forked from neurodatascience/QLS-course-materials
-
Notifications
You must be signed in to change notification settings - Fork 0
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
added python script version of the ABIDE fmri classification task
- Loading branch information
Showing
3 changed files
with
176 additions
and
29 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
158 changes: 158 additions & 0 deletions
158
Lectures/07-Machine_Learning_1/code/ML_classification.py
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,158 @@ | ||
| ## Imports | ||
| from nilearn import datasets | ||
| from nilearn.connectome import ConnectivityMeasure | ||
| import pandas as pd | ||
| import numpy as np | ||
| import argparse | ||
| from sklearn import preprocessing | ||
| from sklearn.model_selection import train_test_split | ||
| from sklearn.linear_model import LogisticRegression | ||
| from sklearn.ensemble import RandomForestClassifier | ||
| from sklearn.metrics import precision_recall_fscore_support | ||
| from sklearn.metrics import confusion_matrix | ||
|
|
||
|
|
||
| def extract_connectome_features(func_data, measure): | ||
| ''' A function to calculate connnectome based on timeseries data and similarity measure | ||
| ''' | ||
| connectome_matrix = measure.fit_transform([func_data])[0] | ||
| tril_idx = np.tril_indices(len(connectome_matrix),k=-1) | ||
| flat_features = connectome_matrix[tril_idx] | ||
|
|
||
| return flat_features | ||
|
|
||
| def load_data(n_subjects,parcel,data_dir): | ||
| ''' Reads data from local directory or nilearn dataset | ||
| ''' | ||
| data = datasets.fetch_abide_pcp(n_subjects=n_subjects,derivatives=[parcel],data_dir=data_dir) | ||
| pheno = pd.DataFrame(data['phenotypic']).drop(columns=['i','Unnamed: 0']) | ||
|
|
||
| return data, pheno | ||
|
|
||
| def get_train_test_splits(X,y,test_subset_fraction=0.2): | ||
| ''' Splits samples into a single train-test split | ||
| ''' | ||
| stratification = y | ||
|
|
||
| X_train, X_test, y_train, y_test = train_test_split( | ||
| X, # input features | ||
| y, # output labels | ||
| test_size = test_subset_fraction, | ||
| shuffle = True, # shuffle dataset | ||
| # before splitting | ||
| stratify = stratification, | ||
| random_state = 123 # same shuffle each time | ||
| ) | ||
|
|
||
| # print the size of our training and test groups | ||
| print('training:', len(X_train), 'testing:', len(X_test)) | ||
|
|
||
| return X_train, X_test, y_train, y_test | ||
|
|
||
| def run(n_subjects,parcel,data_dir,task, model): | ||
| ''' Setup and run ML tasks | ||
| ''' | ||
| print("-"*25) | ||
| print("Loading data") | ||
| print("-"*25) | ||
| data, pheno = load_data(n_subjects,parcel,data_dir) | ||
|
|
||
| # Imaging variables | ||
| features = data[parcel] | ||
| print(f'Number of samples: {len(features)}') | ||
| subject_feature_shape = features[0].shape | ||
| n_rois = subject_feature_shape[1] | ||
| print(f'subject_feature_shape: {subject_feature_shape}') | ||
|
|
||
| # preprocess fmri data (flatten connectome) | ||
| print("-"*25) | ||
| print("Flattening the connectome matrix") | ||
| print("-"*25) | ||
| correlation_measure = ConnectivityMeasure(kind='correlation') | ||
|
|
||
| print(f"Extracting lower triangle values from {n_rois}x{n_rois} connectivity matrix") | ||
| flat_features_list = [] | ||
| for func_data in features: | ||
| flat_features = extract_connectome_features(func_data, correlation_measure) | ||
| flat_features_list.append(flat_features) | ||
|
|
||
| # setup X,y for ML model | ||
| print("-"*25) | ||
| print("Setting up X and y for the ML model") | ||
| print("-"*25) | ||
| X = np.array(flat_features_list) | ||
| print(f'Input data (X) shape: {X.shape}') | ||
|
|
||
| y = pheno[task] | ||
| y_counts = y.value_counts() | ||
|
|
||
| print(f'Unique output clasess:\n{y_counts}') | ||
|
|
||
| # Encode labels to integer categories | ||
| le = preprocessing.LabelEncoder() | ||
| y = le.fit_transform(y) | ||
|
|
||
| # Get a single train-test split (80/20) | ||
| X_train, X_test, y_train, y_test = get_train_test_splits(X,y) | ||
|
|
||
| # train model | ||
| if model == 'RF': | ||
| clf = RandomForestClassifier(max_depth=3, class_weight='balanced', random_state=0) | ||
| elif model == 'LR': | ||
| clf = LogisticRegression(penalty='l1', C=1, class_weight='balanced', solver='saga', random_state=0) | ||
| else: | ||
| print(f'Unknown model: {model}') | ||
|
|
||
| if model in ["RF", "LR"]: | ||
| print("-"*25) | ||
| print("Training {model} model") | ||
| print("-"*25) | ||
| clf.fit(X_train, y_train) | ||
| train_acc = clf.score(X_train, y_train) | ||
| print(f'train acc: {train_acc:.3f}') | ||
|
|
||
| # Evaluate on a test set | ||
| y_pred = clf.predict(X_test) | ||
| test_acc = clf.score(X_test, y_test) | ||
| print(f'test acc: {test_acc:.3f}') | ||
|
|
||
| print("-"*25) | ||
| print("Other useful performance metrics:") | ||
| print("-"*25) | ||
| test_cm = confusion_matrix(y_test, y_pred) | ||
| print(f"Confusion matrix:\n{test_cm}") | ||
| p,r,f1,_ = precision_recall_fscore_support(y_test, y_pred, average='weighted') | ||
|
|
||
| print(f'precision: {p:.2f}, recall: {r:.2f}, f1: {f1:.2f}') | ||
|
|
||
|
|
||
| if __name__ == '__main__': | ||
| # argparse | ||
| HELPTEXT = """ | ||
| Script version of the classification tutorial (diagnosis or scan-site) using ABIDE dataset | ||
| """ | ||
| parser = argparse.ArgumentParser(description=HELPTEXT) | ||
|
|
||
| parser.add_argument('--n_subjects', type=int, default=100, help='number of subjects to download') | ||
| parser.add_argument('--parcel', type=str, default="rois_ho", help='parcellation for connectome (rois_ho or rois_aal)') | ||
| parser.add_argument('--data_dir', type=str, default="./", help='data dir for previously downloaded data') | ||
| parser.add_argument('--task', type=str, default="DX_GROUP", help='ML classification task (DX_GROUP or SITE_ID)') | ||
| parser.add_argument('--model', type=str, default="RF", help='ML model to use (RF or LR)') | ||
|
|
||
| args = parser.parse_args() | ||
|
|
||
| n_subjects = args.n_subjects | ||
| parcel = args.parcel | ||
| data_dir = args.data_dir | ||
| task = args.task | ||
| model = args.model | ||
|
|
||
| print("-"*50) | ||
| print(f"Performing {task} classification task using {model} model with {n_subjects} subjects and {parcel} parcellation") | ||
| print("-"*50) | ||
|
|
||
| run(n_subjects,parcel,data_dir,task,model) | ||
|
|
||
| print("-"*50) | ||
| print(f"Analysis completed for {task} classification task using {model}!") | ||
| print("-"*50) |