def plot_sample_mean_and_sd_maps(selected_patients):
dataDirectory = r"../Data/OnlyProstateResults/AllFields"
patients_who_recur, patients_who_dont_recur = separate_by_recurrence(selected_patients)
# (meanMap1, varMap, stdMap) = load_local_field_recurrence(selected_patients, dataDirectory)
(meanMap1, varMap1, stdMap1) = make_average_field(patients_who_recur, dataDirectory)
meanMap1.to_csv("../outputResults/recurrence_mean_map.csv", header=None, index=False)
stdMap1.to_csv("../outputResults/recurrence_std_map.csv", header=None, index=False)
plot_heat_map(meanMap1, -1, 1, 'mean map - patients_who_recur')
plot_heat_map(varMap1, 0, 1, 'variance map - patients_who_recur')
plot_heat_map(stdMap1, 0, 1, 'standard deviation map - patients_who_recur')
(meanMap2, varMap2, stdMap2) = make_average_field(patients_who_dont_recur, dataDirectory)
plot_heat_map(meanMap2, -1, 1, 'mean map - patients_who_dont_recur')
plot_heat_map(varMap2, 0, 1, 'variance map - patients_who_dont_recur')
plot_heat_map(stdMap2, 0, 1, 'standard deviation map - patients_who_dont_recur')
meanMap2.to_csv("../outputResults/no_recurrence_mean_map.csv", header=None, index=False)
stdMap2.to_csv("../outputResults/no_recurrence_std_map.csv", header=None, index=False)
plot_heat_map(meanMap1 - meanMap2, -0.3, 0.3, 'Difference in mean map')
# Var[X-Y] = Var[X]+Var[Y]
# Standard deviation is the square root of the variance
plot_heat_map(np.sqrt(varMap1 + varMap2), 0, 1.5, 'Difference in std map')
(meanMap1 - meanMap2).to_csv("../outputResults/mean_difference_map.csv", header=None, index=False)
np.sqrt(varMap1 + varMap2).to_csv("../outputResults/std_difference_map.csv", header=None, index=False)
def print_volume_difference_details(patient_database):
"""
:param patient_database: is a dataframe that contains the patients global variables
:return: prints volume difference statistics for collectiver patient, then patients with recurrence and patients /
without
"""
print(patient_database["volumeContourDifference"].describe())
patients_who_recur, patients_who_dont_recur = separate_by_recurrence(patient_database)
print(patients_who_recur["volumeContourDifference"].describe())
print(patients_who_dont_recur["volumeContourDifference"].describe())
def plot_neat_scatter(data, save_name):
fit = np.polyfit(data["volumeContour"], data["volumeContourAuto"], 1)
fit_fn = np.poly1d(fit)
# Fitting the graph
# fig = plt.figure()
x = np.linspace(10, 190, 1000) # Plot straight line
y = x
# ax = plt.gca()
# ax.set_facecolor('white')
# plt.grid()
plt.style.use('seaborn-ticks')
plt.rcParams['font.family'] = 'times'
plt.rcParams['axes.labelweight'] = 'bold'
plt.rcParams['axes.labelsize'] = 12
plt.rcParams['legend.fontsize'] = 12
rec, n_rec = separate_by_recurrence(data)
plt.scatter(rec["volumeContour"],
rec["volumeContourAuto"],
c='#ff0000',
alpha=0.5,
label='Recurrence')
plt.scatter(n_rec["volumeContour"],
n_rec["volumeContourAuto"],
c='#00ff00',
alpha=0.5,
label='No Recurrence')
# plt.scatter(censored["volumeContour"], censored["volumeContourAuto"], c='#0000ff', alpha=0.5, label='Censored')
plt.plot(x,
y,
linestyle='dashed',
color='k',
label='Line of Equal Volumes') # y = x line
# plt.plot(x,x,'yo', AllPatients["volumeContour"], fit_fn(AllPatients["volumeContour"]), '--k') # linear fit
plt.figure(1)
plt.xlim(0, 200)
plt.ylim(0, 200)
plt.xlabel('Manual-contour volume [mm$^3$]')
plt.ylabel('Auto-contour volume [mm$^3$]')
plt.legend(loc="upper left", frameon=True, framealpha=1)
plt.grid(True)
# ax.set_xticks()
# ax.set_yticks()
# # ax.grid()
# # ax.axis('equal')
plt.savefig(save_name, bbox_inches='tight')
plt.show()
def mann_whitney_test_statistic(selected_patients):
"""
Conduct a mann whitney test SUM rank test
:param selected_patients: dataframe of all patients
:return: maps of the statistic and the p-values
"""
# Tag patients with recurrence:1 and non-recurrence:0
patients_who_recur, patients_who_dont_recur = separate_by_recurrence(
selected_patients)
rec_fieldMaps, _ = stack_local_fields(patients_who_recur, 1)
nonrec_fieldMaps, _ = stack_local_fields(patients_who_dont_recur, 0)
stat_map, p_map = mann_whitney_u_test(rec_fieldMaps, nonrec_fieldMaps)
return stat_map, p_map
def global_statistical_analysis(selected_patients):
'''
A function that will carry out the statistical analysis of the global data: The dice coefficient and volume
difference for patients with and without cancer recurrence.
