remove bad config, use torchIO to produce train/test splits. visualiz…

…e CT volumes in tests

TestDatasetLoader.py

@@ -4,6 +4,7 @@
 
 import matplotlib.pyplot as plt
 import matplotlib.patches as patches
+import matplotlib.animation as animation
 
 dataLoader = kaggleDataLoader.KaggleDataLoader()
 
@@ -40,4 +41,38 @@
 fig, ax = plt.subplots(1,1, figsize=(10, 10))
 ax.imshow(imSlicePxArr)
 ax.add_patch(rect)
+plt.show()
+
+train, val = dataLoader.loadDatasetAsClassifier()
+
+subject1 = train[0]
+
+fig, ax = plt.subplots()
+ims = []
+for sagittal_slice_tensor in subject1.ct.data[0]:
+    im = ax.imshow(sagittal_slice_tensor.numpy(), cmap=plt.cm.bone, animated=True)
+    ims.append([im])
+
+ani = animation.ArtistAnimation(fig, ims, interval=50, blit=True,
+                                repeat_delay=1000)
+plt.show()
+
+fig, ax = plt.subplots()
+ims = []
+for coronal_slice_tensor in subject1.ct.data[0].permute(1,2,0):
+    im = ax.imshow(coronal_slice_tensor.numpy(), cmap=plt.cm.bone, animated=True)
+    ims.append([im])
+
+ani = animation.ArtistAnimation(fig, ims, interval=50, blit=True,
+                                repeat_delay=1000)
+plt.show()
+
+fig, ax = plt.subplots()
+ims = []
+for axial_slice_tensor in subject1.ct.data[0].permute(2,1,0):
+    im = ax.imshow(axial_slice_tensor.numpy(), cmap=plt.cm.bone, animated=True)
+    ims.append([im])
+
+ani = animation.ArtistAnimation(fig, ims, interval=50, blit=True,
+                                repeat_delay=1000)
 plt.show()

environment-CPU.yml

@@ -1,18 +1,11 @@
-# To run: conda env create -f environment-lnx-cuda.yml
+# To run: conda env create -f environment-CPU.yml
 name: csna
 channels:
     - conda-forge
 dependencies:
-    - python=3.6
+    - python=3.9
     - pip
-    - Pillow
     - matplotlib
-    - numpy
     - pandas
     - pip:
-        - torch
-        - torchvision
-        - kaggle
-        - pydicom
-        - opencv-python
-        - nibabel
+        - -r requirements.txt

kaggleDataLoader.py

@@ -1,15 +1,16 @@
-
-from PIL import Image
-
+import copy
 import pandas as pd
 import pydicom
 import nibabel as nib
 import numpy as np
 import json
 import cv2
 import os
+os.environ['KMP_DUPLICATE_LIB_OK']='True'
 from tqdm import tqdm
-import random
+
+import torchio as tio
+import torch
 with open('config.json', 'r') as f:
     paths = json.load(f)
 
@@ -95,10 +96,38 @@ def loadSegmentationsForPatient(self, patientID):
         return segmentations
 
     ## Dataset generator functions
-    def loadDatasetAsClassifier(self):
+    def loadDatasetAsClassifier(self, trainPercentage=0.90,train_aug=None,val_aug=None):
         """
         prepare full dataset for training
         """
+
+        HOUNSFIELD_AIR, HOUNSFIELD_BONE = -1000, 1900
+        clamp = tio.Clamp(out_min=HOUNSFIELD_AIR, out_max=HOUNSFIELD_BONE)
+        rescale = tio.RescaleIntensity(percentiles=(0.5, 99.5))
+        preprocess_intensity = tio.Compose([
+            clamp,
+            rescale,
+        ])
+        normalize_orientation = tio.ToCanonical()
+        downsample = tio.Resample(1)
+        preprocess_spatial = tio.Compose([
+            normalize_orientation,
+            downsample,
+        ])
+        preprocess = tio.Compose([
+            preprocess_intensity,
+            preprocess_spatial,
+        ])
+
+        trainSet = tio.datasets.RSNACervicalSpineFracture(RSNA_2022_PATH)
+        num_subjects = len(trainSet)
+        num_train = int(trainPercentage*num_subjects)
+        num_val = num_subjects - num_train
+        train_set, val_set = torch.utils.data.random_split(trainSet,[num_train,num_val])
+        train_set.dataset.set_transform(preprocess)
+        return train_set, val_set
+
+
     def loadDatasetAsDetector(self):
         """
         prepare full dataset for training

requirements.txt

@@ -1,11 +1,13 @@
-scipy==1.2.1
-numpy==1.17.0
-matplotlib==3.1.2
-opencv_python==4.1.2.30
-torch==1.2.0
-torchvision==0.4.0
-tqdm==4.60.0
-Pillow==8.2.0
-h5py==2.10.0
+scipy
+numpy
+matplotlib
+opencv_python
+torch
+torchvision
+tqdm
+Pillow
+h5py
 pylibjpeg
-pylibjpeg-libjpeg
+pylibjpeg-libjpeg
+torchio
+pydicom