0008, 2)), # Connectivity
  521. (1, '↑', (-0.0004, 3)), # Spectral: Other
  522. (2, '↓', (-0.002, 4)) # Information theory
  523. ],
  524. x_limits=eeg_rs_x_limits if clinical_prediction == 'diagnosis' else None,
  525. x_ticks=eeg_rs_x_ticks if clinical_prediction == 'diagnosis' else None,
  526. x_tick_labels=eeg_rs_x_tick_labels if clinical_prediction == 'diagnosis' else None)
  528. plot_ordered_boxplot(eeg_lg, 'EEG LG', clinical_prediction,
  529. f'{fig_name}_eeg_lg',
  530. arrows_plot=arrows_plot,
  531. arrows_info=[(3, '↑', (-0.0077, 0)), # Spectral: High
  532. (3, '↑', (-0.0082, 1)), # Spectral: Other
  533. (2, '↓', (-0.0074, 2)), # Spectral: Low
  534. (1, '↓', (-0.0095, 3)), # Information theory
  535. (3, '↓', (-0.007, 4)), # Connectivity
  536. (0, '', (-0.005, 5)) # Evoked
  537. ])
  539. plot_ordered_boxplot(dti, 'dMRI', clinical_prediction,
  540. f'{fig_name}_dti',
  541. arrows_plot=arrows_plot,
  542. arrows_info=[(0, '', (-0.015, 0)), # FA global
  543. (1, '↑', (-0.02, 1)), # FA brainstem
  544. (7, '↑', (-0.0325, 2)), # MD commissural fibers
  545. (2, '↓', (-0.0275, 3)), # FA projection fibers
  546. (0, '', (-0.03, 4)), # FA associative
  547. (2, '↓', (-0.031, 5)), # FA commissural
  548. (0, '', (-0.03, 6)), # MD projection
  549. (0, '', (-0.03, 7)), # MD associative
  550. (3, '↓', (-0.022, 8)), # MD brainstem
  551. (1, '↓', (-0.017, 9)) # MD global
  552. ])
  554. #%% Brain plots for aMRI, fMRI, PET
  556. ################################
  557. # CORTICAL PLOTS - 4 brains for supplementary material
  559. def plot_nice_surf(data, density='32k', cmap='coolwarm', dpi=250, template='inflated',
  560. atlas=None, cbar_label=None, vmin=None, vmax=None, threshold=None,
  561. organization='2x2', clin_pred=None, fig_name=None):
  562. """
  563. Plot nice plots in fsLR space.
  565. Parameters
  566. ----------
  567. data : array_like or tuple
  568. ROI-wise or vertex-wise data. If tuple, assumes (left, right) hemisphere.
  569. density : str
  570. Density of surface plot, can be '8k', '32k' or '164k'.
  571. cmap : str
  572. Colormap.
  573. template : str
  574. Type of surface plot. Can be 'inflated', 'veryinflated', 'sphere' or 'midthickness'
  575. dpi : int
  576. Resolution of plot.
  577. atlas : Path, optional
  578. Path to an atlas in .dlabel.nii format.
  579. cbar_label: str, optional
  580. Colorbar label.
  581. vmin/vmax : int, optional
  582. Minimun/ maximum value in the plot.
  583. threshold : float, optional
  584. Threshold for the data. If None, no thresholding is applied.
  585. organization : str, optional
  586. How to organize the plots. Can be '1x4' or '2x2'.
  588. # the orig threshold was -1e-14
  589. """
  590. from nilearn.plotting import plot_surf
  591. from neuromaps.datasets import fetch_fslr
  592. from neuromaps.parcellate import Parcellater
  593. from neuromaps.images import dlabel_to_gifti
  595. if atlas is not None:
  596. atlas = dlabel_to_gifti(atlas)
  597. surf_masker = Parcellater(atlas, "fslr")
  598. data = surf_masker.inverse_transform(data)
  599. l_data, r_data = data[0].agg_data(), data[1].agg_data()
  600. else:
  601. if not isinstance(data, tuple):
  602. raise ValueError("Data input must be tuple-of-arrays. Alternatively provide 'atlas' for ROI data.")
  603. l_data, r_data = data[0], data[1]
  605. if None in (vmin, vmax):
  606. # Handle NaNs in left hemisphere data
  607. l_min, l_max = np.nanmin(l_data), np.nanmax(l_data)
  608. l_data = np.nan_to_num(l_data, nan=l_min)
  610. # Handle NaNs in right hemisphere data
  611. r_min, r_max = np.nanmin(r_data), np.nanmax(r_data)
  612. r_data = np.nan_to_num(r_data, nan=r_min)
  614. # min/max values in the data
  615. vmin = np.min([l_min, r_min])
  616. vmax = np.max([l_max, r_max])
  618. print(f'vmin: {vmin}, vmax: {vmax}')
  620. # Fetch surface template for plot
  621. surfaces = fetch_fslr(density=density)
  622. lh, rh = surfaces[template]
  624. if organization == '1x4':
  625. # Plot both hemispheres
  626. fig, ax = plt.subplots(nrows=1,ncols=4,subplot_kw={'projection': '3d'}, figsize=(12, 4), dpi=dpi)
  628. plot_surf(lh, l_data, threshold=threshold, cmap=cmap, alpha=1, view='lateral',
  629. colorbar=False, axes=ax.flat[0],
  630. vmin=vmin, vmax=vmax)
  631. plot_surf(lh, l_data, threshold=threshold, cmap=cmap, alpha=1, view='medial',
  632. colorbar=False, axes=ax.flat[1],
  633. vmin=vmin, vmax=vmax)
  635. plot_surf(rh, r_data, threshold=threshold, cmap=cmap, alpha=1, view='lateral',
  636. colorbar=False, axes=ax.flat[2],
  637. vmin=vmin, vmax=vmax)
  638. p = plot_surf(rh, r_data, threshold=threshold, cmap=cmap, alpha=1, view='medial',
  639. colorbar=True, axes=ax.flat[3],
  640. vmin=vmin, vmax=vmax)
  642. p.axes[-1].set_ylabel(cbar_label, fontsize=14, labelpad=0.5)
  643. p.axes[-1].set_yticks([vmin, vmax])
  644. #p.axes[-1].set_yticklabels(['min', 'max'])
  645. p.axes[-1].tick_params(labelsize=12, width=0, pad=0.1)
  646. plt.subplots_adjust(wspace=-0.05)
  647. p.axes[-1].set_position(p.axes[-1].get_position().translated(0.08, 0))
  649. elif organization == '2x2':
  650. # Plot both hemispheres
  651. fig, ax = plt.subplots(nrows=2, ncols=2, subplot_kw={'projection': '3d'}, figsize=(8, 8), dpi=dpi)
  653. plot_surf(lh, l_data, threshold=threshold, cmap=cmap, alpha=1, view='lateral',
  654. colorbar=False, axes=ax[0, 0],
  655. vmin=vmin, vmax=vmax)
  656. plot_surf(rh, r_data, threshold=threshold, cmap=cmap, alpha=1, view='medial',
  657. colorbar=False, axes=ax[0, 1],
  658. vmin=vmin, vmax=vmax)
  660. plot_surf(lh, l_data, threshold=threshold, cmap=cmap, alpha=1, view='medial',
  661. colorbar=False, axes=ax[1, 0],
  662. vmin=vmin, vmax=vmax)
  663. p = plot_surf(rh, r_data, threshold=threshold, cmap=cmap, alpha=1, view='lateral',
  664. colorbar=True, axes=ax[1, 1],
  665. vmin=vmin, vmax=vmax)
  667. p.axes[-1].set_ylabel(cbar_label, fontsize=10, labelpad=0.5)
  668. p.axes[-1].set_yticks([vmin, vmax])
  669. p.axes[-1].tick_params(labelsize=7, width=0, pad=0.1)
  670. plt.subplots_adjust(wspace=-0.05, hspace=0.05)
  671. p.axes[-1].set_position(p.axes[-1].get_position().translated(0.08, 0))
  673. plt.savefig(base_path+f'/00_derivatives/feature_importances_{clin_pred}/{fig_name}.pdf',
  674. bbox_inches='tight')
  676. return p
  679. # aMRI
  680. # take the first 100 rows of the fi_anat_mean_std:
  681. fi_anat_mean_std_cortical = fi_anat_mean_std.iloc[:100]
  682. fi_anat_mean_std_cortical.index = fi_anat_mean_std_cortical.index.str.replace('anat_lh_', '')
  683. fi_anat_mean_std_cortical.index = fi_anat_mean_std_cortical.index.str.replace('anat_rh_', '')
  684. fi_anat_mean_std_cortical.index = fi_anat_mean_std_cortical.index.str.replace('_thickness', '')
  686. base_path = '/home/Documents/codes/permed'
  687. atlas_path = base_path+'/multimodal/Schaefer2018_100Parcels_7Networks_order.dlabel.nii'
