Skip to content

GitLab

Commit 18552d7e authored by sjromuel's avatar sjromuel
Browse files

dice_ct_seg for mr models

parent 69c0844c
Hide whitespace changes
Inline Side-by-side
RobinNet/dice_gt_seg.py
View file @ 18552d7e
......@@ -18,66 +18,181 @@ from nets.Unet import dice_loss
def main():
mr = True
if mr:
file_path = "data/npy_thresh/"
gt_types = ['ctseg', 'ctthresh', 'unetpred']
for gt_type in gt_types:
#if gt_type == 'ctseg': file_path = "data/npy/"
dice_Score16 = []
hdd_16 = []
all3dhdds = []
for patientnumber in range(16):
y_pred3d = []
patient_dsc = []
patient_hdd = []
if gt_type == 'ctseg':
seg_npy = np.load(file_path + "P" + str(patientnumber + 1).zfill(2) + "_segmr.gipl.npy")
noisygt_npy = np.load(file_path + "P" + str(patientnumber + 1).zfill(2) + "_seg.gipl.npy")
noisygt_npy = cut_mr_ctunetpred("P" + str(patientnumber + 1).zfill(2) + "_seg.gipl.npy", noisygt_npy)
seg_npy = np.reshape(seg_npy, (seg_npy.shape[0], 512, 512, 1))
noisygt_npy = np.reshape(noisygt_npy, (noisygt_npy.shape[0], 512, 512, 1))
test_dataset = tf.data.Dataset.from_tensor_slices((noisygt_npy, seg_npy))
test_dataset = test_dataset.batch(batch_size=1)
elif gt_type == 'unetpred':
seg_npy = np.load(file_path + "P" + str(patientnumber + 1).zfill(2) + "_segmr.gipl.npy")
noisygt_npy = np.load(file_path + "P" + str(patientnumber + 1).zfill(2) + "_mr_ctunetpred.gipl.npy")
seg_npy = np.reshape(seg_npy, (seg_npy.shape[0], 512, 512, 1))
noisygt_npy = np.reshape(noisygt_npy, (noisygt_npy.shape[0], 512, 512, 1))
test_dataset = tf.data.Dataset.from_tensor_slices((noisygt_npy, seg_npy))
test_dataset = test_dataset.batch(batch_size=1)
elif gt_type == 'ctthresh':
seg_npy = np.load(file_path + "P" + str(patientnumber + 1).zfill(2) + "_segmr.gipl.npy")
noisygt_npy = np.load(file_path + "P" + str(patientnumber + 1).zfill(2) + "_100_thresh.gipl.npy")
noisygt_npy = cut_mr_ctunetpred("P" + str(patientnumber + 1).zfill(2) + "_seg.gipl.npy",
noisygt_npy)
seg_npy = np.append(seg_npy,
np.load(file_path + "P" + str(patientnumber + 1).zfill(2) + "_segmr.gipl.npy"),
axis=0)
noisygt_npy = np.append(noisygt_npy, cut_mr_ctunetpred("P" + str(patientnumber + 1).zfill(2) + "_seg.gipl.npy", np.load(
file_path + "P" + str(patientnumber + 1).zfill(2) + "_250_thresh.gipl.npy")), axis=0)
seg_npy = np.append(seg_npy,
np.load(file_path + "P" + str(patientnumber + 1).zfill(2) + "_segmr.gipl.npy"),
axis=0)
noisygt_npy = np.append(noisygt_npy, cut_mr_ctunetpred("P" + str(patientnumber + 1).zfill(2) + "_seg.gipl.npy", np.load(
file_path + "P" + str(patientnumber + 1).zfill(2) + "_400_thresh.gipl.npy")), axis=0)
seg_npy = np.reshape(seg_npy, (seg_npy.shape[0], 512, 512, 1))
noisygt_npy = np.reshape(noisygt_npy, (noisygt_npy.shape[0], 512, 512, 1))
test_dataset = tf.data.Dataset.from_tensor_slices((noisygt_npy, seg_npy))
test_dataset = test_dataset.batch(batch_size=1)
for features in test_dataset:
noisygt, seg = features
noisygt = onehotencode(noisygt, autoencoder=True)
seg = onehotencode(seg, autoencoder=True)
slice_dsc = dice_loss(noisygt, seg)
patient_dsc.append(slice_dsc)
try:
y_true_np = np.squeeze(seg[0, :, :, 0].numpy() > 0.5)
y_true_np = y_true_np.astype(np.float_)
pred_np = np.squeeze(noisygt[0, :, :, 0].numpy() > 0.5)
pred_np = pred_np.astype(np.float_)
hausdorff_distance_filter = sitk.HausdorffDistanceImageFilter()
hausdorff_distance_filter.Execute(sitk.GetImageFromArray(y_true_np),
sitk.GetImageFromArray(pred_np))
