-

app.py

@@ -1,4 +1,144 @@
 import gradio as gr
+from monai.utils import first, set_determinism
+from monai.transforms import(
+ Compose,
+ AddChanneld,
+ LoadImaged,
+ Resized,
+ ToTensord,
+ Spacingd,
+ Orientationd,
+ ScaleIntensityRanged,
+ CropForegroundd,
+ Activations,
+)
+
+from monai.networks.nets import UNet
+from monai.networks.layers import Norm
+from monai.data import CacheDataset, DataLoader, Dataset
+
+import torch
+import matplotlib.pyplot as plt
+
+import os
+from glob import glob
+import numpy as np
+
+from monai.inferers import sliding_window_inference
+
+in_dir = 'Data_Train_Test'
+model_dir = 'results/results'
+
+train_loss = np.load(os.path.join(model_dir, 'loss_train.npy'))
+train_metric = np.load(os.path.join(model_dir, 'metric_train.npy'))
+test_loss = np.load(os.path.join(model_dir, 'loss_test.npy'))
+test_metric = np.load(os.path.join(model_dir, 'metric_test.npy'))
+
+def plot_result:
+ plt.figure("Results 25 june", (12, 6))
+ plt.subplot(2, 2, 1)
+ plt.title("Train dice loss")
+ x = [i + 1 for i in range(len(train_loss))]
+ y = train_loss
+ plt.xlabel("epoch")
+ plt.plot(x, y)
+ 
+ plt.subplot(2, 2, 2)
+ plt.title("Train metric DICE")
+ x = [i + 1 for i in range(len(train_metric))]
+ y = train_metric
+ plt.xlabel("epoch")
+ plt.plot(x, y)
+ 
+ plt.subplot(2, 2, 3)
+ plt.title("Test dice loss")
+ x = [i + 1 for i in range(len(test_loss))]
+ y = test_loss
+ plt.xlabel("epoch")
+ plt.plot(x, y)
+ 
+ plt.subplot(2, 2, 4)
+ plt.title("Test metric DICE")
+ x = [i + 1 for i in range(len(test_metric))]
+ y = test_metric
+ plt.xlabel("epoch")
+ plt.plot(x, y)
+ 
+ plt.show()
+
+
+
+path_train_volumes = sorted(glob(os.path.join(in_dir, "TrainVolumes", "*.nii.gz")))
+path_train_segmentation = sorted(glob(os.path.join(in_dir, "TrainSegmentation", "*.nii.gz")))
+
+path_test_volumes = sorted(glob(os.path.join(in_dir, "TestVolumes", "*.nii.gz")))
+path_test_segmentation = sorted(glob(os.path.join(in_dir, "TestSegmentation", "*.nii.gz")))
+
+train_files = [{"vol": image_name, "seg": label_name} for image_name, label_name in zip(path_train_volumes, path_train_segmentation)]
+test_files = [{"vol": image_name, "seg": label_name} for image_name, label_name in zip(path_test_volumes, path_test_segmentation)]
+test_files = test_files[0:9]
+
+
+test_transforms = Compose(
+ [
+ LoadImaged(keys=["vol", "seg"]),
+ AddChanneld(keys=["vol", "seg"]),
+ Spacingd(keys=["vol", "seg"], pixdim=(1.5,1.5,1.0), mode=("bilinear", "nearest")),
+ Orientationd(keys=["vol", "seg"], axcodes="RAS"),
+ ScaleIntensityRanged(keys=["vol"], a_min=-200, a_max=200,b_min=0.0, b_max=1.0, clip=True), 
+ CropForegroundd(keys=['vol', 'seg'], source_key='vol'),
+ Resized(keys=["vol", "seg"], spatial_size=[128,128,64]), 
+ ToTensord(keys=["vol", "seg"]),
+ ]
+)
+
+test_ds = Dataset(data=test_files, transform=test_transforms)
+test_loader = DataLoader(test_ds, batch_size=1)
+
+device = torch.device("cuda:0")
+model = UNet(
+ dimensions=3,
+ in_channels=1,
+ out_channels=2,
+ channels=(16, 32, 64, 128, 256), 
+ strides=(2, 2, 2, 2),
+ num_res_units=2,
+ norm=Norm.BATCH,
+).to(device)
+
+model.load_state_dict(torch.load(
+ os.path.join(model_dir, "best_metric_model.pth")))
+model.eval()
+
+
+sw_batch_size = 4
+roi_size = (128, 128, 64)
+with torch.no_grad():
+ test_patient = first(test_loader)
+ t_volume = test_patient['vol']
+ #t_segmentation = test_patient['seg']
+ 
+ test_outputs = sliding_window_inference(t_volume.to(device), roi_size, sw_batch_size, model)
+ sigmoid_activation = Activations(sigmoid=True)
+ test_outputs = sigmoid_activation(test_outputs)
+ test_outputs = test_outputs > 0.53
+ 
+# plot the slice [:, :, 80]
+def predict:
+ i =55
+ plt.figure("check", (18, 6))
+ plt.subplot(1, 3, 1)
+ plt.title(f"image {i}")
+ plt.imshow(test_patient["vol"][0, 0, :, :, i], cmap="gray")
+ plt.subplot(1, 3, 2)
+ plt.title(f"label {i}")
+ plt.imshow(test_patient["seg"][0, 0, :, :, i] != 0)
+ plt.subplot(1, 3, 3)
+ plt.title(f"output {i}")
+ plt.imshow(test_outputs.detach().cpu()[0, 1, :, :, i])
+ plt.show()
+
+
 
 gr.Interface(fn=predict, 
  inputs=gr.inputs.Image(type="pil"),