  689. # plot the cortical features first for the fi_anat_mean_std_cortical.mean and the fi_anat_mean_std_cortical.std:
  690. anat_cort_img = plot_nice_surf(fi_anat_mean_std_cortical['mean'], density='32k', dpi=250, template='inflated',
  691. atlas=atlas_path,
  692. cbar_label='thickness (mean)', cmap='pink',
  693. vmin=0, vmax=0.22,
  694. # vmin=None, vmax=None,
  695. threshold=None,
  696. organization='1x4',
  697. clin_pred=clinical_prediction, fig_name=f'{clinical_prediction}_fi_mean_anat_cortical'
  698. )
  700. plot_nice_surf(fi_anat_mean_std_cortical['std'], density='32k', dpi=250, template='inflated',
  701. atlas=atlas_path,
  702. cbar_label='thickness (std)', cmap='pink',
  703. vmin=0, vmax=0.08,
  704. # vmin=None, vmax=None,
  705. threshold=None,
  706. organization='1x4',
  707. clin_pred=clinical_prediction, fig_name=f'{clinical_prediction}_fi_std_anat_cortical'
  708. )
  711. # fMRI
  712. fi_func_mean_std_cortical = fi_func_mean_std[fi_func_mean_std.index.str.contains('7Networks')]
  713. fi_func_mean_std_cortical.index = fi_func_mean_std_cortical.index.str.replace('func-connectivity_mean_', '')
  715. func_cort_img = plot_nice_surf(fi_func_mean_std_cortical['mean'], density='32k', dpi=250, template='inflated',
  716. atlas=atlas_path,
  717. cbar_label='functional correlation (mean)', cmap='pink',
  718. # vmin=-0.1, vmax=0.1,
  719. threshold=None,
  720. organization='1x4',
  721. clin_pred=clinical_prediction, fig_name=f'{clinical_prediction}_fi_mean_fmri_cortical'
  722. )
  724. plot_nice_surf(fi_func_mean_std_cortical['std'], density='32k', dpi=250, template='inflated',
  725. atlas=atlas_path,
  726. cbar_label='functional correlation (std)', cmap='pink',
  727. # vmin=0, vmax=0.08,
  728. threshold=None,
  729. organization='1x4',
  730. clin_pred=clinical_prediction, fig_name=f'{clinical_prediction}_fi_std_fmri_cortical'
  731. )
  733. #
  734. # PET
  735. # take the first 100 rows of the fi_pet_mean_std:
  736. fi_pet_mean_std_cortical = fi_pet_mean_std.iloc[:100]
  737. fi_pet_mean_std_cortical.index = fi_pet_mean_std_cortical.index.str.replace('pet_', '')
  739. pet_cort_img = plot_nice_surf(fi_pet_mean_std_cortical['mean'], density='32k', dpi=250, template='inflated',
  740. atlas=atlas_path,
  741. cbar_label='SUV (mean)', cmap='pink',
  742. # vmin=0, vmax=0.22,
  743. threshold=None,
  744. organization='1x4',
  745. clin_pred=clinical_prediction, fig_name=f'{clinical_prediction}_fi_mean_pet_cortical'
  746. )
  748. plot_nice_surf(fi_pet_mean_std_cortical['std'], density='32k', dpi=250, template='inflated',
  749. atlas=atlas_path,
  750. cbar_label='SUV (std)', cmap='pink',
  751. # vmin=0, vmax=0.08,
  752. threshold=None,
  753. organization='1x4',
  754. clin_pred=clinical_prediction, fig_name=f'{clinical_prediction}_fi_std_pet_cortical'
  755. )
  758. #%%
  759. ##############################
  760. # SUBCORTICAL PLOTS - 4 brains, for supplementary material
  762. def modify_atlas(atlas, atlas_roi_labels, marker_values, atlas_name=None):
  763. """
  764. Modify the values in an atlas image.
  765. marker_values is a dataframe with the following columns:
  766. - roi_label: the label of the ROI in the atlas
  767. - new_value: the new value to assign to the ROI
  768. """
  769. # Load the atlas image
  770. if atlas_name=='Schaefer2018':
  771. atlas_img = nib.load(atlas.maps)
  773. elif atlas_name=='HarvardOxfordSubcortical':
  774. atlas_img = atlas.maps
  776. # Initialize a matrix with the same shape as the atlas
  777. combined_matrix = np.zeros(atlas_img.shape)
  779. if atlas_name=='HarvardOxfordSubcortical':
  780. print('We have the following labels : ', atlas_roi_labels)
  781. roi_to_remove = ['Background',
  782. 'Left Cerebral White Matter',
  783. 'Left Cerebral Cortex ',
  784. 'Left Lateral Ventrical',
  785. 'Right Cerebral White Matter',
  786. 'Right Cerebral Cortex ',
  787. 'Right Lateral Ventricle']
  788. print('Removing the following labels : ', roi_to_remove)
  789. # From the list roi_labels, remove the elements in roi_to_remove
  790. atlas_roi_labels = [x for x in atlas_roi_labels if x not in roi_to_remove]
  791. print('We have the following labels : ', atlas_roi_labels)
  793. for label in atlas_roi_labels:
  794. # Check if the label exists in marker_values
  795. if label in marker_values['roi_label'].values:
  796. # From the dataframe marker_values, get the new_value which corresponds to the current label
  797. new_val = marker_values[marker_values['roi_label'] == label]['new_value'].values[0]
  798. else:
  799. print(f"Label {label} not found in marker_values")
  801. # Find the index of the label in the atlas
  802. if atlas_name=='Schaefer2018':
  803. roi_idx = np.where(atlas.labels == label.encode('UTF-8'))[0]+1
  804. elif atlas_name=='HarvardOxfordSubcortical':
  805. roi_idx = np.where(np.array(atlas.labels) == label)[0]
  806. print(f'Label {label} has index : {roi_idx} and new value : {new_val}.')
  808. # Create a binary mask for the current ROI using the atlas image
  809. from nilearn import image
  810. idx_map = image.math_img('img == %s' % roi_idx, img=atlas.maps)
  811. idx_map_data = idx_map.get_fdata()
  813. # Set the marker value for voxels within the current ROI
  814. idx_map_data[idx_map_data != 0] = new_val
  815. print(f'Unique values in idx_map : {np.unique(idx_map_data)}')
  817. # Add the current ROI mask to the combined matrix
  818. combined_matrix = np.add(combined_matrix, idx_map_data)
  820. # Create a new Nifti image with the modified data
  821. modified_atlas_img = nib.Nifti1Image(combined_matrix, atlas_img.affine, atlas_img.header)
  822. # or:
  823. # modified_atlas_img = new_img_like(atlas_img, combined_matrix)
  825. return modified_atlas_img
  827. def plot_subcortical_regions(atlas, atlas_roi_labels, marker_values, atlas_name=None,
  828. clinical_prediction=None, fig_name=None, savefig=False):
  829. """
  830. Plot the subcortical regions in an atlas image.
  831. marker_values is a dataframe with the following columns:
  832. - roi_label: the label of the ROI in the atlas
  833. - new_value: the new value to assign to the ROI
  834. """
  835. # Modify the atlas using the provided marker values
  836. modified_atlas_img = modify_atlas(atlas, atlas_roi_labels, marker_values, atlas_name)
  838. if savefig:
  839. # Plot the modified atlas
  840. display = plotting.plot_glass_brain(stat_map_img=modified_atlas_img,
  841. colorbar=True,
  842. cmap='twilight',
  843. alpha=1,
  844. display_mode="lyrz")
  845. display.add_contours(modified_atlas_img, filled=True)
  847. # save the plot
  848. display.savefig(base_path+f'/00_derivatives/feature_importances_{clinical_prediction}/{fig_name}.pdf')
  850. return modified_atlas_img
  853. # aMRI
  854. print('aMRI')
  855. # take the first 100 rows of the fi_anat_mean_std:
  856. fi_anat_mean_std_subcortical = fi_anat_mean_std.iloc[100:]
  857. fi_anat_mean_std_subcortical.index = fi_anat_mean_std_subcortical.index.str.replace('anat_', '')
