def support_vector_machine(self, sensors_set):
features = list(self.dataset.get_sensors_set_features(sensors_set))
print("SUPPORT VECTOR MACHINE.....")
print("CLASSIFICATION BASED ON THESE SENSORS: ",
self.dataset.get_remained_sensors(sensors_set))
print("NUMBER OF FEATURES: ", len(features))
train_features, train_classes, test_features, test_classes = self.__get_sets_for_classification(
self.dataset.get_train, self.dataset.get_test, features)
train_features_scaled, test_features_scaled = util.scale_features(
train_features, test_features)
classifier_svm = SVC(C=const.PAR_SVM_C[sensors_set],
gamma=const.PAR_SVM_GAMMA[sensors_set],
verbose=False)
classifier_svm.fit(train_features_scaled, train_classes)
test_prediction = classifier_svm.predict(test_features_scaled)
acc = accuracy_score(test_classes, test_prediction)
print("ACCURACY : " + str(acc))
print("END SUPPORT VECTOR MACHINE.....")
if not os.path.exists(const.DIR_RESULTS):
os.makedirs(const.DIR_RESULTS)
file_content = "acc\n" + str(acc)
with open(
const.DIR_RESULTS + "/" + str(sensors_set) +
const.FILE_SUPPORT_VECTOR_MACHINE_RESULTS, 'w') as f:
f.write(file_content)
def neural_network(self, sensors_set):
features = list(self.dataset.get_sensors_set_features(sensors_set))
print("NEURAL NETWORK.....")
print("CLASSIFICATION BASED ON THESE SENSORS: ",
self.dataset.get_remained_sensors(sensors_set))
print("NUMBER OF FEATURES: ", len(features))
train_features, train_classes, test_features, test_classes = self.__get_sets_for_classification(
self.dataset.get_train, self.dataset.get_test, features)
train_features_scaled, test_features_scaled = util.scale_features(
train_features, test_features)
classifier_nn = MLPClassifier(
hidden_layer_sizes=(const.PAR_NN_NEURONS[sensors_set], ),
alpha=const.PAR_NN_ALPHA[sensors_set],
max_iter=const.PAR_NN_MAX_ITER,
tol=const.PAR_NN_TOL)
classifier_nn.fit(train_features_scaled, train_classes)
test_prediction = classifier_nn.predict(test_features_scaled)
acc = accuracy_score(test_classes, test_prediction)
print("ACCURACY : " + str(acc))
print("END NEURAL NETWORK")
if not os.path.exists(const.DIR_RESULTS):
os.makedirs(const.DIR_RESULTS)
file_content = "acc\n" + str(acc)
with open(
const.DIR_RESULTS + "/" + str(sensors_set) +
const.FILE_NEURAL_NETWORK_RESULTS, 'w') as f:
f.write(file_content)
def main(result_dir: str, data_atlas_dir: str, data_train_dir: str, data_test_dir: str):
"""Brain tissue segmentation using logistic regression.
The main routine executes the medical image analysis pipeline:
- Image loading
- Registration
- Pre-processing
- Feature extraction
- Logistic regression classifier model building
- Segmentation using logistic regression on unseen images
- Post-processing of the segmentation
- Evaluation of the segmentation
"""
# load atlas images
putil.load_atlas_images(data_atlas_dir)
print('-' * 5, 'Training...')
# load feature matrix and label vector
# precomputed by preprocessAndStore.py
file_id = open('data_train.pckl', 'rb')
data_train = pickle.load(file_id)
file_id.close()
file_id = open('labels_train.pckl', 'rb')
labels_train = pickle.load(file_id)
file_id.close()
##########################################
# perform a grid search over the parameter grid and choose the optimal parameters
param_grid = {'C': [0.5, 1, 2.5, 50, 1000]} # grid to search for best parameter C = 0.02
log_reg_classifier = model_selection.GridSearchCV(sk.LogisticRegression(class_weight='balanced')
, param_grid, refit=True)
print('abschnitt 1')
data_train_scaled, scaler = util.scale_features(data_train)
start_time = timeit.default_timer()
log_reg_classifier.fit(data_train_scaled, labels_train)
util.print_feature_importance(log_reg_classifier.best_estimator_.coef_)
util.print_class_count(labels_train)
print('abschnitt 2')
#print("importance of features: ", log_reg_classifier.best_estimator_.coef_)
print("best estimator: ", log_reg_classifier.best_estimator_)
print("best parameter: ", log_reg_classifier.best_params_)
# store trained log_regr
file_id = open('log_regr.pckl', 'wb')
pickle.dump(log_reg_classifier, file_id)
file_id.close()
file_id = open('scaler.pckl', 'wb')
pickle.dump(scaler, file_id)
file_id.close()
print(' Time elapsed:', timeit.default_timer() - start_time, 's')
def main(result_dir: str, data_atlas_dir: str, data_train_dir: str,
data_test_dir: str):
"""Brain tissue segmentation using decision forests.
The main routine executes the medical image analysis pipeline:
- Segmentation using the decision forest classifier model on unseen images
- Post-processing of the segmentation
- Evaluation of the segmentation
"""
# load atlas images
putil.load_atlas_images(data_atlas_dir)
# crawl the training image directories
crawler = load.FileSystemDataCrawler(data_train_dir, IMAGE_KEYS,
futil.BrainImageFilePathGenerator(),
futil.DataDirectoryFilter())
pre_process_params = {
'zscore_pre': True,
'registration_pre': False,
'coordinates_feature': True,
'intensity_feature': True,
'gradient_intensity_feature': True,
'second_oder_coordinate_feature': False,
'label_percentages': [0.0003, 0.004, 0.003, 0.04, 0.04, 0.02]
} #[0.0005, 0.005, 0.005, 0.05, 0.09, 0.022]}
print('-' * 5, 'Testing...')
# load classifier
file_id = open('svm_rbf_fullset_C15_G5_lotofpointspersample.pckl', 'rb')
svm_rbf_classifier = pickle.load(file_id)
file_id.close()
file_id = open('scaler_rbf_fullset_C15_G5_lotofpointspersample.pckl', 'rb')
scaler = pickle.load(file_id)
file_id.close()
# initialize evaluator
evaluator = putil.init_evaluator(result_dir)
# crawl the training image directories
crawler = load.FileSystemDataCrawler(data_test_dir, IMAGE_KEYS,
futil.BrainImageFilePathGenerator(),
futil.DataDirectoryFilter())
# load images for testing and pre-process
pre_process_params['training'] = False
images_test = putil.pre_process_batch(crawler.data,
pre_process_params,
multi_process=False)
images_prediction = []
images_probabilities = []
for img in images_test:
print('-' * 10, 'Testing', img.id_)
scaled_features, s = util.scale_features(img.feature_matrix[0], scaler)
start_time = timeit.default_timer()
predictions = svm_rbf_classifier.predict(scaled_features)
#probabilities = svm_classifier.predict_proba(img.feature_matrix[0])
#predictions = forest.predict(img.feature_matrix[0])
#probabilities = forest.predict_proba(img.feature_matrix[0])
print(' Time elapsed:', timeit.default_timer() - start_time, 's')
# convert prediction and probabilities back to SimpleITK images
image_prediction = conversion.NumpySimpleITKImageBridge.convert(
predictions.astype(np.uint8), img.image_properties)
#image_probabilities = conversion.NumpySimpleITKImageBridge.convert(probabilities, img.image_properties)
# evaluate segmentation without post-processing
evaluator.evaluate(image_prediction,
img.images[structure.BrainImageTypes.GroundTruth],
img.id_)
images_prediction.append(image_prediction)
#images_probabilities.append(image_probabilities)
# post-process segmentation and evaluate with post-processing
#post_process_params = {'crf_post': False}
#images_post_processed = putil.post_process_batch(images_test, images_prediction, images_probabilities,
# post_process_params, multi_process=True)
for i, img in enumerate(images_test):
# evaluator.evaluate(images_post_processed[i], img.images[structure.BrainImageTypes.GroundTruth],
# img.id_ + '-PP')
# save results
sitk.WriteImage(
images_prediction[i],
os.path.join(
result_dir, images_test[i].id_ +
'_SEG_SVM_fullset_C15-_G5_lotofpointspersample.mha'), True)
def main(result_dir: str, data_atlas_dir: str, data_train_dir: str, data_test_dir: str, ml_method: str, verbose: bool):
"""Brain tissue segmentation using decision forests.
The main routine executes the medical image analysis pipeline:
- Image loading
- Registration
- Pre-processing
- Feature extraction
- Decision forest classifier model building
- Segmentation using the decision forest classifier model on unseen images
- Post-processing of the segmentation
- Evaluation of the segmentation
"""
# load atlas images
putil.load_atlas_images(data_atlas_dir)
print('-' * 5, 'Training '+ ml_method + '...')
# crawl the training image directories
crawler = load.FileSystemDataCrawler(data_train_dir,
IMAGE_KEYS,
futil.BrainImageFilePathGenerator(),
futil.DataDirectoryFilter())
pre_process_params = {'zscore_pre': True,
'registration_pre': False,
'coordinates_feature': True,
'intensity_feature': True,
'gradient_intensity_feature': True,
'second_oder_coordinate_feature': False,
'label_percentages': [0.0005, 0.005, 0.005, 0.05, 0.09, 0.022]}
# load images for training and pre-process
images = putil.pre_process_batch(crawler.data, pre_process_params, multi_process=False)
# generate feature matrix and label vector
data_train = np.concatenate([img.feature_matrix[0] for img in images])
labels_train = np.concatenate([img.feature_matrix[1] for img in images]).squeeze()
if verbose:
util.print_class_count(labels_train)
start_time = timeit.default_timer()
if ml_method == 'random_forest':
classifier = sk_ensemble.RandomForestClassifier(max_features=images[0].feature_matrix[0].shape[1],
n_estimators=20,
max_depth=25)
data_train_scaled = data_train # do not scale features to keep original RF
elif ml_method == 'svm_linear':
classifier = svm.SVC(kernel='linear', C=1, class_weight='balanced')
data_train_scaled, scaler = util.scale_features(data_train)
elif ml_method == 'svm_rbf':
classifier = svm.SVC(kernel='rbf', C=15, gamma=5, class_weight='balanced',
decision_function_shape='ovo')
data_train_scaled, scaler = util.scale_features(data_train)
elif ml_method == 'logistic_regression':
classifier = linear_model.LogisticRegression(class_weight='balanced')
data_train_scaled, scaler = util.scale_features(data_train)
else:
assert False, "No valid segmentation algorithm selected in argument ml_method"
classifier.fit(data_train_scaled, labels_train)
print(' Time elapsed:', timeit.default_timer() - start_time, 's')
# create a result directory with timestamp
t = datetime.datetime.now().strftime('%Y-%m-%d-%H-%M-%S')
result_dir = os.path.join(result_dir, t)
os.makedirs(result_dir, exist_ok=True)
# print and plot feature importance for each structure
if verbose:
if ml_method == 'svm_linear':
util.print_feature_importance(classifier.coef_)
util.plot_feature_importance(classifier.coef_, result_dir)
if ml_method == 'random_forest':
util.print_feature_importance(classifier.feature_importances_)
util.plot_feature_importance(classifier.feature_importances_, result_dir)
print('-' * 5, 'Testing...')
# initialize evaluator
evaluator = putil.init_evaluator(result_dir)
# crawl the training image directories
crawler = load.FileSystemDataCrawler(data_test_dir,
IMAGE_KEYS,
futil.BrainImageFilePathGenerator(),
futil.DataDirectoryFilter())
# load images for testing and pre-process
pre_process_params['training'] = False
images_test = putil.pre_process_batch(crawler.data, pre_process_params, multi_process=True)
images_prediction = []
images_probabilities = []
for img in images_test:
print('-' * 10, 'Testing', img.id_)
start_time = timeit.default_timer()
if ml_method == 'random_forest':
scaled_features = img.feature_matrix[0]
else:
scaled_features, s = util.scale_features(img.feature_matrix[0], scaler)
predictions = classifier.predict(scaled_features)
print(' Time elapsed:', timeit.default_timer() - start_time, 's')
# convert prediction and probabilities back to SimpleITK images
image_prediction = conversion.NumpySimpleITKImageBridge.convert(predictions.astype(np.uint8),
img.image_properties)
probabilities = classifier.predict_proba(scaled_features)
image_probabilities = conversion.NumpySimpleITKImageBridge.convert(probabilities, img.image_properties)
images_probabilities.append(image_probabilities)
# evaluate segmentation without post-processing
evaluator.evaluate(image_prediction, img.images[structure.BrainImageTypes.GroundTruth], img.id_)
images_prediction.append(image_prediction)
# post-process segmentation and evaluate with post-processing
post_process_params = {'crf_post': False}
images_post_processed = putil.post_process_batch(images_test, images_prediction, images_probabilities,
post_process_params, multi_process=True)
for i, img in enumerate(images_test):
evaluator.evaluate(images_post_processed[i], img.images[structure.BrainImageTypes.GroundTruth],
img.id_ + '-PP')
# save results
sitk.WriteImage(images_prediction[i], os.path.join(result_dir, images_test[i].id_ + '_SEG.mha'), True)
sitk.WriteImage(images_post_processed[i], os.path.join(result_dir, images_test[i].id_ + '_SEG-PP.mha'), True)
def main(result_dir: str, data_atlas_dir: str, data_train_dir: str,
data_test_dir: str):
"""Brain tissue segmentation using decision forests.
The main routine executes the medical image analysis pipeline:
- Image loading
- Registration
- Pre-processing
- Feature extraction
- Decision forest classifier model building
- Segmentation using the decision forest classifier model on unseen images
- Post-processing of the segmentation
- Evaluation of the segmentation
"""
# load atlas images
putil.load_atlas_images(data_atlas_dir)
print('-' * 5, 'Training...')
