Commit
·
e00c8e8
1
Parent(s):
7d43ed6
Added input arguments to change the CSV file, separator and output directory
Browse files- EvaluationScripts/statistics.py +31 -27
EvaluationScripts/statistics.py
CHANGED
@@ -1,21 +1,20 @@
|
|
1 |
import numpy as np
|
|
|
|
|
2 |
from statsmodels.stats.multicomp import pairwise_tukeyhsd
|
3 |
from statsmodels.stats.multitest import fdrcorrection
|
4 |
from scipy import stats
|
5 |
import pandas as pd
|
6 |
|
|
|
|
|
7 |
|
8 |
# increase length of string in pandas
|
9 |
pd.options.display.max_colwidth = 100
|
10 |
|
11 |
|
12 |
-
def post_hoc_ixi():
|
13 |
-
|
14 |
-
|
15 |
-
df = pd.read_csv(file_path, sep=";")
|
16 |
-
df = df.iloc[:, 1:]
|
17 |
-
df = df[df["Experiment"] == "IXI"]
|
18 |
-
df["Model"] = [x.replace("_", "-") for x in df["Model"]]
|
19 |
|
20 |
TRE_values = df["TRE"]
|
21 |
m_comp = pairwise_tukeyhsd(df["TRE"], df["Model"], alpha=0.05)
|
@@ -53,20 +52,14 @@ def post_hoc_ixi():
|
|
53 |
out_pd = out_pd.replace(-1.0, "-")
|
54 |
out_pd = out_pd.replace(-0.0, '\cellcolor{green!25}$<$0.001')
|
55 |
|
56 |
-
with open("
|
57 |
pfile.write("{}".format(out_pd.to_latex(escape=False, column_format="r" + "c"*all_pvalues.shape[1], bold_rows=True)))
|
58 |
|
59 |
print(out_pd)
|
60 |
|
61 |
|
62 |
-
def study_transfer_learning_benefit():
|
63 |
-
|
64 |
-
|
65 |
-
df = pd.read_csv(file_path, sep=";")
|
66 |
-
df = df.iloc[:, 1:]
|
67 |
-
df["Model"] = [x.replace("_", "-") for x in df["Model"]]
|
68 |
-
|
69 |
-
df_tl = df[df["Experiment"] == "COMET_TL_Ft2Stp"]
|
70 |
df_orig = df[df["Experiment"] == "COMET"]
|
71 |
|
72 |
pvs = []
|
@@ -91,14 +84,8 @@ def study_transfer_learning_benefit():
|
|
91 |
print("UW-NSD:", corrected_pvs[2])
|
92 |
|
93 |
|
94 |
-
def post_hoc_comet():
|
95 |
-
|
96 |
-
|
97 |
-
df = pd.read_csv(file_path, sep=";")
|
98 |
-
df = df.iloc[:, 1:]
|
99 |
-
df["Model"] = [x.replace("_", "-") for x in df["Model"]]
|
100 |
-
|
101 |
-
df_tl = df[df["Experiment"] == "COMET_TL_Ft2Stp"]
|
102 |
|
103 |
filter_ = np.array([x in ["BL-N", "SG-NSD", "UW-NSD"] for x in df_tl["Model"]])
|
104 |
|
@@ -129,11 +116,28 @@ def post_hoc_comet():
|
|
129 |
|
130 |
|
131 |
if __name__ == "__main__":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
132 |
print("\nComparing all contrasts in TRE of all models in the IXI dataset:")
|
133 |
-
post_hoc_ixi()
|
134 |
|
135 |
print("\nTransfer learning benefit (COMET):")
|
136 |
-
study_transfer_learning_benefit()
|
137 |
|
138 |
print("\nAssessing whether there is a benefit to segmentation-guiding and uncertainty weighting (COMET):")
|
139 |
-
post_hoc_comet()
|
|
|
|
|
|
|
|
1 |
import numpy as np
|
2 |
+
import statsmodels
|
3 |
+
import scipy
|
4 |
from statsmodels.stats.multicomp import pairwise_tukeyhsd