  859. base_path = '/home/Documents/codes/permed'
  861. # create a dataframe for marker_values
  862. marker_values_mean = pd.DataFrame({
  863. 'roi_label': fi_anat_mean_std_subcortical.index,
  864. 'new_value': fi_anat_mean_std_subcortical['mean']
  865. })
  867. marker_values_std = pd.DataFrame({
  868. 'roi_label': fi_anat_mean_std_subcortical.index,
  869. 'new_value': fi_anat_mean_std_subcortical['std']
  870. })
  872. # rename the marker_values_mean roi_label values to match those of the Harvard-Oxford atlas by putting the first letters of the words in capital
  873. marker_values_mean['roi_label'] = marker_values_mean['roi_label'].str.title()
  874. marker_values_std['roi_label'] = marker_values_std['roi_label'].str.title()
  877. atlas = fetch_atlas_harvard_oxford('sub-maxprob-thr25-1mm')
  878. # load the atlas
  879. # atlas = nib.load(atlas_path)
  881. # plot the subcortical features first for the fi_anat_mean_std_subcortical.mean and the fi_anat_mean_std_subcortical.std:
  882. anat_img = plot_subcortical_regions(atlas, atlas.labels, marker_values_mean, atlas_name='HarvardOxfordSubcortical',
  883. clinical_prediction=clinical_prediction, fig_name=f'{clinical_prediction}_fi_mean_anat_subcortical')
  885. # do the same for fmri
  886. print('fMRI')
  887. fi_func_mean_std_subcortical = fi_func_mean_std.iloc[100:]
  888. fi_func_mean_std_subcortical.index = fi_func_mean_std_subcortical.index.str.replace('func-connectivity_mean_', '')
  889. # create a dataframe for marker_values
  890. marker_values_mean = pd.DataFrame({
  891. 'roi_label': fi_func_mean_std_subcortical.index,
  892. 'new_value': fi_func_mean_std_subcortical['mean']
  893. })
  894. marker_values_std = pd.DataFrame({
  895. 'roi_label': fi_func_mean_std_subcortical.index,
  896. 'new_value': fi_func_mean_std_subcortical['std']
  897. })
  898. # rename the marker_values_mean roi_label values to match those of the Harvard-Oxford atlas by putting the first letters of the words in capital
  899. marker_values_mean['roi_label'] = marker_values_mean['roi_label'].str.title()
  900. marker_values_std['roi_label'] = marker_values_std['roi_label'].str.title()
  901. func_img = plot_subcortical_regions(atlas, atlas.labels, marker_values_mean, atlas_name='HarvardOxfordSubcortical',
  902. clinical_prediction=clinical_prediction, fig_name=f'{clinical_prediction}_fi_mean_fmri_subcortical')
  904. # do the same for PET
  905. print('PET')
  906. fi_pet_mean_std_subcortical = fi_pet_mean_std.iloc[100:]
  907. fi_pet_mean_std_subcortical.index = fi_pet_mean_std_subcortical.index.str.replace('pet_', '')
  908. # create a dataframe for marker_values
  909. marker_values_mean = pd.DataFrame({
  910. 'roi_label': fi_pet_mean_std_subcortical.index,
  911. 'new_value': fi_pet_mean_std_subcortical['mean']
  912. })
  913. marker_values_std = pd.DataFrame({
  914. 'roi_label': fi_pet_mean_std_subcortical.index,
  915. 'new_value': fi_pet_mean_std_subcortical['std']
  916. })
  917. # rename the marker_values_mean roi_label values to match those of the Harvard-Oxford atlas by putting the first letters of the words in capital
  918. marker_values_mean['roi_label'] = marker_values_mean['roi_label'].str.title()
  919. marker_values_std['roi_label'] = marker_values_std['roi_label'].str.title()
  920. pet_img = plot_subcortical_regions(atlas, atlas.labels, marker_values_mean, atlas_name='HarvardOxfordSubcortical',
  921. clinical_prediction=clinical_prediction, fig_name=f'{clinical_prediction}_fi_mean_pet_subcortical')
  924. #%% FINAL FIGURE Supplementary Figure - 3x2 brain plots (cortical and subcortical) mean feature importances of aMRI, fMRI, PET
  926. clin_pred = clinical_prediction
  927. # Load the images
  928. anat_cort_img = convert_from_path(base_path+f'/00_derivatives/feature_importances_{clin_pred}/{clin_pred}_fi_mean_anat_cortical.pdf')[0]
  929. func_cort_img = convert_from_path(base_path+f'/00_derivatives/feature_importances_{clin_pred}/{clin_pred}_fi_mean_fmri_cortical.pdf')[0]
  930. pet_cort_img = convert_from_path(base_path+f'/00_derivatives/feature_importances_{clin_pred}/{clin_pred}_fi_mean_pet_cortical.pdf')[0]
  932. # Convert PIL Images to numpy arrays
  933. anat_cort_img = np.array(anat_cort_img)
  934. func_cort_img = np.array(func_cort_img)
  935. pet_cort_img = np.array(pet_cort_img)
  938. # Display the images in subplots
  939. fig, ax = plt.subplots(nrows=3, ncols=2, figsize=(16, 8), dpi=300)
  941. # Cortical plots
  942. # aMRI
  943. ax[0, 0].imshow(anat_cort_img)
  944. ax[0, 0].axis('off')
  945. ax[0, 0].set_title('aMRI cortical mean feature importances',
  946. color='black', bbox={'facecolor': 'white', 'alpha': 0.5, 'pad': 10},
  947. fontsize=14, y=1.2
  948. )
  950. # fMRI
  951. ax[1, 0].imshow(func_cort_img)
  952. ax[1, 0].axis('off')
  953. ax[1, 0].set_title('fMRI cortical mean feature importances',
  954. color='black', bbox={'facecolor': 'white', 'alpha': 0.5, 'pad': 10},
  955. fontsize=14, y=1.2
  956. )
  958. # PET
  959. ax[2, 0].imshow(pet_cort_img)
  960. ax[2, 0].axis('off')
  961. ax[2, 0].set_title('PET cortical mean feature importances',
  962. color='black', bbox={'facecolor': 'white', 'alpha': 0.5, 'pad': 10},
  963. fontsize=14, y=1.2
  964. )
  966. # Subcortical plots
  967. # aMRI
  968. display = plotting.plot_glass_brain(stat_map_img=anat_img, colorbar=True, cmap='twilight', alpha=1, display_mode="lyrz", axes=ax[0, 1])
  969. display.add_contours(anat_img, filled=True)
  970. ax[0, 1].set_title('aMRI subcortical mean feature importances',
  971. color='black', bbox={'facecolor': 'white', 'alpha': 0.5, 'pad': 10}, fontsize=14)
  972. # Get the colorbar object
  973. cbar = display._cbar
  974. # Change the tick location to right
  975. cbar.ax.yaxis.set_tick_params(left=False, right=True, labelleft=False, labelright=True)
  977. # fMRI
  978. display = plotting.plot_glass_brain(stat_map_img=func_img, colorbar=True, cmap='twilight', alpha=1, display_mode="lyrz", axes=ax[1, 1])
  979. display.add_contours(func_img, filled=True)
  980. ax[1, 1].set_title('fMRI subcortical mean feature importances',
  981. color='black', bbox={'facecolor': 'white', 'alpha': 0.5, 'pad': 10}, fontsize=14)
  982. cbar = display._cbar
  983. cbar.ax.yaxis.set_tick_params(left=False, right=True, labelleft=False, labelright=True)
  985. # PET
  986. display = plotting.plot_glass_brain(stat_map_img=pet_img, colorbar=True, cmap='twilight', alpha=1, display_mode="lyrz", axes=ax[2, 1])
  987. display.add_contours(pet_img, filled=True)
  988. ax[2, 1].set_title('PET subcortical mean feature importances',
  989. color='black', bbox={'facecolor': 'white', 'alpha': 0.5, 'pad': 10}, fontsize=14)
  990. cbar = display._cbar
  991. cbar.ax.yaxis.set_tick_params(left=False, right=True, labelleft=False, labelright=True)
  994. # save the figure
  995. plt.savefig(base_path+f'/00_derivatives/feature_importances_groups_diagnosis_6m1y2y-prognosis/{clinical_prediction}_fi_mean_cort-subcort.pdf',
  996. bbox_inches='tight')
  999. # Combination of Diagnosis and prognosis with lines how the feature groups ranking changes
  1000. # diagnosis (glass brains, boxplots) -lines- prognosis (boxplots, glass brains)
  1002. # The difference with the code above is that the code above has 4 brains
  1003. # in the plots from all sides and the cortical plot is a different brain design.
  1006. def plot_regions(atlas, atlas_roi_labels, marker_values, atlas_name=None,
  1007. clinical_prediction=None, fig_name=None):
  1008. """
  1009. Plot the regions in an atlas image.
  1010. marker_values is a dataframe with the following columns:
  1011. - roi_label: the label of the ROI in the atlas
  1012. - new_value: the new value to assign to the ROI
  1013. """
  1014. # Modify the atlas using the provided marker values
  1015. modified_atlas_img = modify_atlas(atlas, atlas_roi_labels, marker_values, atlas_name)
  1017. # Plot the modified atlas
  1018. display = plotting.plot_glass_brain(stat_map_img=modified_atlas_img,
  1019. colorbar=True,
  1020. cmap='twilight',
  1021. alpha=0.7,
  1022. display_mode="z")
  1023. display.add_contours(modified_atlas_img, filled=True)
  1025. # save the plot
  1026. display.savefig(base_path+
  1027. f'/00_derivatives/feature_importances_groups_diagnosis_6m1y2y-prognosis/{fig_name}.pdf')
  1029. return modified_atlas_img
  1031. ############################# SUBCORTICAL #####################################
  1032. # aMRI
  1033. print('aMRI')
  1034. # take the first 100 rows of the fi_anat_mean_std:
  1035. fi_anat_mean_std_subcortical = fi_anat_mean_std.iloc[100:]
  1036. fi_anat_mean_std_subcortical.index = fi_anat_mean_std_subcortical.index.str.replace('anat_', '')
  1038. base_path = '/home/Documents/codes/permed'
  1040. # create a dataframe for marker_values
  1041. marker_values_mean = pd.DataFrame({
  1042. 'roi_label': fi_anat_mean_std_subcortical.index,
  1043. 'new_value': fi_anat_mean_std_subcortical['mean']
  1044. })
  1046. marker_values_std = pd.DataFrame({
  1047. 'roi_label': fi_anat_mean_std_subcortical.index,
  1048. 'new_value': fi_anat_mean_std_subcortical['std']
  1049. })
  1051. # rename the marker_values_mean roi_label values to match those of the Harvard-Oxford atlas by putting the first letters of the words in capital
  1052. marker_values_mean['roi_label'] = marker_values_mean['roi_label'].str.title()
  1053. marker_values_std['roi_label'] = marker_values_std['roi_label'].str.title()
  1056. atlas = fetch_atlas_harvard_oxford('sub-maxprob-thr25-1mm')
  1057. # load the atlas
  1058. # atlas = nib.load(atlas_path)
  1060. # plot the subcortical features first for the fi_anat_mean_std_subcortical.mean and the fi_anat_mean_std_subcortical.std:
  1061. modified_atlas_img = plot_regions(atlas, atlas.labels, marker_values_mean, atlas_name='HarvardOxfordSubcortical',
  1062. clinical_prediction=clinical_prediction, fig_name=f'{clinical_prediction}_fi_mean_anat_subcortical')
  1063. plot_regions(atlas, atlas.labels, marker_values_std, atlas_name='HarvardOxfordSubcortical',
  1064. clinical_prediction=clinical_prediction, fig_name=f'{clinical_prediction}_fi_std_anat_subcortical')
  1066. # do the same for fmri
  1067. print('fMRI')
  1068. fi_func_mean_std_subcortical = fi_func_mean_std.iloc[100:]
  1069. fi_func_mean_std_subcortical.index = fi_func_mean_std_subcortical.index.str.replace('func-connectivity_mean_', '')
  1070. # create a dataframe for marker_values
  1071. marker_values_mean = pd.DataFrame({
  1072. 'roi_label': fi_func_mean_std_subcortical.index,
  1073. 'new_value': fi_func_mean_std_subcortical['mean']
  1074. })
  1075. marker_values_std = pd.DataFrame({
  1076. 'roi_label': fi_func_mean_std_subcortical.index,
  1077. 'new_value': fi_func_mean_std_subcortical['std']
  1078. })
  1079. # rename the marker_values_mean roi_label values to match those of the Harvard-Oxford atlas by putting the first letters of the words in capital
  1080. marker_values_mean['roi_label'] = marker_values_mean['roi_label'].str.title()
  1081. marker_values_std['roi_label'] = marker_values_std['roi_label'].str.title()
  1082. plot_regions(atlas, atlas.labels, marker_values_mean, atlas_name='HarvardOxfordSubcortical',
  1083. clinical_prediction=clinical_prediction, fig_name=f'{clinical_prediction}_fi_mean_fmri_subcortical')
  1084. plot_regions(atlas, atlas.labels, marker_values_std, atlas_name='HarvardOxfordSubcortical',
  1085. clinical_prediction=clinical_prediction, fig_name=f'{clinical_prediction}_fi_std_fmri_subcortical')
  1087. # do the same for PET
  1088. print('PET')
  1089. fi_pet_mean_std_subcortical = fi_pet_mean_std.iloc[100:]
  1090. fi_pet_mean_std_subcortical.index = fi_pet_mean_std_subcortical.index.str.replace('pet_', '')
  1091. # create a dataframe for marker_values
  1092. marker_values_mean = pd.DataFrame({
  1093. 'roi_label': fi_pet_mean_std_subcortical.index,
  1094. 'new_value': fi_pet_mean_std_subcortical['mean']
  1095. })
  1096. marker_values_std = pd.DataFrame({
  1097. 'roi_label': fi_pet_mean_std_subcortical.index,
  1098. 'new_value': fi_pet_mean_std_subcortical['std']
  1099. })
  1100. # rename the marker_values_mean roi_label values to match those of the Harvard-Oxford atlas by putting the first letters of the words in capital
  1101. marker_values_mean['roi_label'] = marker_values_mean['roi_label'].str.title()
  1102. marker_values_std['roi_label'] = marker_values_std['roi_label'].str.title()
  1103. plot_regions(atlas, atlas.labels, marker_values_mean, atlas_name='HarvardOxfordSubcortical',
  1104. clinical_prediction=clinical_prediction, fig_name=f'{clinical_prediction}_fi_mean_pet_subcortical')
  1105. plot_regions(atlas, atlas.labels, marker_values_std, atlas_name='HarvardOxfordSubcortical',
  1106. clinical_prediction=clinical_prediction, fig_name=f'{clinical_prediction}_fi_std_pet_subcortical')
  1109. ############################ CORTICAL ##########################################
  1111. atlas_name = 'Schaefer2018'
  1113. atlas = datasets.fetch_atlas_schaefer_2018(
  1114. n_rois=100,
  1115. yeo_networks=7,
  1116. resolution_mm=1)
  1117. roi_labels = atlas.labels
  1118. roi_labels = [row.tobytes().decode('UTF-8') for row in roi_labels]
  1120. # aMRI
  1121. # take the first 100 rows of the fi_anat_mean_std:
  1122. fi_anat_mean_std_cortical = fi_anat_mean_std.iloc[:100]
  1123. fi_anat_mean_std_cortical.index = fi_anat_mean_std_cortical.index.str.replace('anat_lh_', '')
  1124. fi_anat_mean_std_cortical.index = fi_anat_mean_std_cortical.index.str.replace('anat_rh_', '')
  1125. fi_anat_mean_std_cortical.index = fi_anat_mean_std_cortical.index.str.replace('_thickness', '')
  1127. marker_values_mean = pd.DataFrame({
  1128. 'roi_label': fi_anat_mean_std_cortical.index,
  1129. 'new_value': fi_anat_mean_std_cortical['mean']
  1130. })
  1132. marker_values_std = pd.DataFrame({