patient_hdd.append(hausdorff_distance_filter.GetHausdorffDistance())
if y_pred3d == []:
y_pred3d = np.reshape(pred_np, newshape=(1, 512, 512))
y_true3d = np.reshape(y_true_np, newshape=(1, 512, 512))
else:
y_pred3d = np.append(np.reshape(pred_np, newshape=(1, 512, 512)), y_pred3d, axis=0)
y_true3d = np.append(np.reshape(y_true_np, newshape=(1, 512, 512)), y_true3d, axis=0)
except:
pass
print("Dice Score for patient", patientnumber, ":", np.mean(patient_dsc))
dice_Score16.append(patient_dsc)
file_path = "data/npy/"
gt_types = ['ctfgt2', 'thresh']
for gt_type in gt_types:
dice_Score16 = []
hdd_16 = []
for patientnumber in range(16):
patient_dsc = []
patient_hdd = []
try:
hausdorff_distance_filter = sitk.HausdorffDistanceImageFilter()
hausdorff_distance_filter.Execute(sitk.GetImageFromArray(y_true3d),
sitk.GetImageFromArray(y_pred3d))
hdd3d = hausdorff_distance_filter.GetHausdorffDistance()
print("3D HDD for patient", patientnumber, ":", hdd3d)
all3dhdds.append(hdd3d)
except:
pass
if gt_type == 'ctfgt2':
seg_npy = np.load(file_path+"P"+str(patientnumber+1).zfill(2)+"_seg.gipl.npy")
noisygt_npy = np.load(file_path+"P"+str(patientnumber+1).zfill(2)+"_ctfgt2.gipl.npy")
seg_npy = np.reshape(seg_npy, (seg_npy.shape[0], 512, 512, 1))
noisygt_npy = np.reshape(noisygt_npy, (noisygt_npy.shape[0], 512, 512, 1))
test_dataset = tf.data.Dataset.from_tensor_slices((noisygt_npy, seg_npy))
test_dataset = test_dataset.batch(batch_size=1)
elif gt_type == 'thresh':
file_path = "data/npy_thresh/"
seg_npy = np.load(file_path + "P" + str(patientnumber + 1).zfill(2) + "_seg.gipl.npy")
noisygt_npy = np.load(file_path + "P" + str(patientnumber + 1).zfill(2) + "_100_thresh.gipl.npy")
seg_npy = np.append(seg_npy, np.load(file_path + "P" + str(patientnumber + 1).zfill(2) + "_seg.gipl.npy"), axis=0)
noisygt_npy = np.append(noisygt_npy, np.load(file_path + "P" + str(patientnumber + 1).zfill(2) + "_250_thresh.gipl.npy"), axis=0)
seg_npy = np.append(seg_npy, np.load(file_path + "P" + str(patientnumber + 1).zfill(2) + "_seg.gipl.npy"), axis=0)
noisygt_npy = np.append(noisygt_npy, np.load(file_path + "P" + str(patientnumber + 1).zfill(2) + "_400_thresh.gipl.npy"), axis=0)
seg_npy = np.reshape(seg_npy, (seg_npy.shape[0], 512, 512, 1))
noisygt_npy = np.reshape(noisygt_npy, (noisygt_npy.shape[0], 512, 512, 1))
test_dataset = tf.data.Dataset.from_tensor_slices((noisygt_npy, seg_npy))
test_dataset = test_dataset.batch(batch_size=1)
print("All 3d-HDDs for", gt_type, ": ", all3dhdds)
print("Dice-Score mean: ", np.mean(dice_Score16))
print("Dice-Score std: ", np.std(dice_Score16))
print("HDD mean: ", np.mean(all3dhdds))
print("HDD std: ", np.std(all3dhdds))
else:
file_path = "data/npy/"
gt_types = ['ctfgt2', 'thresh']
for gt_type in gt_types:
dice_Score16 = []
hdd_16 = []
all3dhdds = []
for patientnumber in range(16):
y_pred3d = []
patient_dsc = []
patient_hdd = []
if gt_type == 'ctfgt2':
seg_npy = np.load(file_path+"P"+str(patientnumber+1).zfill(2)+"_seg.gipl.npy")
noisygt_npy = np.load(file_path+"P"+str(patientnumber+1).zfill(2)+"_ctfgt2.gipl.npy")
seg_npy = np.reshape(seg_npy, (seg_npy.shape[0], 512, 512, 1))
noisygt_npy = np.reshape(noisygt_npy, (noisygt_npy.shape[0], 512, 512, 1))
test_dataset = tf.data.Dataset.from_tensor_slices((noisygt_npy, seg_npy))
test_dataset = test_dataset.batch(batch_size=1)
elif gt_type == 'thresh':
file_path = "data/npy_thresh/"
