def _build_data_set(self, to_file=False):
list_of_dfs = []
df_patient_data = pd.read_excel(os.path.join(patient_data_path, 'PREDICT-AF_Measurements.xlsx'), # AduHeart_PatientData_Relevant
index_col='patient_ID', header=0)
for curv_file in self.files:
print('case: {}'.format(curv_file))
ven = Trace(self.source_path, case_name=curv_file, view=self.view, interpolate=self.interpolate_traces)
list_of_dfs.append(ven.get_biomarkers())
self.df_all_cases = pd.concat(list_of_dfs)
self.df_all_cases.index.name = 'ID'
self.df_all_cases['patient_ID'] = self.df_all_cases.index.map({k: k.split('_')[0] for k
in self.df_all_cases.index})
print(self.df_all_cases)
print(df_patient_data)
self.df_all_cases = self.df_all_cases.join(df_patient_data.SB, how='inner', on='patient_ID')
self.df_all_cases = self.df_all_cases.set_index(['patient_ID', self.df_all_cases.index])
self.df_all_cases['min_index'] = np.abs(self.df_all_cases.min_delta * self.df_all_cases['min'])
self.df_all_cases['min_index_ED'] = np.abs(self.df_all_cases.min_delta_ED * self.df_all_cases.min_ED)
self.df_all_cases['curv_len_inter'] = np.abs(self.df_all_cases.min_ED * self.df_all_cases.trace_length_ED)
print(self.df_all_cases)
if to_file:
data_set_output_dir = check_directory(os.path.join(self.output_path, 'output_EDA'))
self.df_all_cases.to_csv(os.path.join(data_set_output_dir, self.indices_file))
print('master table saved')
def save_curvatures(self):
_output_path = check_directory(os.path.join(self.output_path, 'curvatures'))
for case in self.files:
ven = Trace(self.source_path, case_name=case, view=self.view, interpolate=self.interpolate_traces)
print(ven.id)
pd.DataFrame(ven.ventricle_curvature).to_csv(os.path.join(_output_path, ven.id+'.csv'))
def _set_paths_and_files(self, view=None, output_path=''):
# self.view = view
if view is not None: # Read specified view data
self.files = glob.glob(os.path.join(self.source_path, self.view, '*.CSV'))
else: # Read all CSVs in the source directory
self.files = glob.glob(os.path.join(self.source_path, '*.CSV'))
self.files.sort()
if not output_path == '':
self.output_path = check_directory(output_path)
def plot_curvatures(self, coloring_scheme='curvature', plot_mean=False):
_source_path = os.path.join(self.source_path, self.view)
if plot_mean:
_output_path = check_directory(os.path.join(self.output_path, 'output_curvature', 'mean'))
else:
_output_path = check_directory(os.path.join(self.output_path, 'output_curvature'))
for case in self.files:
ven = Trace(self.source_path, case_name=case, view=self.view, interpolate=self.interpolate_traces)
print(ven.id)
print('Points: {}'.format(ven.number_of_points))
plot_tool = PlottingCurvature(source=_source_path,
output_path=_output_path,
ventricle=ven)
if plot_mean:
plot_tool.plot_mean_curvature()
else:
plot_tool.plot_all_frames(coloring_scheme=coloring_scheme)
plot_tool.plot_heatmap()
def _plot_master(self):
if self.df_master is None:
self._try_get_data(master_table=True)
_master_output_path = check_directory(os.path.join(self.output_path, 'output_master'))
master_plot_tool = PlottingDistributions(self.df_master, '', _master_output_path)
for col in self.biomarkers:
print(col)
if self.table_name != 'master_table.csv':
master_plot_tool.plot_with_labels('4C_' + col, '3C_' + col)
else:
master_plot_tool.plot_2_distributions('4C_' + col, '3C_' + col, kind='kde')
def save_extemes(self, n=30):
self._try_get_data(data=True)
list_of_extremes = []
for col in self.df_all_cases.columns:
list_of_extremes.append(self.df_all_cases[col].sort_values(ascending=False).index.values[:n])
list_of_extremes.append(self.df_all_cases[col].sort_values(ascending=False).values[:n])
index_lists = [2 * [i] for i in self.df_all_cases.columns]
index = [item for sublist in index_lists for item in sublist]
df_extremes = pd.DataFrame(list_of_extremes, index=index)
_output_path = check_directory(os.path.join(self.output_path, self.view, 'output_EDA'))
df_extremes.to_csv(os.path.join(_output_path, 'extremes.csv'))
def __init__(self, source_path='data', view='4C', output_path='data', indices_file='indices_all_cases.csv',
interpolate_traces=None):
self.view = view
self.source_path = source_path
self.output_path = check_directory(output_path)
self.indices_file = indices_file
self.interpolate_traces = interpolate_traces
self.files = glob.glob(os.path.join(self.source_path, '*.CSV'))
self.files.sort()
self.df_all_cases = None
self.df_master = None
self.curv = None
self.biomarkers = None
self.table_name = None
def _plot_data(self):
if self.df_all_cases is None:
self._try_get_data(data=True)
_view_output_path = check_directory(os.path.join(self.output_path, self.view, 'output_EDA'))
plot_tool = PlottingDistributions(self.df_all_cases, '', _view_output_path)
for col in self.biomarkers:
plot_tool.set_series(col)
plot_tool.plot_distribution()
col_combs = combinations(self.biomarkers, 2)
for comb in col_combs:
if self.table_name != 'master_table.csv':
plot_tool.plot_with_labels(comb[0], comb[1])
else:
plot_tool.plot_2_distributions(comb[0], comb[1], kind='kde')
def plots_wt_and_curvature_vs_markers(self, save_figures=False):
plot_dir = check_directory(os.path.join(self.output_path, 'plots'))
x_labels = ['min_ED', 'avg_min_basal_curv', r'PLAX basal/mid', r'4C basal/mid', 'avg_basal_ED']
for x_label in x_labels:
for y_label in self.FACTORS_BASIC:
if x_label in ['PLAX basal_mid', '4C basal_mid']:
plt.axvline(1.4, linestyle='--', c='k')
self.df_comparison.plot(x=x_label, y=y_label, c='SB', kind='scatter', legend=True, colorbar=True,
cmap='winter', title='Relation of {} to {}'.format(y_label, x_label))
else:
self.df_comparison.plot(x=x_label, y=y_label, c=x_label, kind='scatter', legend=True, colorbar=True,
cmap='autumn', title='Relation of {} to {}'.format(y_label, x_label))
if save_figures:
plt.savefig(os.path.join(plot_dir, r'{} vs {} HTNs.png'.format(y_label, x_label.replace('/', '_'))))
else:
plt.show()
plt.close()
def linear_regression_basic_factors(self, to_file=False, show_plots=False):
from sklearn.linear_model import LinearRegression
from sklearn.metrics import mean_squared_error, r2_score
markers = ['Average septal curvature [cm-1]',
r'Wall thickness ratio in 4CH view',
r'Wall thickness ratio in PLAX view']
list_results = []
for marker in markers:
for factor in self.FACTORS_BASIC:
x = self.df_comparison[marker].values.reshape(-1, 1)
y = self.df_comparison[factor].values.reshape(-1, 1)
lr = LinearRegression()
lr.fit(x, y)
y_pred = lr.predict(x)
rho_sp, p_sp = spearmanr(x, y)
r_pe = pearsonr(x, y)
dict_results = {'marker': marker, 'factor': factor, 'coefficients': lr.coef_, 'R2': r2_score(y, y_pred),
'mse': mean_squared_error(y, y_pred), 'spearmanr': rho_sp, 'spearmanp': p_sp,
'pearsonr': r_pe}
list_results.append(dict_results)
if show_plots:
plots = PlottingDistributions(self.df_comparison, 'min',
check_directory(os.path.join(self.output_path, 'plots')))
plots.plot_with_labels(series1=marker, series2=factor, w_labels=False)
df_results = pd.DataFrame(list_results)
if to_file:
df_results.to_csv(os.path.join(self.output_path, 'Linear_regression_results.csv'))
def plot_curv_vs_wt(self, save_figures=False, w_reg=False):
plot_dir = check_directory(os.path.join(self.output_path, 'plots'))
x_labels = [r'PLAX basal/mid', r'4C basal/mid', 'IVSd (basal) PLAX', 'IVSd (mid) PLAX', 'IVSd (basal) 4C',
'IVSd (mid) 4C']
y_labels = ['min_ED', 'avg_basal_ED', 'avg_min_basal_curv']
# for x_label in x_labels:
# for y_label in y_labels:
# self.df_comparison.plot(x=x_label, y=y_label, c='SB', kind='scatter', legend=True, colorbar=True,
# cmap='winter', title='Relation of {} to {}'.format(y_label, x_label))
# means_x = self.df_comparison.groupby('SB')[x_label].mean()
# means_y = self.df_comparison.groupby('SB')[y_label].mean()
# plt.plot(means_x, means_y, 'kd')
#
# if x_label in [r'PLAX basal/mid', r'4C basal/mid']:
# plt.axvline(1.4, linestyle='--', c='k')
# if save_figures:
# plt.savefig(os.path.join(plot_dir, r'Meas {} vs {} HTNs.png'.format(y_label,
# x_label.replace('/', '_'))))
# else:
# plt.show()
# plt.close()
# print('Curvature below -1: {}'.format(self.df_comparison.curv_threshold.sum()))
# print('4C above 1.4: {}'.format((self.df_comparison['4C basal/mid'] > 1.4).sum()))
# print('PLAX above 1.4: {}'.format((self.df_comparison['PLAX basal/mid'] > 1.4).sum()))
# print('SB cases: {}'.format((self.df_comparison.SB > 1).sum()))
# 'Average septal curvature [cm-1]',
# # r'Wall thickness ratio in 4CH view',
# # r'Wall thickness ratio in PLAX view'
from matplotlib import cm
from matplotlib.colors import ListedColormap
top = cm.get_cmap('Oranges_r', 128)
bottom = cm.get_cmap('Blues', 128)
newcolors = np.vstack((top(np.linspace(0, 1, 512)),
bottom(np.linspace(0, 1, 512))))
newcmp = ListedColormap(newcolors, name='OrangeBlue')
self.df_comparison.plot(x=r'Wall thickness ratio in PLAX view', y=r'Wall thickness ratio in 4CH view',
c='Average septal curvature [cm-1]', kind='scatter', legend=False, s=200,
colorbar=True, cmap=newcmp, title='',
figsize=(9, 7.2))
# plt.title('Curvature values w.r.t. both WTR metrics', fontsize=26)
plt.xlabel(r'Wall thickness ratio in PLAX view', fontsize=23)
plt.ylabel(r'Wall thickness ratio in 4CH view', fontsize=23)
plt.xticks(fontsize=16)
plt.yticks(fontsize=16)
plt.axvline(1.4, ymax=0.47, linestyle='--', c='k')
plt.axhline(1.4, xmax=0.462, linestyle='--', c='k')
plt.xlim((0.7, 2.2))
plt.ylim((0.7, 2.2))
plt.tight_layout()
f = plt.gcf()
f.get_axes()[1].set_ylabel('Average septal curvature $[dm^{-1}]$', fontsize=23)
f.get_axes()[1].tick_params(labelsize=16)
if save_figures:
plt.savefig(os.path.join(plot_dir, r'Ratios_curvature.svg'))
else:
plt.show()
view = ''
segment = ''
df_range = var.calculate_sem_single_index(o2=o2)
var.bland_altman_plot_single_index(o2=o2)
# ranges = pd.concat((ranges, df_range), axis=0)
# for view in ['PLAX', '4C']:
# var.bland_altman_plot_multi_index(o2=o2, view=view, segment='ratio')
# df_range = var.calculate_sem_multi_index(o2=o2, view=view, segment='ratio')
# ranges = pd.concat((ranges, df_range), axis=0)
# ranges.to_csv(os.path.join(var.output_path, 'ranges.csv'))
# print(ranges)
# STRAIN ANALYSIS
#
patient_data_path = os.path.join('C:\Data\ProjectCurvature\Analysis\Output_HTN\Statistics')
curvature_results = os.path.join('C:/', 'Data', 'ProjectCurvature', 'Analysis', 'Output')
output = check_directory(os.path.join('C:\Data\ProjectCurvature\Analysis\Output_HTN\Statistics\plots', 'EDA'))
# measurements = 'AduHeart_Measurements.xlsx'
# twodstrain = 'AduHeart_Strain_MW.xlsx'
# curvature = 'master_table_full.csv'
# patient_info = 'AduHeart_PatientData_Full.xlsx'
merged_data = 'Measurements_and_2DstrainPlotting.csv'
#
# anal = StrainAnalysis(patient_data_path, curvature_results, output, merged_data_filename=merged_data)
# anal.plots_wt_and_curvature_vs_markers(True)
# anal.plot_curv_vs_wt(True)
# anal.get_statistics()
# anal.linear_regression_basic_factors(False, show_plots=True)
# STATANALYSIS