  1133. 'roi_label': fi_anat_mean_std_cortical.index,
  1134. 'new_value': fi_anat_mean_std_cortical['std']
  1135. })
  1137. # plot the aMRI plot_regions
  1138. plot_regions(atlas, roi_labels, marker_values_mean, atlas_name=atlas_name,
  1139. clinical_prediction=clinical_prediction, fig_name=f'{clinical_prediction}_fi_mean_anat_cortical')
  1140. plot_regions(atlas, roi_labels, marker_values_std, atlas_name=atlas_name,
  1141. clinical_prediction=clinical_prediction, fig_name=f'{clinical_prediction}_fi_std_anat_cortical')
  1143. # fMRI
  1144. fi_func_mean_std_cortical = fi_func_mean_std[fi_func_mean_std.index.str.contains('7Networks')]
  1145. fi_func_mean_std_cortical.index = fi_func_mean_std_cortical.index.str.replace('func-connectivity_mean_', '')
  1147. marker_values_mean = pd.DataFrame({
  1148. 'roi_label': fi_func_mean_std_cortical.index,
  1149. 'new_value': fi_func_mean_std_cortical['mean']
  1150. })
  1152. marker_values_std = pd.DataFrame({
  1153. 'roi_label': fi_func_mean_std_cortical.index,
  1154. 'new_value': fi_func_mean_std_cortical['std']
  1155. })
  1157. # plot the fMRI plot_regions
  1158. plot_regions(atlas, roi_labels, marker_values_mean, atlas_name=atlas_name,
  1159. clinical_prediction=clinical_prediction, fig_name=f'{clinical_prediction}_fi_mean_fmri_cortical')
  1160. plot_regions(atlas, roi_labels, marker_values_std, atlas_name=atlas_name,
  1161. clinical_prediction=clinical_prediction, fig_name=f'{clinical_prediction}_fi_std_fmri_cortical')
  1163. # PET
  1164. # take the first 100 rows of the fi_pet_mean_std:
  1165. fi_pet_mean_std_cortical = fi_pet_mean_std.iloc[:100]
  1166. fi_pet_mean_std_cortical.index = fi_pet_mean_std_cortical.index.str.replace('pet_', '')
  1168. marker_values_mean = pd.DataFrame({
  1169. 'roi_label': fi_pet_mean_std_cortical.index,
  1170. 'new_value': fi_pet_mean_std_cortical['mean']
  1171. })
  1173. marker_values_std = pd.DataFrame({
  1174. 'roi_label': fi_pet_mean_std_cortical.index,
  1175. 'new_value': fi_pet_mean_std_cortical['std']
  1176. })
  1178. # plot the PET plot_regions
  1179. plot_regions(atlas, roi_labels, marker_values_mean, atlas_name=atlas_name,
  1180. clinical_prediction=clinical_prediction, fig_name=f'{clinical_prediction}_fi_mean_pet_cortical')
  1181. plot_regions(atlas, roi_labels, marker_values_std, atlas_name=atlas_name,
  1182. clinical_prediction=clinical_prediction, fig_name=f'{clinical_prediction}_fi_std_pet_cortical')
  1184. #%%
  1185. # Correlations between the diagnostic and prognostic feature importances for: fMRI, EEG LG, dMRI
  1187. sns.set_theme()
  1188. sns.set_context("talk")
  1189. sns.set_style("white")
  1191. mean_feat_path = os.path.join(base_path, '00_derivatives/mean_feat_importances_with_feature-groups_splits')
  1193. # diagnosis
  1194. clinical_prediction = 'diagnosis' # '6m1y2y-prognosis' # 'diagnosis'
  1195. # fMRI
  1196. fmri_diag = pd.read_csv(os.path.join(mean_feat_path, f'{clinical_prediction}_mean_feature-groups_func.csv'), index_col=0)
  1197. # dMRI
  1198. dMRI_diag = pd.read_csv(os.path.join(mean_feat_path, f'{clinical_prediction}_mean_feature-groups_dti.csv'), index_col=0)
  1199. # EEG LG
  1200. eeg_lg_diag = pd.read_csv(os.path.join(mean_feat_path, f'{clinical_prediction}_mean_feature-groups_eeg_lg.csv'), index_col=0)
  1202. # prognosis
  1203. clinical_prediction = '6m1y2y-prognosis'
  1204. # fMRI
  1205. fmri_prog = pd.read_csv(os.path.join(mean_feat_path, f'{clinical_prediction}_mean_feature-groups_func.csv'), index_col=0)
  1206. # dMRI
  1207. dMRI_prog = pd.read_csv(os.path.join(mean_feat_path, f'{clinical_prediction}_mean_feature-groups_dti.csv'), index_col=0)
  1208. # EEG LG
  1209. eeg_lg_prog = pd.read_csv(os.path.join(mean_feat_path, f'{clinical_prediction}_mean_feature-groups_eeg_lg.csv'), index_col=0)
  1212. # in fmri_diag and fmri_prog change the values in the column region_groups from Subcortical to Cortical to Sub to Cort
  1213. fmri_diag['region_groups'] = fmri_diag['region_groups'].replace('Subcortical to Cortical', 'Sub to Cort')
  1214. fmri_prog['region_groups'] = fmri_prog['region_groups'].replace('Subcortical to Cortical', 'Sub to Cort') # ??? this was not included
  1215. #
  1217. fig, ax = plt.subplots(3, 1, figsize=(6, 16), dpi=300)
  1218. scatter_fmri = sns.scatterplot(x=fmri_diag['mean'], y=fmri_prog['mean'], hue=fmri_diag['region_groups'],
  1219. palette='deep', alpha=0.85, ax=ax[0])
  1220. scatter_dMRI = sns.scatterplot(x=dMRI_diag['mean'], y=dMRI_prog['mean'], hue=dMRI_diag['region_groups'],
  1221. palette='deep', alpha=0.85, ax=ax[1])
  1222. scatter_eeg_lg = sns.scatterplot(x=eeg_lg_diag['mean'], y=eeg_lg_prog['mean'], hue=eeg_lg_diag['region_groups'],
  1223. palette='deep', alpha=0.85, ax=ax[2])
  1224. ax[0].set_title('fMRI')
  1225. ax[1].set_title('dMRI')
  1226. ax[2].set_title('EEG LG')
  1227. ax[0].set_xlabel('Diagnosis mean FI')
  1228. ax[0].set_ylabel('Prognosis mean FI')
  1229. ax[1].set_xlabel('Diagnosis mean FI')
  1230. ax[1].set_ylabel('Prognosis mean FI')
  1231. ax[2].set_xlabel('Diagnosis mean FI')
  1232. ax[2].set_ylabel('Prognosis mean FI')
  1234. # set the limits of the first subplot to 0 to 0.04 on both x and y
  1235. ax[0].set_xlim(0, 0.04)
  1236. ax[0].set_ylim(0, 0.04)
  1237. ax[1].set_xlim(0, 0.075)
  1238. ax[1].set_ylim(0, 0.075)
  1239. ax[2].set_xlim(0, 0.03)
  1240. ax[2].set_ylim(0, 0.03)
  1242. # On all three plots but the same ticks on the x and y axes, and always put 4 ticks, two at the extremes and one in the middle:
  1243. ax[0].set_xticks([0, 0.02, 0.04])
  1244. ax[0].set_yticks([0, 0.02, 0.04])
  1245. ax[1].set_xticks([0, 0.04, 0.075])
  1246. ax[1].set_yticks([0, 0.04, 0.075])
  1247. ax[2].set_xticks([0, 0.015, 0.03])
  1248. ax[2].set_yticks([0, 0.015, 0.03])
  1250. # Set the tick labels to display 0 without commas
  1251. ax[0].set_xticklabels(['0', '0.02', '0.04'])
  1252. ax[0].set_yticklabels(['0', '0.02', '0.04'])
  1253. ax[1].set_xticklabels(['0', '0.04', '0.075'])
  1254. ax[1].set_yticklabels(['0', '0.04', '0.075'])
  1255. ax[2].set_xticklabels(['0', '0.015', '0.03'])
  1256. ax[2].set_yticklabels(['0', '0.015', '0.03'])
  1259. plt.tight_layout()
  1261. legend_fmri = scatter_fmri.legend(loc='upper right', bbox_to_anchor=(1, 1), fontsize=10)
  1262. for handle in legend_fmri.legendHandles:
  1263. handle._sizes = [35] # adjust as needed
  1264. legend_dmri = scatter_dMRI.legend(loc='upper right', bbox_to_anchor=(1, 1), fontsize=10)
  1265. for handle in legend_dmri.legendHandles:
  1266. handle._sizes = [35] # adjust as needed
  1267. legend_eeg_lg = scatter_eeg_lg.legend(loc='upper right', bbox_to_anchor=(1, 1), fontsize=10)
  1268. for handle in legend_eeg_lg.legendHandles:
  1269. handle._sizes = [35] # adjust as needed
  1272. plt.savefig(os.path.join(base_path, '00_derivatives/Figure_3_feat-imp/correlation_diag-prog_feature-importances.pdf'), bbox_inches='tight')

01_efm_model-results.py at commit 476e8bf, no license · at the source

Overview

Authors: Dragana Manasova1,2, Laouen Mayal Louan Belloli1,3,4, Martin Justinus Rosenfelder5,6,7, Lina Willacker5, Emilia Fló Rama1, Chiara Valota8,9, Bertrand Hermann10,11, Brigitte Charlotte Kaufmann1, Alice Pirastru9, Chiara Camilla Derchi9, Theresa Raiser5, Melanie Valente1, Aude Sangare1, Başak Türker1, Nadya Pyatigorskaya1, Benoît Béranger1, Michele Colombo8, Esteban Munoz-Musat1,12, Anira Escrichs13, Tiziana Atzori9
and 17 other authorsFrancesca Baglio9, Constantin Lapa14, Ansgar Berlis15, Kristina Krüger15, Tina Luther5,6, Vincent Perlbarg16, Gustavo Deco13,17, Yonathan Sanz-Perl1,17, Enzo Tagliazucchi4,18, Louis Puybasset16,19, Benjamin Rohaut1,20, Lionel Naccache1,21, Angela Comanducci9, Anat Arzi1,22,23, Mario Rosanova8, Andreas Bender5,6,24, Jacobo Diego Sitt1
ORCID iDs: Dragana Manasova, Bertrand Hermann, Brigitte Charlotte Kaufmann, Alice Pirastru, Anira Escrichs, Constantin Lapa, Yonathan Sanz-Perl, Benjamin Rohaut, Lionel Naccache
24 affiliations
  1. Institut du Cerveau - Paris Brain Institute - ICM, Inserm, CNRS, Sorbonne Université, Paris 75013, France
  2. Université Paris Cité, Paris 75006, France
  3. Laboratorio de Inteligencia Artificial Aplicada, Instituto de Ciencias de la Computación, Universidad de Buenos Aires, Buenos Aires C1053, Argentina
  4. Consejo Nacional de Investigaciones Científicas y Técnicas (CONICET), Ministry of Science, Technology and Innovation, Buenos Aires C1053, Argentina
  5. Department of Neurology, University Hospital of the Ludwig-Maximilians-Universität München, Munich 82152, Germany
  6. Therapiezentrum Burgau, Hospital for Neurological Rehabilitation, Burgau 89331, Germany
  7. Clinical and Biological Psychology, Institute of Psychology and Education, Ulm University, Ulm 89081, Germany
  8. Department of Biomedical and Clinical Sciences, University of Milano, Milan 20157, Italy
  9. IRCCS Fondazione Don Carlo Gnocchi ONLUS, Milan 20148, Italy
  10. Inserm 1266, Institute of Psychiatry and Neurosciences of Paris, Université Paris Cité, Paris F-75014, France
  11. Medical Intensive Care Unit, HEGP Hôpital, Assistance Publique - Hôpitaux de Paris-Centre (APHP-Centre), Paris 75014, France
  12. Centre Mémoire de Ressources et de Recherche, Paris Nord/Université Paris-Cité, Paris 75006, France
  13. Center for Brain and Cognition, Computational Neuroscience Group, Universitat Pompeu Fabra, Barcelona 08005, Spain
  14. Nuclear Medicine, Faculty of Medicine, University of Augsburg, Augsburg 86156, Germany
  15. Diagnostic and Interventional Neuroradiology, Faculty of Medicine, University of Augsburg, Augsburg 86156, Germany
  16. BRAINTALE SAS, Paris 75013, France
  17. Institució Catalana de la Recerca I Estudis Avançats (ICREA), Barcelona 08010, Spain
  18. Latin American Brain Health Institute (BrainLat), Universidad Adolfo Ibáñez, Santiago 7941169, Chile
  19. GRC 29, AP-HP, DMU DREAM, Department of Anaesthesiology and Critical Care Medicine, Pitié-Salpêtrière Hospital, Sorbonne University, Paris 75013, France
  20. AP-HP, Hôpital de la Pitié Salpêtrière, Neuro ICU, DMU Neurosciences, Paris 75013, France
  21. AP-HP, Hôpital Pitié - Salpêtrière, Service de Neurophysiologie Clinique, Paris 75013, France
  22. Department of Medical Neurobiology, Institute for Medical Research Israel-Canada, Faculty of Medicine, The Hebrew University of Jerusalem, Jerusalem 9112102, Israel
  23. Department of Cognitive and Brain Sciences, The Hebrew University of Jerusalem, Jerusalem 9112102, Israel
  24. Department of Neurorehabilitation, Medical Faculty, University of Augsburg, Augsburg 86156, Germany
Institutions: Centre National de la Recherche Scientifique (France); Inserm (France); Université Paris Cité (France); Sorbonne Université (France); Institut du Cerveau (France); Consejo Nacional de Investigaciones Científicas y Técnicas (Argentina); Universidad de Buenos Aires (Argentina); Ministerio de Ciencia, Tecnología e Innovación (Argentina); Universität Ulm (Germany); LMU Klinikum (Germany); Therapiezentrum Burgau (Germany); Ludwig-Maximilians-Universität München (Germany); University of Milan (Italy); Don Carlo Gnocchi Foundation (Italy); Assistance Publique – Hôpitaux de Paris (France); Hôpital Européen Georges-Pompidou (France); Institut de Psychiatrie et Neurosciences de Paris (France); Universitat Pompeu Fabra (Spain); University of Augsburg (Germany); Institució Catalana de Recerca i Estudis Avançats (Spain); Adolfo Ibáñez University (Chile); Pitié-Salpêtrière Hospital (France); Groupe de recherche clinique en anesthésie réanimation médecine périopératoire (France); Hebrew University of Jerusalem (Israel)
Journal: Brain : a journal of neurology, volume 149, issue 4, pages 1381-1395
Dates: received 21 November 2024; accepted 3 September 2025; published online 7 January 2026; in print April 2026
Type: Research article · Language: English
License: CC BY-NC
Identifiers: DOI 10.1093/brain/awaf412 · PMID 41499248 · PMCID PMC13058464 · OpenAlex W7119468174
Open access: hybrid, a free copy (OpenAlex)
Status: code verified
Categories: structural MRI / diffusion (modality), EEG (modality), PET / SPECT (modality), human (organism), other condition (population)
Methods: Machine learning, fMRI & imaging, Preprocessing
Keywords: disorders of consciousness, electrophysiology, neuroimaging, multimodal, machine learning
MeSH: Brain*, Consciousness Disorders*, Multimodal Imaging*, Neuroimaging*, Adult, Electroencephalography, Female, Humans, Machine Learning, Magnetic Resonance Imaging, Male, Middle Aged, Positron-Emission Tomography, Prognosis (* major topic)
Topic: Traumatic Brain Injury Research (Epidemiology, Medicine), according to OpenAlex
Funding: Agence Nationale de la Recherche (ANR-19-PERM-0002); Fondazione Regionale per la Ricerca Biomedica (GA 77982); MODELDxConsciousness Consortium (JTC 2023); Paris Brain Institute; Ecole Doctorale Frontières de l'Innovation en Recherche et Education-Programme Bettencourt; Federal Ministry of Education and Research (01KU2003); Italian Ministry of Health; European Research Council (101071900)
Citations: cited by 7 papers (Europe PMC); 77 references in the paper

Abstract

Severely brain-injured patients may enter a spectrum of conditions collectively known as disorders of consciousness. This spectrum includes clinical conditions such as unresponsive wakefulness syndrome or minimally conscious state, where the behavioural assessment of consciousness can often be deceptive.

To bridge this dissociation, neuroimaging techniques are employed to identify the residual brain functions. Each neuroimaging modality imperfectly captures distinct aspects of brain preservation—functional, anatomical, or both. In this study, we adopt a comprehensive approach by integrating the neurophysiology and neuroimaging modalities available from the standard and advanced clinical assessments through interpretable machine learning. The electrophysiological modalities included high-density EEG (resting state and task), whereas neuroimaging modalities included anatomical and resting-state functional MRI, diffusion MRI and 18F-fluorodeoxyglucose PET.

Our investigation reveals that specific modalities, such as functional assessments, provide comprehensive insights into the currently evaluated state of consciousness, the diagnosis of the patients. Conversely, structural modalities offer valuable information about the patient's evolution within the consciousness spectrum. We validate the proposed analysis with data coming from other centres with different acquisition parameters. Importantly, we demonstrate that model performance improves with an increase in the number of modalities. We observe a higher inter-modality disagreement for minimally conscious state patients and those patients who improve. Lastly, we observe a difference in feature importances between diagnosis and prognosis, with an interaction between modality and anatomical structures: some subcortical markers tend to contribute more to prognosis, while other cortical markers are more informative for diagnosis.

This integrative multimodal and machine learning methodology presents a promising avenue for a more nuanced understanding of disorders of consciousness, contributing to enhanced diagnostic precision, prognostic capabilities and the personalization of rehabilitative strategies in clinical practice.

Reproduced under the paper's license (CC BY-NC), from the paper cited above.

Repository

Its files are read in the Code ↔ Paper reader above, with 4 matches between paragraphs and lines of code.

DraganaMana/multimod_doc

License: none: the authors keep all their rights
State: the link answers, verified on 28 September 2026
Evidence: files inventoried
Commit: 476e8bf792c6258056bba25d82d73535edfc5379, 7 January 2026
Languages: Python (27), Shell (10)
Size: 41 files, 37 scripts
Software Heritage: not archived
Found in: “Data availability”
Holds: README
Not found: license file, CITATION.cff, environment file, tests, continuous integration, documentation
Tools: pandas (24 files), NumPy (19 files), Matplotlib (17 files), Nilearn (9 files), seaborn (9 files), scikit-learn (8 files), PyBIDS (7 files), FreeSurfer (6 files), NiBabel (4 files), SciPy (4 files), fMRIPrep (1 file), MNE-Python (1 file), neuromaps (1 file), specparam (formerly FOOOF) (1 file), statannotations (1 file), XGBoost (1 file)
Availability: 1 check, the latest on 28 September 2026: the link answers
  • 28 September 2026: the link answers
38 files
At the source: github.com/DraganaMana/multimod_doc

The paper's code and data availability statement is in the Data section.

Tracing map

Proposed by the machine: these links were found in the paper and verified at the source, without human review. The map will receive a Zenodo DOI once one of the paper's authors has validated it with their ORCID.

What the map holds:

  • 1 repository of the authors' code, each at its verified commit, with its license and how the link was found in the paper;
  • 37 scripts, each with its path and the digest of its content;
  • 4 matches between paragraphs of the paper and lines of the code (method lexical-v1);
  • neither the text of the paper nor the code itself.

Its JSON (tracing-map.json) is deposited on Zenodo with its DOI once the map is validated.

Data

No dataset and no data link were found in the paper.

Data availability

The data are not publicly available. Codes used in the analyses will be made publicly available upon publication at https://github.com/DraganaMana/multimod_doc.

Reproduced under the paper's license (CC BY-NC), from the paper cited above.

Versions

The history of this record: each version stored by the harvester or made by a correction of its authors or of the maintainers of its code, and what changed in its facts. The texts of the paper (its abstract, its availability statements) are not part of it; versions that changed only those are not listed.

Version 1, 28 September 2026: the first record

Recorded: type, language, journal, volume, issue, pages, dates, 37 authors, 5 keywords, 14 MeSH terms, 8 funders, 73 references.

Cite

This paper

Manasova, D., Belloli, L. M. L., Rosenfelder, M. J., Willacker, L., Fló Rama, E., Valota, C., Hermann, B., Kaufmann, B. C., Pirastru, A., Derchi, C. C., Raiser, T., Valente, M., Sangare, A., Türker, B., Pyatigorskaya, N., Béranger, B., Colombo, M., Munoz-Musat, E., Escrichs, A., . . . Sitt, J. D. (2026). Multimodal multicentre investigation of diagnostic and prognostic markers in disorders of consciousness. Brain : a journal of neurology, 149(4), 1381-1395. https://doi.org/10.1093/brain/awaf412

BibTeX

@article{manasova2026multimodal,
author = {Manasova, Dragana and Belloli, Laouen Mayal Louan and Rosenfelder, Martin Justinus and Willacker, Lina and Fló Rama, Emilia and Valota, Chiara and Hermann, Bertrand and Kaufmann, Brigitte Charlotte and Pirastru, Alice and Derchi, Chiara Camilla and Raiser, Theresa and Valente, Melanie and Sangare, Aude and Türker, Başak and Pyatigorskaya, Nadya and Béranger, Benoît and Colombo, Michele and Munoz-Musat, Esteban and Escrichs, Anira and Atzori, Tiziana and Baglio, Francesca and Lapa, Constantin and Berlis, Ansgar and Krüger, Kristina and Luther, Tina and Perlbarg, Vincent and Deco, Gustavo and Sanz-Perl, Yonathan and Tagliazucchi, Enzo and Puybasset, Louis and Rohaut, Benjamin and Naccache, Lionel and Comanducci, Angela and Arzi, Anat and Rosanova, Mario and Bender, Andreas and Sitt, Jacobo Diego},
title = {{Multimodal multicentre investigation of diagnostic and prognostic markers in disorders of consciousness}},
journal = {Brain : a journal of neurology},
year = {2026},
month = apr,
volume = {149},
number = {4},
pages = {1381--1395},
publisher = {Oxford University Press},
issn = {0006-8950},
doi = {10.1093/brain/awaf412},
url = {https://doi.org/10.1093/brain/awaf412},
pmid = {41499248},
pmcid = {PMC13058464}
}

RIS

TY - JOUR
AU - Manasova, Dragana
AU - Belloli, Laouen Mayal Louan
AU - Rosenfelder, Martin Justinus
AU - Willacker, Lina
AU - Fló Rama, Emilia
AU - Valota, Chiara
AU - Hermann, Bertrand
AU - Kaufmann, Brigitte Charlotte
AU - Pirastru, Alice
AU - Derchi, Chiara Camilla
AU - Raiser, Theresa
AU - Valente, Melanie
AU - Sangare, Aude
AU - Türker, Başak
AU - Pyatigorskaya, Nadya
AU - Béranger, Benoît
AU - Colombo, Michele
AU - Munoz-Musat, Esteban
AU - Escrichs, Anira
AU - Atzori, Tiziana
AU - Baglio, Francesca
AU - Lapa, Constantin
AU - Berlis, Ansgar
AU - Krüger, Kristina
AU - Luther, Tina
AU - Perlbarg, Vincent
AU - Deco, Gustavo
AU - Sanz-Perl, Yonathan
AU - Tagliazucchi, Enzo
AU - Puybasset, Louis
AU - Rohaut, Benjamin
AU - Naccache, Lionel
AU - Comanducci, Angela
AU - Arzi, Anat
AU - Rosanova, Mario
AU - Bender, Andreas
AU - Sitt, Jacobo Diego
TI - Multimodal multicentre investigation of diagnostic and prognostic markers in disorders of consciousness
T2 - Brain : a journal of neurology
J2 - Brain
PY - 2026
DA - 2026/04/01
VL - 149
IS - 4
SP - 1381
EP - 1395
SN - 0006-8950
PB - Oxford University Press
DO - 10.1093/brain/awaf412
UR - https://doi.org/10.1093/brain/awaf412
LA - en
ER -