seg_npy = np.load(file_path + "P" + str(patientnumber + 1).zfill(2) + "_seg.gipl.npy")
noisygt_npy = np.load(file_path + "P" + str(patientnumber + 1).zfill(2) + "_100_thresh.gipl.npy")
seg_npy = np.append(seg_npy, np.load(file_path + "P" + str(patientnumber + 1).zfill(2) + "_seg.gipl.npy"), axis=0)
noisygt_npy = np.append(noisygt_npy, np.load(file_path + "P" + str(patientnumber + 1).zfill(2) + "_250_thresh.gipl.npy"), axis=0)
seg_npy = np.append(seg_npy, np.load(file_path + "P" + str(patientnumber + 1).zfill(2) + "_seg.gipl.npy"), axis=0)
noisygt_npy = np.append(noisygt_npy, np.load(file_path + "P" + str(patientnumber + 1).zfill(2) + "_400_thresh.gipl.npy"), axis=0)
seg_npy = np.reshape(seg_npy, (seg_npy.shape[0], 512, 512, 1))
noisygt_npy = np.reshape(noisygt_npy, (noisygt_npy.shape[0], 512, 512, 1))
test_dataset = tf.data.Dataset.from_tensor_slices((noisygt_npy, seg_npy))
test_dataset = test_dataset.batch(batch_size=1)
for features in test_dataset:
noisygt, seg = features
noisygt = onehotencode(noisygt, autoencoder=True)
seg = onehotencode(seg, autoencoder=True)
slice_dsc = dice_loss(noisygt, seg)
patient_dsc.append(slice_dsc)
y_true_np = np.squeeze(seg[0, :, :, 0].numpy() > 0.5)
y_true_np = y_true_np.astype(np.float_)
pred_np = np.squeeze(noisygt[0, :, :, 0].numpy() > 0.5)
pred_np = pred_np.astype(np.float_)
hausdorff_distance_filter = sitk.HausdorffDistanceImageFilter()
hausdorff_distance_filter.Execute(sitk.GetImageFromArray(y_true_np), sitk.GetImageFromArray(pred_np))
patient_hdd.append(hausdorff_distance_filter.GetHausdorffDistance())
if y_pred3d == []:
y_pred3d = np.reshape(pred_np, newshape=(1, 512, 512))
y_true3d = np.reshape(y_true_np, newshape=(1, 512, 512))
else:
y_pred3d = np.append(np.reshape(pred_np, newshape=(1, 512, 512)), y_pred3d, axis=0)
y_true3d = np.append(np.reshape(y_true_np, newshape=(1, 512, 512)), y_true3d, axis=0)
for features in test_dataset:
noisygt, seg = features
noisygt = onehotencode(noisygt, autoencoder=True)
seg = onehotencode(seg, autoencoder=True)
slice_dsc = dice_loss(noisygt, seg)
patient_dsc.append(slice_dsc)
y_true_np = np.squeeze(seg[0, :, :, 0].numpy() > 0.5)
y_true_np = y_true_np.astype(np.float_)
pred_np = np.squeeze(noisygt[0, :, :, 0].numpy() > 0.5)
pred_np = pred_np.astype(np.float_)
hausdorff_distance_filter = sitk.HausdorffDistanceImageFilter()
hausdorff_distance_filter.Execute(sitk.GetImageFromArray(y_true_np), sitk.GetImageFromArray(pred_np))
patient_hdd.append(hausdorff_distance_filter.GetHausdorffDistance())
hdd_16.append(np.mean(patient_hdd))
dice_Score16.append(np.mean(patient_dsc))
print(gt_type, dice_Score16)
print(gt_type, "HDD: ", hdd_16)
print("Dice-Score mean: ", np.mean(dice_Score16))
print("Dice-Score std: ", np.std(dice_Score16))
print("HDD mean: ", np.mean(hdd_16))
print("HDD std: ", np.std(hdd_16))
try:
hausdorff_distance_filter = sitk.HausdorffDistanceImageFilter()
hausdorff_distance_filter.Execute(sitk.GetImageFromArray(y_true3d),
sitk.GetImageFromArray(y_pred3d))
hdd3d = hausdorff_distance_filter.GetHausdorffDistance()
print("3D HDD for patient", patientnumber, ":", hdd3d)
all3dhdds.append(hdd3d)
except:
pass
print("All 3d-HDDs for", gt_type, ": ", all3dhdds)
#print("Dice-Score mean: ", np.mean(dice_Score16))
#print("Dice-Score std: ", np.std(dice_Score16))
print("HDD mean: ", np.mean(all3dhdds))
print("HDD std: ", np.std(all3dhdds))
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment