def plot_heat_temperature(self):
"""Plot the measurement temperature as a heat map in (H, Hr) or (Hc, Hb) space.
"""
self.send_latest_data.emit()
plotting.heat_map(ax=self.p_map.axes,
forc=self.data_queue.get(),
data_str='temperature',
mask=self.f_2d_mask.currentText(),
coordinates=self.coordinates(),
cmap=self.f_2d_cmap.text())
self.tabWidget.setCurrentIndex(1)
return
def plot_heat_rho_uncertainty(self):
"""Plot the FORC distribution uncertainty as a heat map in (H, Hr) or (Hc, Hb) space.
"""
self.send_latest_data.emit()
plotting.heat_map(ax=self.p_map.axes,
forc=self.data_queue.get(),
data_str='rho_uncertainty',
mask=self.f_2d_mask.currentText(),
coordinates=self.coordinates(),
cmap=self.f_2d_cmap.text())
self.tabWidget.setCurrentIndex(1)
return
def post_processing(clf, basename):
"""Given a PairwiseClassifier object and a basename, perform post processing"""
# Average classification accuracy brain map
mc.make_mask_map(clf, basename + "avg_class.nii.gz",
mc.get_mask_averages(clf, ))
print "Made average accuracy map"
print "Making consistency brain maps.."
mc.make_mask_map(clf, basename + "imps_shannons_0.nii.gz",
mc.importance_stats(clf, method='shannons'))
mc.make_mask_map(clf, basename + "imps_var_0.nii.gz",
mc.importance_stats(clf, method='var'))
# mc.make_mask_map(basename + "imps_cor_0.nii.gz", mc.importance_stats(method='cor'))
mc.make_mask_map(clf, basename + "acc_shannons_0.nii.gz",
mc.accuracy_stats(clf, method='shannons'))
mc.make_mask_map(clf, basename + "clf_var_0.nii.gz",
mc.accuracy_stats(clf, method='var'))
print "Making sparsity brain maps.."
mc.make_mask_map(clf, basename + "imps_shannons_1.nii.gz",
mc.importance_stats(clf, method='shannons', axis=1))
mc.make_mask_map(clf, basename + "imps_var_1.nii.gz",
mc.importance_stats(clf, method='var', axis=1))
# mc.make_mask_map(clf, basename + "imps_cor_1.nii.gz", mc.importance_stats(clf, method='cor', axis=1))
mc.make_mask_map(clf, basename + "acc_shannons_1.nii.gz",
mc.accuracy_stats(clf, method='shannons'))
mc.make_mask_map(clf, basename + "clf_var_1.nii.gz",
mc.accuracy_stats(clf, method='var'))
# print "Making consistency heat maps..."
heat_map(mc.importance_stats(clf, method='shannons', axis=0,
average=False).T,
range(1, clf.mask_num),
clf.feature_names,
file_name=basename + "shannons_hm_0.png")
heat_map(mc.importance_stats(clf, method='var', axis=0, average=False).T,
range(1, clf.mask_num),
clf.feature_names,
file_name=basename + "var_hm_0.png")
print "Making sparsity heat maps..."
heat_map(mc.importance_stats(clf, method='shannons', axis=1,
average=False).T,
range(1, clf.mask_num),
range(0, clf.mask_num),
file_name=basename + "shannons_hm_1.png")
heat_map(mc.importance_stats(clf, method='var', axis=1, average=False).T,
range(1, clf.mask_num),
range(0, clf.mask_num),
file_name=basename + "var_hm_1.png")
print "Making feature importance heatmaps..."
mc.region_heatmap(clf, basename)
mc.region_heatmap(clf, basename, zscore_regions=True)
mc.region_heatmap(clf, basename, zscore_features=True)
mc.region_heatmap(clf, basename, zscore_regions=True, zscore_features=True)
def post_processing(clf, basename):
"""Given a PairwiseClassifier object and a basename, perform post processing"""
# Average classification accuracy brain map
mc.make_mask_map(clf, basename + "avg_class.nii.gz", mc.get_mask_averages(clf, ))
print "Made average accuracy map"
print "Making consistency brain maps.."
mc.make_mask_map(clf, basename + "imps_shannons_0.nii.gz", mc.importance_stats(clf, method='shannons'))
mc.make_mask_map(clf, basename + "imps_var_0.nii.gz", mc.importance_stats(clf, method='var'))
# mc.make_mask_map(basename + "imps_cor_0.nii.gz", mc.importance_stats(method='cor'))
mc.make_mask_map(clf, basename + "acc_shannons_0.nii.gz", mc.accuracy_stats(clf, method='shannons'))
mc.make_mask_map(clf, basename + "clf_var_0.nii.gz", mc.accuracy_stats(clf, method='var'))
print "Making sparsity brain maps.."
mc.make_mask_map(clf, basename + "imps_shannons_1.nii.gz", mc.importance_stats(clf, method='shannons', axis=1))
mc.make_mask_map(clf, basename + "imps_var_1.nii.gz", mc.importance_stats(clf, method='var', axis=1))
# mc.make_mask_map(clf, basename + "imps_cor_1.nii.gz", mc.importance_stats(clf, method='cor', axis=1))
mc.make_mask_map(clf, basename + "acc_shannons_1.nii.gz", mc.accuracy_stats(clf, method='shannons'))
mc.make_mask_map(clf, basename + "clf_var_1.nii.gz", mc.accuracy_stats(clf, method='var'))
# print "Making consistency heat maps..."
heat_map(mc.importance_stats(clf, method='shannons', axis=0, average=False).T,
range(1, clf.mask_num), clf.feature_names, file_name=basename + "shannons_hm_0.png")
heat_map(mc.importance_stats(clf, method='var', axis=0, average=False).T,
range(1, clf.mask_num), clf.feature_names, file_name=basename +
"var_hm_0.png")
print "Making sparsity heat maps..."
heat_map(mc.importance_stats(clf, method='shannons', axis=1, average=False).T,
range(1, clf.mask_num), range(0, clf.mask_num), file_name=basename + "shannons_hm_1.png")
heat_map(mc.importance_stats(clf, method='var', axis=1, average=False).T,
range(1, clf.mask_num), range(0, clf.mask_num), file_name=basename+"var_hm_1.png")
print "Making feature importance heatmaps..."
mc.region_heatmap(clf, basename)
mc.region_heatmap(clf, basename, zscore_regions=True)
mc.region_heatmap(clf, basename, zscore_features=True)
mc.region_heatmap(clf, basename, zscore_regions=True, zscore_features=True)
def region_heatmap(clf, basename=None, zscore_regions=False, zscore_features=False, thresh=None, subset=None, compare=False, each_region=True):
"""" Makes a heatmap of the importances of the classification. Makes an overall average heatmap
as well as a heatmap for each individual region. Optionally, you can specify the heatmap to be
z-scored. You can also specify a threshold.
Args:
basename: string, base directory and file name
zscore_regions: boolean, should heatmap be z-scored based within regions
zscore_regions: boolean, should heatmap be z-scored based within features
thresh: value to threshold heatmap. Only values above this value are kept
subset: what regions should be plotted? default is all
each_region: make a heat map for every single subregion
Outputs:
Outputs a .png file for the overall heatmap and for each region. If z-scored on thresholded,
will denote in file name using z0 (regions), z1 (features), and/or t followed by threshold.
"""
from plotting import heat_map
if subset is None:
subset = range(0, clf.mask_num)
if compare is True:
overall_fi = clf.feature_importances[subset][:, subset]
else:
overall_fi = clf.feature_importances[:, subset]
if np.array(subset).max() > clf.mask_num:
print "Warning: you entered an incorrect mask index!"
fi = overall_fi.mean(axis=0).T
z0 = ""
z1 = ""
t = ""
if zscore_regions:
fi = np.apply_along_axis(stats.zscore, 0, fi)
z0 = "z0_"
if zscore_features:
fi = np.apply_along_axis(stats.zscore, 1, fi)
z1 = "z1_"
if thresh is not None:
fi = np.ma.masked_array(fi)
fi.mask = fi < thresh
t = "zt" + str(thresh) + "_"
if basename is None:
file_name = None
else:
file_name = basename + "imps_hm_" + z0 + z1 + t + "overall.png"
heat_map(fi, np.array(subset) + 1, clf.feature_names, file_name)
if each_region:
for i in subset:
fi = overall_fi[subset.index(i)].T
if zscore_regions:
fi = np.ma.masked_invalid(stats.zscore(fi, axis=0))
if zscore_features:
fi = stats.zscore(fi, axis=1)
if thresh is not None:
fi.mask = fi < thresh
if basename is None:
file_name = None
else:
file_name = basename + "imps_hm_" + \
z0 + z1 + t + str(i) + ".png"
heat_map(fi, np.array(subset) + 1, clf.feature_names, file_name)