:param selected_patients: A dataframe of all patients
'''
# Split the patients by recurrence
patients_who_recur, patients_who_dont_recur = separate_by_recurrence(
selected_patients)
print('#rec: %.f, #n_rec: %.f' %
(patients_who_recur.shape[0], patients_who_dont_recur.shape[0]))
# Test the relationship between volume difference and recurrence
global_p, lower_p, upper_p = non_parametric_permutation_test(
patients_who_recur["volumeContourDifference"],
patients_who_dont_recur["volumeContourDifference"])
print(
'Vdiff: p_value(rec=/=n_rec): %.6f p_value(rec<n_rec): %.6f p_value(rec>n_rec): %.6f'
% (global_p, lower_p, upper_p))
# Test the relationship between dice coefficient and recurrence
global_p, lower_p, upper_p = non_parametric_permutation_test(
patients_who_recur["DSC"], patients_who_dont_recur["DSC"])
print(
'DSC: p_value(rec=/=n_rec): %.6f p_value(rec<n_rec): %.6f p_value(rec>n_rec): %.6f'
% (global_p, lower_p, upper_p))
# Test the relationship between ratio of volumes and recurrence: V_man/V_auto
global_p, lower_p, upper_p = non_parametric_permutation_test(
patients_who_recur["volumeRatio"],
patients_who_dont_recur["volumeRatio"])
print(
'VRatio: p_value(rec=/=n_rec): %.6f p_value(rec<n_rec): %.6f p_value(rec>n_rec): %.6f'
% (global_p, lower_p, upper_p))
# Test the relationship between mean dose diff and recurrence: V_man/V_auto
global_p, lower_p, upper_p = non_parametric_permutation_test(
patients_who_recur["meanDoseDiff"],
patients_who_dont_recur["meanDoseDiff"])
print(
'VRatio: p_value(rec=/=n_rec): %.6f p_value(rec<n_rec): %.6f p_value(rec>n_rec): %.6f'
% (global_p, lower_p, upper_p))
def pymining_t_test(selected_patients):
# Tag patients with recurrence:1 and non-recurrence:0
patients_who_recur, patients_who_dont_recur = separate_by_recurrence(
selected_patients)
(rec_fieldMaps,
rec_label_array) = stack_local_fields(patients_who_recur, 1)
(non_rec_field_maps,
non_rec_label_array) = stack_local_fields(patients_who_dont_recur, 0)
# Concatenate the two
totalPatients = np.concatenate((rec_fieldMaps, non_rec_field_maps),
axis=-1)
labels = np.concatenate((rec_label_array, non_rec_label_array))
# Now use pymining to get DSC cuts global p value. It should be similar to that from scipy
global_neg_pvalue, global_pos_pvalue, neg_tthresh, pos_tthresh = pm.permutationTest(
totalPatients, labels, 1000)
t_value_map = pm.imagesTTest(totalPatients,
labels) # no longer.[0] element
return global_neg_pvalue, global_pos_pvalue, neg_tthresh, pos_tthresh, t_value_map