|
5 |
from statsmodels.stats.multitest import fdrcorrection
|
6 |
from scipy import stats
|
7 |
import pandas as pd
|
8 |
|
9 |
+
import argparse
|
10 |
+
import os
|
11 |
|
12 |
# increase length of string in pandas
|
13 |
pd.options.display.max_colwidth = 100
|
14 |
|
15 |
|
16 |
+
def post_hoc_ixi(df_in: pd.DataFrame, outdir: str = './'):
|
17 |
+
df = df_in[df_in["Experiment"] == "IXI"]
|
|
|
|
|
|
|
|
|
|
|
18 |
|
19 |
TRE_values = df["TRE"]
|
20 |
m_comp = pairwise_tukeyhsd(df["TRE"], df["Model"], alpha=0.05)
|
|
|
52 |
out_pd = out_pd.replace(-1.0, "-")
|
53 |
out_pd = out_pd.replace(-0.0, '\cellcolor{green!25}$<$0.001')
|
54 |
|
55 |
+
with open(os.path.join(outdir, "tukey_pvalues_result_IXI.txt"), "w") as pfile:
|
56 |
pfile.write("{}".format(out_pd.to_latex(escape=False, column_format="r" + "c"*all_pvalues.shape[1], bold_rows=True)))
|
57 |
|
58 |
print(out_pd)
|
59 |
|
60 |
|
61 |
+
def study_transfer_learning_benefit(df_in: pd.DataFrame):
|
62 |
+
df_tl = df_in[df_in["Experiment"] == "COMET_TL_Ft2Stp"]
|
|
|
|
|
|
|
|
|
|
|
|
|
63 |
df_orig = df[df["Experiment"] == "COMET"]
|
64 |
|
65 |
pvs = []
|
|
|
84 |
print("UW-NSD:", corrected_pvs[2])
|
85 |
|
86 |
|
87 |
+
def post_hoc_comet(df_in: pd.DataFrame):
|
88 |
+
df_tl = df_in[df_in["Experiment"] == "COMET_TL_Ft2Stp"]
|
|
|
|
|
|
|
|
|
|
|
|
|
89 |
|
90 |
filter_ = np.array([x in ["BL-N", "SG-NSD", "UW-NSD"] for x in df_tl["Model"]])
|
91 |
|
|
|
116 |
|
117 |
|
118 |
if __name__ == "__main__":
|
119 |
+
parser = argparse.ArgumentParser()
|
120 |
+
parser.add_argument('--file', '-f', type=str, help='CSV file with the metrics')
|
121 |
+
parser.add_argument('--sep', type=str, help='CSV Separator (default: ;)', default=';')
|
122 |
+
parser.add_argument('--outdir', type=str, help='Output directory (default: .)', default='./')
|
123 |
+
args = parser.parse_args()
|
124 |
+
|
125 |
+
assert os.path.exists(args.file), 'CSV file not found'
|
126 |
+
|
127 |
+
df = pd.read_csv(args.file, sep=args.sep)
|
128 |
+
if "Unnamed" in df.columns[0]:
|
129 |
+
df = df.iloc[:, 1:]
|
130 |
+
if any(['_' in m for m in df["Model"].unique()]):
|
131 |
+
df["Model"] = [x.replace("_", "-") for x in df["Model"]]
|
132 |
+
|
133 |
print("\nComparing all contrasts in TRE of all models in the IXI dataset:")
|
134 |
+
post_hoc_ixi(df, args.outdir)
|
135 |
|
136 |
print("\nTransfer learning benefit (COMET):")
|
137 |
+
study_transfer_learning_benefit(df)
|
138 |
|
139 |
print("\nAssessing whether there is a benefit to segmentation-guiding and uncertainty weighting (COMET):")
|
140 |
+
post_hoc_comet(df)
|
141 |
+
|
142 |
+
print('Statsmodels v.: ' + statsmodels.__version__)
|
143 |
+
print('Scipy v.: ' + scipy.__version__)
|