CSL-JSON

{
"id": "10.1093/brain/awaf412",
"type": "article-journal",
"title": "Multimodal multicentre investigation of diagnostic and prognostic markers in disorders of consciousness",
"container-title": "Brain : a journal of neurology",
"author": [
{
"family": "Manasova",
"given": "Dragana"
},
{
"family": "Belloli",
"given": "Laouen Mayal Louan"
},
{
"family": "Rosenfelder",
"given": "Martin Justinus"
},
{
"family": "Willacker",
"given": "Lina"
},
{
"family": "Fló Rama",
"given": "Emilia"
},
{
"family": "Valota",
"given": "Chiara"
},
{
"family": "Hermann",
"given": "Bertrand"
},
{
"family": "Kaufmann",
"given": "Brigitte Charlotte"
},
{
"family": "Pirastru",
"given": "Alice"
},
{
"family": "Derchi",
"given": "Chiara Camilla"
},
{
"family": "Raiser",
"given": "Theresa"
},
{
"family": "Valente",
"given": "Melanie"
},
{
"family": "Sangare",
"given": "Aude"
},
{
"family": "Türker",
"given": "Başak"
},
{
"family": "Pyatigorskaya",
"given": "Nadya"
},
{
"family": "Béranger",
"given": "Benoît"
},
{
"family": "Colombo",
"given": "Michele"
},
{
"family": "Munoz-Musat",
"given": "Esteban"
},
{
"family": "Escrichs",
"given": "Anira"
},
{
"family": "Atzori",
"given": "Tiziana"
},
{
"family": "Baglio",
"given": "Francesca"
},
{
"family": "Lapa",
"given": "Constantin"
},
{
"family": "Berlis",
"given": "Ansgar"
},
{
"family": "Krüger",
"given": "Kristina"
},
{
"family": "Luther",
"given": "Tina"
},
{
"family": "Perlbarg",
"given": "Vincent"
},
{
"family": "Deco",
"given": "Gustavo"
},
{
"family": "Sanz-Perl",
"given": "Yonathan"
},
{
"family": "Tagliazucchi",
"given": "Enzo"
},
{
"family": "Puybasset",
"given": "Louis"
},
{
"family": "Rohaut",
"given": "Benjamin"
},
{
"family": "Naccache",
"given": "Lionel"
},
{
"family": "Comanducci",
"given": "Angela"
},
{
"family": "Arzi",
"given": "Anat"
},
{
"family": "Rosanova",
"given": "Mario"
},
{
"family": "Bender",
"given": "Andreas"
},
{
"family": "Sitt",
"given": "Jacobo Diego"
}
],
"container-title-short": "Brain",
"volume": "149",
"issue": "4",
"page": "1381-1395",
"DOI": "10.1093/brain/awaf412",
"PMID": "41499248",
"PMCID": "PMC13058464",
"ISSN": "0006-8950",
"publisher": "Oxford University Press",
"URL": "https://doi.org/10.1093/brain/awaf412",
"language": "en",
"issued": {
"date-parts": [
[
2026,
4,
1
]
]
}
}

The tracing map gets a citation of its own once an author has validated it and it has a DOI.

Similar papers

The papers with a page that share the most with this one: the tools found in their code, their categories, datasets, cited references and authors, the rarest counting most.

[1] doi:10.21203/rs.3.rs-9914920/v1 [code]
Prediction of cognitive performance by demographics, sleep, and brain morphometry: machine learning findings from ENIGMA-Sleep Working Group
Journal: Research Square (preprint)
In common: neuromaps, statannotations, XGBoost, 9 other tools, structural MRI / diffusion
[2] doi:10.1007/s00234-026-04103-8 [code]
Enhanced detection of subtle cortical abnormalities in focal epilepsy using 7 T MRI surface-based models and graph neural networks.
Journal: Neuroradiology
In common: PyBIDS, statannotations, FreeSurfer, 8 other tools, EEG, structural MRI / diffusion
[3] doi:10.1186/s12967-026-08108-y [code]
An explainable multimodal machine learning model for diagnosing disorders of consciousness: evidence from a large multicenter Chinese cohort.
Journal: Journal of translational medicine
In common: MNE-Python, pandas, SciPy, 1 other tool, EEG, other condition, 6 references
[4] doi:10.64898/2026.08.18.26360725 [code]
Temporal pole blurring in hippocampal sclerosis reflects seizure-disrupted myelination
Journal: medRxiv (preprint)
In common: PyBIDS, statannotations, FreeSurfer, 8 other tools, structural MRI / diffusion
[5] doi:10.1038/s41398-026-04157-5 [code]
Association of glymphatic function with 40-Hz neural oscillations, systemic metabolic markers, and cognitive performance in healthy aging adults: An EEG and MRI study.
Journal: Translational psychiatry
In common: fMRIPrep, PyBIDS, FreeSurfer, 7 other tools, EEG, structural MRI / diffusion
[6] doi:10.1093/nc/niag043 [code]
Demographics-robust spontaneous eye blinking slowing in patients with severe acquired brain injury.
Journal: Neuroscience of consciousness
In common: statannotations, MNE-Python, seaborn, 5 other tools, other condition, 3 references
[7] doi:10.21203/rs.3.rs-9326213/v1 [code]
Multi-task fMRI outperforms resting-state fMRI for revealing task-invariant organization of the human brain
Journal: Research Square (preprint)
In common: neuromaps, statannotations, FreeSurfer, 8 other tools
[8] doi:10.1162/imag.a.1269 [code]
From early to contemporary normative modeling: Mapping individual differences in neurophysiological signals.
Journal: Imaging neuroscience (Cambridge, Mass.)
In common: specparam (formerly FOOOF), FreeSurfer, MNE-Python, 8 other tools, EEG
[9] doi:10.1162/imag.a.1347 [code]
Neural and behavioural correlates of theory of mind reasoning in five-year-old children born preterm.
Journal: Imaging neuroscience (Cambridge, Mass.)
In common: fMRIPrep, PyBIDS, FreeSurfer, 7 other tools, other condition
[10] doi:10.1038/s41586-026-10631-3 [code]
A prognostic human brain network for diffuse midline glioma.
Journal: Nature
In common: neuromaps, FreeSurfer, Nilearn, 7 other tools, other condition, 1 reference

Contribute

The authors of this paper can claim it, correct its record and validate its tracing map, and the maintainers of its code (its owner, or a public member of its organization) correct what it says of their repository; anyone signed in can ask for its removal. Every request goes to OSCR's own machine, which answers it; your account page follows them.

Sign in with ORCID to claim this paper as one of its authors, correct its record or validate its tracing map: when the paper's metadata lists your ORCID iD, you are recognized at once. Maintainers of its code: sign in with GitHub, then claim the repository on your account page.

Request its removal

To ask OSCR to remove this record, the copies of its authors' scripts or its tracing map, use the removal request page: signed in, you say who you are, what to remove and why, then review and confirm the request. Published rules decide every request (how).

Discussion, reproductions, activity

Discussion: questions and error reports about this paper and its code, from signed-in readers and its authors. It opens with sign-in.

Reproductions: reports from readers who ran the authors' code: what they reproduced, with which environment, commit and data. It opens with sign-in.

Activity: what happens around this paper: new versions of its record, its map's validation, discussions and reproductions. It opens with sign-in.