def main():
parser = argparse.ArgumentParser(description='Radiomics results')
parser.add_argument('-feat', '--feat', metavar='feat',
nargs='+', dest='feat', type=str, required=True,
help='List of patient feature files (HDF)')
parser.add_argument('-class', '--class', metavar='class',
nargs='+', dest='classs', type=str, required=True,
help='Classification of patients (text)')
parser.add_argument('-out', '--out', metavar='out',
nargs='+', dest='out', type=str, required=True,
help='Output png file')
args = parser.parse_args()
if type(args.classs) is list:
args.classs = ''.join(args.classs)
if type(args.out) is list:
args.out = ''.join(args.out)
if type(args.feat) is list and len(args.feat) == 1:
args.feat = ''.join(args.feat)
if os.path.isdir(args.feat):
args.feat = glob.glob(args.feat + '/features_*.hdf5')
args.feat = natsorted(args.feat)
# Read and stack the features
print("Reading features.")
image_features_temp = list()
for i_feat in range(len(args.feat)):
feat = dict()
print args.feat[i_feat]
feat_temp = pd.read_hdf(args.feat[i_feat])
feat_temp = feat_temp.image_features
for feattype in feat_temp.keys():
feat_type = feat_temp[feattype]
for subtype in feat_type.keys():
subfeat = feat_type[subtype]
for k in subfeat.keys():
feat[k] = subfeat[k]
image_features_temp.append(feat)
# Get the mutation labels and patient IDs
print("Reading class labels.")
mutation_type = [['WT']]
mutation_data, image_features = gp.findmutationdata(args.classs,
mutation_type,
args.feat,
image_features_temp)
labels = image_features[0].keys()
featvect = dict()
flab = dict()
for l in labels:
featvect[l] = {"all": [], "1": [], "0": []}
flab[l] = {"all": [], "1": [], "0": []}
# Stack per feature type and class
print("Stacking features.")
mutation_label = mutation_data['mutation_label'].tolist()[0]
patient_IDs = mutation_data['patient_IDs'].tolist()
for imfeat, label, pid in zip(image_features, mutation_label, patient_IDs):
for fl in labels:
featvect[fl]['all'].append(imfeat[fl])
flab[fl]['all'].append(pid)
if label[0] == 0:
featvect[fl]['0'].append(imfeat[fl])
flab[fl]['0'].append(pid)
else:
featvect[fl]['1'].append(imfeat[fl])
flab[fl]['1'].append(pid)
# Create the boxplots
print("Generating boxplots.")
# Split in 5x5 figures.
nfig = np.ceil(len(labels) / 25.0)
labels = sorted(labels)
for fi in range(0, int(nfig)):
f = plt.figure()
fignum = 1
for i in range(fi*25, min((fi+1)*25, len(labels))):
ax = plt.subplot(5, 5, fignum)
lab = labels[i]
plt.subplots_adjust(hspace=0.3, wspace=0.2)
ax.scatter(np.ones(len(featvect[lab]['all'])),
featvect[lab]['all'],
color='blue')
ax.scatter(np.ones(len(featvect[lab]['1']))*2.0,
featvect[lab]['1'],
color='red')
ax.scatter(np.ones(len(featvect[lab]['0']))*3.0,
featvect[lab]['0'],
color='green')
plt.boxplot([featvect[lab]['all'], featvect[lab]['1'], featvect[lab]['0']])
fz = 5 # Works best after saving
ax.set_title(lab, fontsize=fz)
for tick in ax.xaxis.get_major_ticks():
tick.label.set_fontsize(fz)
for tick in ax.yaxis.get_major_ticks():
tick.label.set_fontsize(fz)
fignum += 1
# Maximize figure to get correct spacings
mng = plt.get_current_fig_manager()
mng.resize(*mng.window.maxsize())
# plt.show()
# High DTI to make sure we save the maximized image
f.savefig(args.out + ('_{}.png').format(str(fi)), dpi=600)
print(("Boxplot saved as {} !").format(args.out + ('_{}.png').format(str(fi))))
def main():
parser = argparse.ArgumentParser(description='Radiomics results')
parser.add_argument('-feat',
'--feat',
metavar='feat',
nargs='+',
dest='feat',
type=str,
required=True,
help='List of patient feature files (HDF)')
parser.add_argument('-class',
'--class',
metavar='class',
nargs='+',
dest='classs',
type=str,
required=True,
help='Classification of patients (text)')
parser.add_argument('-lab',
'--lab',
metavar='lab',
nargs='+',
dest='lab',
type=str,
required=True,
help='Label of two features to plot')
parser.add_argument('-out',
'--out',
metavar='out',
nargs='+',
dest='out',
type=str,
required=True,
help='Output png file')
args = parser.parse_args()
if type(args.classs) is list:
args.classs = ''.join(args.classs)
if type(args.out) is list:
args.out = ''.join(args.out)
if type(args.feat) is list and len(args.feat) == 1:
args.feat = ''.join(args.feat)
if os.path.isdir(args.feat):
args.feat = glob.glob(args.feat + '/features_*.hdf5')
args.feat = natsorted(args.feat)
# Read and stack the features
image_features_temp = list()
for i_feat in range(len(args.feat)):
feat = dict()
feat_temp = pd.read_hdf(args.feat[i_feat])
feat_temp = feat_temp.image_features
for feattype in feat_temp.keys():
feat_type = feat_temp[feattype]
for subtype in feat_type.keys():
subfeat = feat_type[subtype]
for k in subfeat.keys():
feat[k] = subfeat[k]
image_features_temp.append(feat)
# Get the mutation labels and patient IDs
mutation_type = [['GP']]
mutation_data, image_features = gp.findmutationdata(
args.classs, mutation_type, args.feat, image_features_temp)
image_features = image_features.tolist()
# Select the two relevant features
feat1_c0 = list()
feat2_c0 = list()
feat1_c1 = list()
feat2_c1 = list()
mutation_label = mutation_data['mutation_label'].tolist()[0]
patient_IDs = mutation_data['patient_IDs'].tolist()
for imfeat, label, pid in zip(image_features, mutation_label, patient_IDs):
print imfeat[args.lab[0]], pid
if label[0] == 0:
feat1_c0.append(imfeat[args.lab[0]])
feat2_c0.append(imfeat[args.lab[1]])
else:
feat1_c1.append(imfeat[args.lab[0]])
feat2_c1.append(imfeat[args.lab[1]])
# Make a scatter plot
f = plt.figure()
subplot = f.add_subplot(111)
subplot.plot(feat1_c0,
feat2_c0,
linestyle='',
ms=12,
marker='o',
color='navy')
subplot.plot(feat1_c1,
feat2_c1,
linestyle='',
ms=12,
marker='x',
color='red')
# NOTE: arbitrary limits!
plt.xlim([0, 10])
plt.ylim([0, 10])
plt.xlabel(args.lab[0])
plt.ylabel(args.lab[1])
plt.title('Feature scatter plot')
plt.legend()
plt.show()
f.savefig(args.out)
print(("Snapshot saved as {} !").format(args.out))