# load feature matrix and label vector
# precomputed by preprocessAndStore.py
file_id = open('data_train.pckl', 'rb')
data_train = pickle.load(file_id)
file_id.close()
file_id = open('labels_train.pckl', 'rb')
labels_train = pickle.load(file_id)
file_id.close()
##########################################
# use if GridSearchCV is used
# perform a grid search over the parameter grid and choose the optimal parameters
#Cs = [10, 12, 15, 20]#a list best = 15
#gammas = [1 ,2 ,3, 5,10]#a list best = 10
#param_grid = {'C': Cs, 'gamma': gammas}#a dictionary
#svm_rbf_classifier = model_selection.GridSearchCV(svm.SVC(kernel='rbf'), param_grid, verbose=1)
data_train_scaled, scaler = util.scale_features(data_train)
# printing out how much labels of each group were taken by the mask
util.print_class_count(labels_train)
# use if GridSearchCV is not used
svm_rbf_classifier = svm.SVC(kernel='rbf',
C=15,
gamma=10,
class_weight='balanced',
decision_function_shape='ovo')
start_time = timeit.default_timer()
print("start training")
# for position features only: svm_rbf_classifier.fit(data_train_scaled[:, 0:3], labels_train)
svm_rbf_classifier.fit(data_train_scaled, labels_train)
#####svm_rbf_classifier.coef_ can not be used with rbf kernel
#util.print_feature_importance(svm_rbf_classifier.best_estimator_.coef_)
#use if GridSearchCV is used
#print("importance of features: ", svm_rbf_classifier.best_estimator_.coef_)#####svm_rbf_classifier.coef_ can not be used with rbf kernel
#print("best estimator: ", svm_rbf_classifier.best_estimator_)
#print("best parameter: ", svm_rbf_classifier.best_params_)
#use if GridSearchCV is not used
print("best estimator: ", svm_rbf_classifier)
print("estimator dual_coef_: ", svm_rbf_classifier.dual_coef_)
file_id = open('svm_rbf_fullset_C15_G5.pckl', 'wb')
pickle.dump(svm_rbf_classifier, file_id)
file_id.close()
file_id = open('scaler_rbf_fullset_C15_G5.pckl', 'wb')
pickle.dump(scaler, file_id)
file_id.close()
print(' Time elapsed:', timeit.default_timer() - start_time, 's')
def main(result_dir: str, data_atlas_dir: str, data_train_dir: str, data_test_dir: str):
"""Brain tissue segmentation using decision forests.
The main routine executes the medical image analysis pipeline:
- Image loading
- Registration
- Pre-processing
- Feature extraction
- Decision forest classifier model building
- Segmentation using the decision forest classifier model on unseen images
- Post-processing of the segmentation
- Evaluation of the segmentation
"""
# load atlas images
putil.load_atlas_images(data_atlas_dir)
print('-' * 5, 'Training...')
# load feature matrix and label vector
# precomputed by preprocessAndStore.py1057278
file_id = open('data_train_reduced2.pckl', 'rb')
data_train = pickle.load(file_id)
file_id.close()
file_id = open('labels_train_reduced.pckl', 'rb')
labels_train = pickle.load(file_id)
file_id.close()
##########################################
# perform a grid search over the parameter grid and choose the optimal parameters
param_grid = {'C': [ 2, 3, 4, 5, 10, 20, 100]} # grid to search for best parameter C = 0.02
#svm_classifier = model_selection.GridSearchCV(svm.LinearSVC(C=1, class_weight='balanced', dual=False), param_grid, verbose=1)
data_train_scaled, scaler = util.scale_features(data_train)
util.print_class_count(labels_train)
# use balanced class weights to include classes with small sample size
# solve the primal problem since n_features < n_samples
svm_classifier = svm.LinearSVC(C=1, class_weight='balanced', dual=False) # probability=False, kernel= 'rbf') #kernel='linear')
start_time = timeit.default_timer()
svm_classifier.fit(data_train_scaled, labels_train)
#util.print_feature_importance(svm_classifier.coef_)
util.plot_feature_importance(svm_classifier.coef_)
#print(svm_classifier.best_params_)
#print(svm_classifier.best_estimator_)
# store trained SVM
file_id = open('svm_linear.pckl', 'wb')
pickle.dump(svm_classifier, file_id)
file_id.close()
file_id = open('scaler.pckl', 'wb')
pickle.dump(scaler, file_id)
file_id.close()
print(' Time elapsed:', timeit.default_timer() - start_time, 's')