def test_spatiotemporal(path, subjects, conf_file, type_, **kwargs): conf = read_configuration(path, conf_file, type_) for arg in kwargs: conf[arg] = kwargs[arg] if arg == 'balance': balance = kwargs[arg] total_results = dict() data_path = conf['data_path'] conf['analysis_type'] = 'spatiotemporal' conf['analysis_task'] = type_ for subj in subjects: try: ds = load_dataset(data_path, subj, type_, **conf) except Exception, err: print err continue ds = detrend_dataset(ds, type_, **conf) if 'balance' in locals() and balance == True: if conf['label_included'] == 'all' and \ conf['label_dropped'] == 'none': ds = balance_dataset_timewise(ds, 'fixation') r = spatiotemporal(ds, **conf) total_results[subj] = r
def test_spatiotemporal(path, subjects, conf_file, type_, **kwargs): conf = read_configuration(path, conf_file, type_) for arg in kwargs: conf[arg] = kwargs[arg] if arg == 'balance': balance = kwargs[arg] total_results = dict() data_path = conf['data_path'] conf['analysis_type'] = 'spatiotemporal' conf['analysis_task'] = type_ for subj in subjects: try: ds = load_dataset(data_path, subj, type_, **conf) except Exception, err: print err continue ds = detrend_dataset(ds, type_, **conf) if 'balance' in locals() and balance == True: if conf['label_included'] == 'all' and \ conf['label_dropped'] == 'none': ds = balance_dataset_timewise(ds, 'fixation') r = spatiotemporal(ds, **conf) total_results[subj] = r
def __init__(self, name="searchlight", **kwargs): """ This method is used to set up the configuration of the analysis. """ self.name = name self._default_conf = { 'path':'/home/robbis/fmri/memory/', # Data 'configuration_file':"memory.conf", # Data "project":"carlo_memory", # Data "partecipants": "subjects.csv", # Data+Analysis 'data_type': 'BETA_MVPA', # Data 'n_folds':3, # Analysis "condition_names":["evidence", "task"], # Data+Analysis 'evidence': 3, # Default_value (memory : 3) # Data+Analysis 'task':'decision', # Data 'split_attr':'subject', # Analysis 'mask_area':'intersect', # Data 'normalization':'both', # Analysis "radius": 3, # Analysis "n_balanced_ds": 1, # Analysis "set_targets": carlo_memory_set_targets, # Data+Analysis (Outdated) "classifier":LinearCSVMC(C=1) # Analysis (Not proper) } self._default_conf.update(**kwargs) for k, v in self._default_conf.iteritems(): setattr(self, "_"+k, v) # Setting default fields # if __debug__: debug.active += ["SLC"] self._conf = read_configuration(self._path, self._configuration_file, self._data_type) self._conf['analysis_type'] = 'searchlight' self._conf['analysis_task'] = self._project # Avoid non serializable objects saving self._default_conf.pop("set_targets") self._default_conf.pop("classifier") self._conf.update(**self._default_conf) self._data_path = self._conf['data_path'] self.result_dict = dict() self.maps = [] self._subjects, self._extra_sa = load_subject_file(os.path.join(self._path, self._partecipants))
def test_clustering(path, subjects, analysis, conf_file, source='task', **kwargs): if source == 'task': target = 'rest' else: target = 'task' conf_src = read_configuration(path, conf_file, source) conf_tar = read_configuration(path, conf_file, target) ############################################## conf_src['label_included'] = 'all' conf_src['label_dropped'] = 'none' conf_src['mean_samples'] = 'True' ############################################## for arg in kwargs: conf_src[arg] = kwargs[arg] conf_tar[arg] = kwargs[arg] total_results = dict() data_path = conf_src['data_path'] for subj in subjects: try: ds_src = load_dataset(data_path, subj, source, **conf_src) ds_tar = load_dataset(data_path, subj, target, **conf_tar) except Exception, err: print err continue ds_src = detrend_dataset(ds_src, source, **conf_src) ds_tar = detrend_dataset(ds_tar, target, **conf_tar) if conf_src['label_included'] == 'all' and \ conf_src['label_dropped'] != 'fixation': ds_src = balance_dataset_timewise(ds_src, 'fixation') r = clustering_analysis(ds_src, ds_tar, analysis, **kwargs) total_results[subj] = r
def test_searchlight_cross_decoding(path, subjects, conf_file, type, **kwargs): conf = read_configuration(path, conf_file, type) for arg in kwargs: conf[arg] = kwargs[arg] if arg == 'radius': radius = kwargs[arg] debug.active += ["SLC"] ds_merged = get_merged_ds(path, subjects, conf_file, type, **kwargs) clf = LinearCSVMC(C=1, probability=1, enable_ca=['probabilities']) cv = CrossValidation(clf, NFoldPartitioner(attr='task')) maps = [] for ds in ds_merged: ds.targets[ds.targets == 'point'] = 'face' ds.targets[ds.targets == 'saccade'] = 'place' sl = sphere_searchlight(cv, radius, space = 'voxel_indices') sl_map = sl(ds) sl_map.samples *= -1 sl_map.samples += 1 nif = map2nifti(sl_map, imghdr=ds.a.imghdr) maps.append(nif) datetime = get_time() analysis = 'cross_searchlight' mask = conf['mask_area'] task = type new_dir = datetime+'_'+analysis+'_'+mask+'_'+task command = 'mkdir '+os.path.join(path, '0_results', new_dir) os.system(command) parent_dir = os.path.join(path, '0_results', new_dir) for s, map in zip(subjects, maps): name = s command = 'mkdir '+os.path.join(parent_dir, name) os.system(command) results_dir = os.path.join(parent_dir, name) fname = name+'_radius_'+str(radius)+'_searchlight_map.nii.gz' map.to_filename(os.path.join(results_dir, fname)) return maps
def test_searchlight_cross_decoding(path, subjects, conf_file, type, **kwargs): conf = read_configuration(path, conf_file, type) for arg in kwargs: conf[arg] = kwargs[arg] if arg == 'radius': radius = kwargs[arg] debug.active += ["SLC"] ds_merged = get_merged_ds(path, subjects, conf_file, type, **kwargs) clf = LinearCSVMC(C=1, probability=1, enable_ca=['probabilities']) cv = CrossValidation(clf, NFoldPartitioner(attr='task')) maps = [] for ds in ds_merged: ds.targets[ds.targets == 'point'] = 'face' ds.targets[ds.targets == 'saccade'] = 'place' sl = sphere_searchlight(cv, radius, space='voxel_indices') sl_map = sl(ds) sl_map.samples *= -1 sl_map.samples += 1 nif = map2nifti(sl_map, imghdr=ds.a.imghdr) maps.append(nif) datetime = get_time() analysis = 'cross_searchlight' mask = conf['mask_area'] task = type new_dir = datetime + '_' + analysis + '_' + mask + '_' + task command = 'mkdir ' + os.path.join(path, '0_results', new_dir) os.system(command) parent_dir = os.path.join(path, '0_results', new_dir) for s, map in zip(subjects, maps): name = s command = 'mkdir ' + os.path.join(parent_dir, name) os.system(command) results_dir = os.path.join(parent_dir, name) fname = name + '_radius_' + str(radius) + '_searchlight_map.nii.gz' map.to_filename(os.path.join(results_dir, fname)) return maps
def test_clustering(path, subjects, analysis, conf_file, source='task', **kwargs): if source == 'task': target = 'rest' else: target = 'task' conf_src = read_configuration(path, conf_file, source) conf_tar = read_configuration(path, conf_file, target) ############################################## conf_src['label_included'] = 'all' conf_src['label_dropped'] = 'none' conf_src['mean_samples'] = 'True' ############################################## for arg in kwargs: conf_src[arg] = kwargs[arg] conf_tar[arg] = kwargs[arg] total_results = dict() data_path = conf_src['data_path'] for subj in subjects: try: ds_src = load_dataset(data_path, subj, source, **conf_src) ds_tar = load_dataset(data_path, subj, target, **conf_tar) except Exception, err: print err continue ds_src = detrend_dataset(ds_src, source, **conf_src) ds_tar = detrend_dataset(ds_tar, target, **conf_tar) if conf_src['label_included'] == 'all' and \ conf_src['label_dropped'] != 'fixation': ds_src = balance_dataset_timewise(ds_src, 'fixation') r = clustering_analysis(ds_src, ds_tar, analysis, **kwargs) total_results[subj] = r
def _test_spatial(path, subjects, conf_file, type_, **kwargs): warnings.warn("Deprecated use test_spatial.", DeprecationWarning) conf = read_configuration(path, conf_file, type_) conf['analysis_type'] = 'spatial' conf['analysis_task'] = type_ for arg in kwargs: conf[arg] = kwargs[arg] total_results = dict() data_path = conf['data_path'] summarizers = [rs.DecodingSummarizer()] savers = [rs.DecodingSaver()] result = rs.ResultsCollection(conf, path, summarizers) for subj in subjects: print '------' try: ds = load_dataset(data_path, subj, type_, **conf) except Exception as err: print err continue ds = detrend_dataset(ds, type_, **conf) balancer = balance_dataset(**conf) for i, ds_ in enumerate(balancer.generate(ds)): logger.info("Balanced dataset n. %d" % (i + 1)) subj_ = "%s_%03d" % (subj, i + 1) ds_ = normalize_dataset(ds_, **conf) logger.info(ds_.summary()) r = spatial(ds_, **conf) total_results[subj_] = r subj_result = rs.SubjectResult(subj_, r, savers) result.add(subj_result) #result.save() result.summarize() conf['classes'] = np.unique(ds.targets) #save_results() #save_results(path, total_results, conf) return total_results, subj_result
def _test_spatial(path, subjects, conf_file, type_, **kwargs): warnings.warn("Deprecated use test_spatial.", DeprecationWarning) conf = read_configuration(path, conf_file, type_) conf['analysis_type'] = 'spatial' conf['analysis_task'] = type_ for arg in kwargs: conf[arg] = kwargs[arg] total_results = dict() data_path = conf['data_path'] summarizers = [rs.DecodingSummarizer()] savers = [rs.DecodingSaver()] result = rs.ResultsCollection(conf, path, summarizers) for subj in subjects: print '------' try: ds = load_dataset(data_path, subj, type_, **conf) except Exception, err: print err continue ds = detrend_dataset(ds, type_, **conf) balancer = balance_dataset(**conf) for i, ds_ in enumerate(balancer.generate(ds)): logger.info("Balanced dataset n. %d" % (i+1)) subj_ = "%s_%03d" % (subj, i+1) ds_ = normalize_dataset(ds_, **conf) logger.info(ds_.summary()) r = spatial(ds_, **conf) total_results[subj_] = r subj_result = rs.SubjectResult(subj_, r, savers) result.add(subj_result)
def test_searchlight(path, subjects, conf_file, type_, **kwargs): conf = read_configuration(path, conf_file, type_) for arg in kwargs: conf[arg] = kwargs[arg] conf['analysis_type'] = 'searchlight' conf['analysis_task'] = type_ total_results = dict() data_path = conf['data_path'] # summarizers = [rs.SearchlightSummarizer()] savers = [rs.SearchlightSaver()] result = rs.ResultsCollection(conf, path, summarizers) # for subj in subjects: ds = load_dataset(data_path, subj, type_, **conf) ds = detrend_dataset(ds, type_, **conf) r = searchlight(ds, **kwargs) subj_result = rs.SubjectResult(subj, r, savers) total_results[subj] = r result.add(subj_result) result.summarize() conf['classes'] = np.unique(ds.targets) #save_results() #save_results(path, total_results, conf) return result, r, subj_result
def test_searchlight(path, subjects, conf_file, type_, **kwargs): conf = read_configuration(path, conf_file, type_) for arg in kwargs: conf[arg] = kwargs[arg] conf['analysis_type'] = 'searchlight' conf['analysis_task'] = type_ total_results = dict() data_path = conf['data_path'] # summarizers = [rs.SearchlightSummarizer()] savers = [rs.SearchlightSaver()] result = rs.ResultsCollection(conf, path, summarizers) # for subj in subjects: ds = load_dataset(data_path, subj, type_, **conf) ds = detrend_dataset(ds, type_, **conf) r = searchlight(ds, **kwargs) subj_result = rs.SubjectResult(subj, r, savers) total_results[subj] = r result.add(subj_result) result.summarize() conf['classes'] = np.unique(ds.targets) #save_results() #save_results(path, total_results, conf) return result, r, subj_result
'111123roblan', '111129mirgra', '111202cincal', '111206marant', '111214angfav', '111214cardin', '111220fraarg', '111220martes', '120119maulig', '120126andspo', '120112jaclau'] tasks = ['memory', 'decision'] res = [] result_dict = dict() conf = read_configuration(path, 'memory.conf', 'BETA_MVPA') conf['analysis_type'] = 'searchlight' conf['analysis_task'] = 'memory' summarizers = [rs.SearchlightSummarizer()] savers = [rs.SearchlightSaver()] collection = rs.ResultsCollection(conf, path, summarizers) #for subj in subjects: data_type = 'BETA_MVPA' conf = read_configuration(path, 'memory.conf', data_type) data_path = conf['data_path'] ds_original = load_dataset(data_path, subj, data_type, **conf)
def test_searchlight_similarity(path, subjects, conf_file, type, dim=4, measure=CorrelationThresholdMeasure(), partitioner=TargetCombinationPartitioner(attr='targets'), **kwargs): conf = read_configuration(path, conf_file, type) for arg in kwargs: conf[arg] = kwargs[arg] if arg == 'radius': radius = kwargs[arg] if arg == 'duration': duration = kwargs[arg] ######################################################### datetime = get_time() analysis = 'searchlight_measure' mask = conf['mask_area'] task = type new_dir = datetime+'_'+analysis+'_'+mask+'_'+task command = 'mkdir '+os.path.join(path, '0_results', new_dir) os.system(command) parent_dir = os.path.join(path, '0_results', new_dir) ########################################################## debug.active += ["SLC"] ds_merged = get_merged_ds(path, subjects, conf_file, type, dim=dim, **kwargs) cv = CrossValidation(measure, partitioner, splitter=Splitter(attr='partitions', attr_values=(3,2)), errorfx=None ) maps = [] kwa = dict(voxel_indices=Sphere(radius), event_offsetidx=Sphere(duration)) queryengine = IndexQueryEngine(**kwa) for s, ds in zip(subjects, ds_merged): print ds.samples.shape voxel_num = ds.samples.shape[1]/duration ids = np.arange(voxel_num) sl = Searchlight(cv, queryengine=queryengine, roi_ids=ids) map = sl(ds) maps.append(map) name = s command = 'mkdir '+os.path.join(parent_dir, name) os.system(command) results_dir = os.path.join(parent_dir, name) fname = name+'_radius_'+str(radius)+'_searchlight_map.nii.gz' map_ = ds.a.mapper[:2].reverse(map) map_ = np.rollaxis(map_.samples, 0, 4) img = ni.Nifti1Image(map_, ds.a.imghdr.get_base_affine()) ni.save(img, os.path.join(results_dir, fname)) return maps
def test_transfer_learning(path, subjects, analysis, conf_file, source='task', \ analysis_type='single', calculateSimilarity='True', **kwargs): if source == 'task': target = 'rest' else: if source == 'rest': target = 'task' if source == 'saccade': target = 'face' else: if source == 'face': target = 'saccade' p = kwargs['p'] ############################################## ############################################## ## conf_src['label_included'] = 'all' ## ## conf_src['label_dropped'] = 'none' ## ## conf_src['mean_samples'] = 'False' ## ############################################## ############################################## if analysis_type == 'group': if path.__class__ == conf_file.__class__ == list: ds_src, _, conf_src = sources_merged_ds(path, subjects, conf_file, source, **kwargs) ds_tar, subjects, conf_tar = sources_merged_ds( path, subjects, conf_file, target, **kwargs) conf_src['permutations'] = 0 conf_tar['permutations'] = 0 else: print 'In group analysis path, subjects and conf_file must be lists: \ Check configuration file and/or parameters!!' return 0 else: conf_src = read_configuration(path, conf_file, source) conf_tar = read_configuration(path, conf_file, target) for arg in kwargs: conf_src[arg] = kwargs[arg] conf_tar[arg] = kwargs[arg] data_path = conf_src['data_path'] conf_src['analysis_type'] = 'transfer_learning' conf_src['analysis_task'] = source conf_src['analysis_func'] = analysis.func_name for arg in conf_src: if arg == 'map_list': map_list = conf_src[arg].split(',') if arg == 'p_dist': p = float(conf_src[arg]) print p total_results = dict() summarizers = [ rs.CrossDecodingSummarizer(), rs.SimilaritySummarizer(), rs.DecodingSummarizer(), rs.SignalDetectionSummarizer(), ] savers = [ rs.CrossDecodingSaver(), rs.SimilaritySaver(), rs.DecodingSaver(), rs.SignalDetectionSaver(), ] collection = rs.ResultsCollection(conf_src, path, summarizers) for subj in subjects: print '-------------------' if (len(subjects) > 1) or (subj != 'group'): try: ds_src = load_dataset(data_path, subj, source, **conf_src) ds_tar = load_dataset(data_path, subj, target, **conf_tar) except Exception, err: print err continue # Evaluate if is correct to do further normalization after merging two ds. ds_src = detrend_dataset(ds_src, source, **conf_src) ds_tar = detrend_dataset(ds_tar, target, **conf_tar) if conf_src['label_included'] == 'all' and \ conf_src['label_dropped'] != 'fixation': print 'Balancing dataset...' ds_src = balance_dataset_timewise(ds_src, 'fixation') # Make cross-decoding r = transfer_learning(ds_src, ds_tar, analysis, **conf_src) # Now we have cross-decoding results we could process it pred = np.array(r['classifier'].ca.predictions) targets = r['targets'] c_m = ConfusionMatrix(predictions=pred, targets=targets) c_m.compute() r['confusion_target'] = c_m c_new = cross_decoding_confusion(pred, targets, map_list) r['confusion_total'] = c_new print c_new # Similarity Analysis if calculateSimilarity == 'True': if 'p' not in locals(): print 'Ciao!' mahala_data = similarity_measure(r['ds_tar'], r['ds_src'], r, p_value=p, method='mahalanobis') #r['mahalanobis_similarity'] = mahala_data for k_, v_ in mahala_data.items(): r[k_] = v_ r['confusion_mahala'] = mahala_data['confusion_mahalanobis'] else: #r['mahalanobis_similarity'] = [] r['confusion_mahala'] = 'Null' # Signal Detection Theory Analysis sdt_res = signal_detection_measures(c_new) for k_, v_ in sdt_res.items(): r[k_] = v_ ''' Same code of: r['d_prime'] = d_prime r['beta'] = beta r['c'] = c ''' total_results[subj] = r subj_result = rs.SubjectResult(subj, r, savers=savers) collection.add(subj_result)
def test_transfer_learning(path, subjects, analysis, conf_file, source='task', \ analysis_type='single', calculateSimilarity='True', **kwargs): if source == 'task': target = 'rest' else: if source == 'rest': target = 'task' if source == 'saccade': target = 'face' else: if source == 'face': target = 'saccade' p = kwargs['p'] ############################################## ############################################## ## conf_src['label_included'] = 'all' ## ## conf_src['label_dropped'] = 'none' ## ## conf_src['mean_samples'] = 'False' ## ############################################## ############################################## if analysis_type == 'group': if path.__class__ == conf_file.__class__ == list: ds_src, _, conf_src = sources_merged_ds(path, subjects, conf_file, source, **kwargs) ds_tar, subjects, conf_tar = sources_merged_ds(path, subjects, conf_file, target, **kwargs) conf_src['permutations'] = 0 conf_tar['permutations'] = 0 else: print 'In group analysis path, subjects and conf_file must be lists: \ Check configuration file and/or parameters!!' return 0 else: conf_src = read_configuration(path, conf_file, source) conf_tar = read_configuration(path, conf_file, target) for arg in kwargs: conf_src[arg] = kwargs[arg] conf_tar[arg] = kwargs[arg] data_path = conf_src['data_path'] conf_src['analysis_type'] = 'transfer_learning' conf_src['analysis_task'] = source conf_src['analysis_func'] = analysis.func_name for arg in conf_src: if arg == 'map_list': map_list = conf_src[arg].split(',') if arg == 'p_dist': p = float(conf_src[arg]) print p total_results = dict() summarizers = [rs.CrossDecodingSummarizer(), rs.SimilaritySummarizer(), rs.DecodingSummarizer(), rs.SignalDetectionSummarizer(), ] savers = [rs.CrossDecodingSaver(), rs.SimilaritySaver(), rs.DecodingSaver(), rs.SignalDetectionSaver(), ] collection = rs.ResultsCollection(conf_src, path, summarizers) for subj in subjects: print '-------------------' if (len(subjects) > 1) or (subj != 'group'): try: ds_src = load_dataset(data_path, subj, source, **conf_src) ds_tar = load_dataset(data_path, subj, target, **conf_tar) except Exception, err: print err continue # Evaluate if is correct to do further normalization after merging two ds. ds_src = detrend_dataset(ds_src, source, **conf_src) ds_tar = detrend_dataset(ds_tar, target, **conf_tar) if conf_src['label_included'] == 'all' and \ conf_src['label_dropped'] != 'fixation': print 'Balancing dataset...' ds_src = balance_dataset_timewise(ds_src, 'fixation') # Make cross-decoding r = transfer_learning(ds_src, ds_tar, analysis, **conf_src) # Now we have cross-decoding results we could process it pred = np.array(r['classifier'].ca.predictions) targets = r['targets'] c_m = ConfusionMatrix(predictions=pred, targets=targets) c_m.compute() r['confusion_target'] = c_m c_new = cross_decoding_confusion(pred, targets, map_list) r['confusion_total'] = c_new print c_new # Similarity Analysis if calculateSimilarity == 'True': if 'p' not in locals(): print 'Ciao!' mahala_data = similarity_measure(r['ds_tar'], r['ds_src'], r, p_value=p, method='mahalanobis') #r['mahalanobis_similarity'] = mahala_data for k_,v_ in mahala_data.items(): r[k_] = v_ r['confusion_mahala'] = mahala_data['confusion_mahalanobis'] else: #r['mahalanobis_similarity'] = [] r['confusion_mahala'] = 'Null' # Signal Detection Theory Analysis sdt_res = signal_detection_measures(c_new) for k_,v_ in sdt_res.items(): r[k_] = v_ ''' Same code of: r['d_prime'] = d_prime r['beta'] = beta r['c'] = c ''' total_results[subj] = r subj_result = rs.SubjectResult(subj, r, savers=savers) collection.add(subj_result)
frames = [0] #[1,2,3,4,5,6] mask = ni.load( p.join(path, '1_single_ROIs', 'cluster_conjunction_level-0-003_omnibus-0-001.nii.gz')) subjects, extra_sa = load_subject_file(p.join(path, 'subjects.csv')) target = 'decision' selected_variables = { target: ['L', 'F'], } evidences = [1, 2, 3] # Setup result utilities conf = read_configuration(path, conf_file, task) conf['analysis_type'] = 'searchlight' conf['analysis_task'] = 'temporal_residual' summarizers = [rs.SearchlightSummarizer()] #savers = [rs.DecodingSaver(fields=['classifier', 'stats'])] savers = [rs.SearchlightSaver()] result = rs.ResultsCollection(conf, path, summarizers) for subj in subjects[:1]: conf = read_configuration(path, conf_file, task) # Load dataset ds_orig = load_dataset(path, subj, task,
def get_merged_ds(path, subjects, conf_file, source='task', dim=3, **kwargs): #Mettere source e target nel conf! if source == 'task': target = 'rest' else: if source == 'rest': target = 'task' if source == 'saccade': target = 'face' else: if source == 'face': target = 'saccade' ds_merged_list = [] conf_src = read_configuration(path, conf_file, source) conf_tar = read_configuration(path, conf_file, target) ############################################## ############################################## ## conf_src['label_included'] = 'all' ## ## conf_src['label_dropped'] = 'none' ## ## conf_src['mean_samples'] = 'False' ## ############################################## ############################################## for arg in kwargs: conf_src[arg] = kwargs[arg] conf_tar[arg] = kwargs[arg] data_path = conf_src['data_path'] for subj in subjects: print '--------' try: ds_src = load_dataset(data_path, subj, source, **conf_src) ds_tar = load_dataset(data_path, subj, target, **conf_tar) except Exception, err: print err continue ds_src = detrend_dataset(ds_src, source, **conf_src) ds_tar = detrend_dataset(ds_tar, target, **conf_tar) if dim == 4: duration = np.min([e['duration'] for e in ds_src.a.events]) ds_tar = build_events_ds(ds_tar, duration, overlap=duration-1) ds_src = load_spatiotemporal_dataset(ds_src, duration=duration) print ds_src.samples.shape print ds_tar.samples.shape ds_src.sa['task'] = [source for s in range(ds_src.samples.shape[0])] ds_tar.sa['task'] = [target for s in range(ds_tar.samples.shape[0])] ds_merged = vstack((ds_src, ds_tar)) ds_merged.a.update(ds_src.a) print ds_merged.sa.task ds_merged_list.append(ds_merged) '''
def _group_transfer_learning(path, subjects, analysis, conf_file, source='task', analysis_type='single', **kwargs): if source == 'task': target = 'rest' else: if source == 'rest': target = 'task' if source == 'saccade': target = 'face' else: if source == 'face': target = 'saccade' ############################################## ############################################## ## conf_src['label_included'] = 'all' ## ## conf_src['label_dropped'] = 'none' ## ## conf_src['mean_samples'] = 'False' ## ############################################## ############################################## if analysis_type == 'group': if path.__class__ == conf_file.__class__ == list: ds_src, s, conf_src = sources_merged_ds(path, subjects, conf_file, source, **kwargs) conf_src['permutations'] = 0 else: print 'In group analysis path, subjects and conf_file must be lists: \ Check configuration file and/or parameters!!' return 0 else: conf_src = read_configuration(path, conf_file, source) for arg in conf_src: if arg == 'map_list': map_list = conf_src[arg].split(',') r_group = spatial(ds_src, **conf_src) total_results = dict() total_results['group'] = r_group clf = r_group['classifier'] for subj_, conf_, path_ in zip(subjects, conf_file, path): for subj in subj_: print '-----------' r = dict() if len(subj_) > 1: conf_tar = read_configuration(path_, conf_, target) for arg in kwargs: conf_tar[arg] = kwargs[arg] data_path = conf_tar['data_path'] try: ds_tar = load_dataset(data_path, subj, target, **conf_tar) except Exception, err: print err continue ds_tar = detrend_dataset(ds_tar, target, **conf_tar) if conf_src['label_included'] == 'all' and \ conf_src['label_dropped'] != 'fixation': print 'Balancing dataset...' ds_src = balance_dataset_timewise(ds_src, 'fixation') predictions = clf.predict(ds_tar) pred = np.array(predictions) targets = ds_tar.targets for arg in r_group.keys(): r[arg] = copy.copy(r_group[arg]) r['targets'] = targets r['predictions'] = predictions r['fclf'] = clf c_m = ConfusionMatrix(predictions=pred, targets=targets) c_m.compute() r['confusion_target'] = c_m print c_m tr_pred = similarity_measure_mahalanobis(ds_tar, ds_src, r) r['mahalanobis_similarity'] = tr_pred #print tr_pred c_mat_mahala = ConfusionMatrix(predictions=tr_pred.T[1], targets=tr_pred.T[0]) c_mat_mahala.compute() r['confusion_mahala'] = c_mat_mahala d_prime, beta, c, c_new = signal_detection_measures( pred, targets, map_list) r['d_prime'] = d_prime print d_prime r['beta'] = beta r['c'] = c r['confusion_total'] = c_new ''' d_prime_maha, c_new_maha = d_prime_statistics(tr_pred.T[1], tr_pred.T[0], map_list) r['d_prime_maha'] = d_prime_maha r['confusion_tot_maha'] = c_new_maha ''' total_results[subj] = r
def get_merged_ds(path, subjects, conf_file, source='task', dim=3, **kwargs): #Mettere source e target nel conf! if source == 'task': target = 'rest' else: if source == 'rest': target = 'task' if source == 'saccade': target = 'face' else: if source == 'face': target = 'saccade' ds_merged_list = [] conf_src = read_configuration(path, conf_file, source) conf_tar = read_configuration(path, conf_file, target) ############################################## ############################################## ## conf_src['label_included'] = 'all' ## ## conf_src['label_dropped'] = 'none' ## ## conf_src['mean_samples'] = 'False' ## ############################################## ############################################## for arg in kwargs: conf_src[arg] = kwargs[arg] conf_tar[arg] = kwargs[arg] data_path = conf_src['data_path'] for subj in subjects: print '--------' try: ds_src = load_dataset(data_path, subj, source, **conf_src) ds_tar = load_dataset(data_path, subj, target, **conf_tar) except Exception, err: print err continue ds_src = detrend_dataset(ds_src, source, **conf_src) ds_tar = detrend_dataset(ds_tar, target, **conf_tar) if dim == 4: duration = np.min([e['duration'] for e in ds_src.a.events]) ds_tar = build_events_ds(ds_tar, duration, overlap=duration - 1) ds_src = load_spatiotemporal_dataset(ds_src, duration=duration) print ds_src.samples.shape print ds_tar.samples.shape ds_src.sa['task'] = [source for s in range(ds_src.samples.shape[0])] ds_tar.sa['task'] = [target for s in range(ds_tar.samples.shape[0])] ds_merged = vstack((ds_src, ds_tar)) ds_merged.a.update(ds_src.a) print ds_merged.sa.task ds_merged_list.append(ds_merged) '''
def test_searchlight_similarity( path, subjects, conf_file, type, dim=4, measure=CorrelationThresholdMeasure(), partitioner=TargetCombinationPartitioner(attr='targets'), **kwargs): conf = read_configuration(path, conf_file, type) for arg in kwargs: conf[arg] = kwargs[arg] if arg == 'radius': radius = kwargs[arg] if arg == 'duration': duration = kwargs[arg] ######################################################### datetime = get_time() analysis = 'searchlight_measure' mask = conf['mask_area'] task = type new_dir = datetime + '_' + analysis + '_' + mask + '_' + task command = 'mkdir ' + os.path.join(path, '0_results', new_dir) os.system(command) parent_dir = os.path.join(path, '0_results', new_dir) ########################################################## debug.active += ["SLC"] ds_merged = get_merged_ds(path, subjects, conf_file, type, dim=dim, **kwargs) cv = CrossValidation(measure, partitioner, splitter=Splitter(attr='partitions', attr_values=(3, 2)), errorfx=None) maps = [] kwa = dict(voxel_indices=Sphere(radius), event_offsetidx=Sphere(duration)) queryengine = IndexQueryEngine(**kwa) for s, ds in zip(subjects, ds_merged): print ds.samples.shape voxel_num = ds.samples.shape[1] / duration ids = np.arange(voxel_num) sl = Searchlight(cv, queryengine=queryengine, roi_ids=ids) map = sl(ds) maps.append(map) name = s command = 'mkdir ' + os.path.join(parent_dir, name) os.system(command) results_dir = os.path.join(parent_dir, name) fname = name + '_radius_' + str(radius) + '_searchlight_map.nii.gz' map_ = ds.a.mapper[:2].reverse(map) map_ = np.rollaxis(map_.samples, 0, 4) img = ni.Nifti1Image(map_, ds.a.imghdr.get_base_affine()) ni.save(img, os.path.join(results_dir, fname)) return maps
def _group_transfer_learning(path, subjects, analysis, conf_file, source='task', analysis_type='single', **kwargs): if source == 'task': target = 'rest' else: if source == 'rest': target = 'task' if source == 'saccade': target = 'face' else: if source == 'face': target = 'saccade' ############################################## ############################################## ## conf_src['label_included'] = 'all' ## ## conf_src['label_dropped'] = 'none' ## ## conf_src['mean_samples'] = 'False' ## ############################################## ############################################## if analysis_type == 'group': if path.__class__ == conf_file.__class__ == list: ds_src, s, conf_src = sources_merged_ds(path, subjects, conf_file, source, **kwargs) conf_src['permutations'] = 0 else: print 'In group analysis path, subjects and conf_file must be lists: \ Check configuration file and/or parameters!!' return 0 else: conf_src = read_configuration(path, conf_file, source) for arg in conf_src: if arg == 'map_list': map_list = conf_src[arg].split(',') r_group = spatial(ds_src, **conf_src) total_results = dict() total_results['group'] = r_group clf = r_group['classifier'] for subj_, conf_, path_ in zip(subjects, conf_file, path): for subj in subj_: print '-----------' r = dict() if len(subj_) > 1: conf_tar = read_configuration(path_, conf_, target) for arg in kwargs: conf_tar[arg] = kwargs[arg] data_path = conf_tar['data_path'] try: ds_tar = load_dataset(data_path, subj, target, **conf_tar) except Exception, err: print err continue ds_tar = detrend_dataset(ds_tar, target, **conf_tar) if conf_src['label_included'] == 'all' and \ conf_src['label_dropped'] != 'fixation': print 'Balancing dataset...' ds_src = balance_dataset_timewise(ds_src, 'fixation') predictions = clf.predict(ds_tar) pred = np.array(predictions) targets = ds_tar.targets for arg in r_group.keys(): r[arg] = copy.copy(r_group[arg]) r['targets'] = targets r['predictions'] = predictions r['fclf'] = clf c_m = ConfusionMatrix(predictions=pred, targets=targets) c_m.compute() r['confusion_target'] = c_m print c_m tr_pred = similarity_measure_mahalanobis(ds_tar, ds_src, r) r['mahalanobis_similarity'] = tr_pred #print tr_pred c_mat_mahala = ConfusionMatrix(predictions=tr_pred.T[1], targets=tr_pred.T[0]) c_mat_mahala.compute() r['confusion_mahala'] = c_mat_mahala d_prime, beta, c, c_new = signal_detection_measures(pred, targets, map_list) r['d_prime'] = d_prime print d_prime r['beta'] = beta r['c'] = c r['confusion_total'] = c_new ''' d_prime_maha, c_new_maha = d_prime_statistics(tr_pred.T[1], tr_pred.T[0], map_list) r['d_prime_maha'] = d_prime_maha r['confusion_tot_maha'] = c_new_maha ''' total_results[subj] = r
from mvpa2.suite import mean_group_sample, map2nifti from mvpa2.mappers.fx import BinaryFxNode from mvpa2.misc.errorfx import mean_mismatch_error import nibabel as ni import numpy as np from mvpa2.clfs.base import Classifier from mvpa2.generators.resampling import Balancer import mvpa_itab.results as rs from mvpa2.misc.neighborhood import Sphere, IndexQueryEngine from mvpa2.measures.searchlight import Searchlight #path = '/media/robbis/DATA/fmri/memory/' conf = read_configuration(path, 'remote_memory.conf', 'BETA_MVPA') conf['analysis_type'] = 'searchlight' conf['analysis_task'] = 'memory_regression_sample_wise' conf['mask_area'] = 'total' task_ = 'BETA_MVPA' subj = '110929anngio' partitioners = [NGroupPartitioner(k) for k in np.arange(2, 5)] result_dict = dict() summarizers = [rs.SearchlightSummarizer()] savers = [rs.SearchlightSaver()] collection = rs.ResultsCollection(conf, path, summarizers)