--- a
+++ b/src/preprocess_b0.py
@@ -0,0 +1,108 @@
+#!/usr/bin/env python
+import os
+from os import path
+import sys
+import numpy as np
+import nibabel as nib
+import argparse
+import pathlib
+
+# Function to preprocess the dwi cases
+def process_trainingdata(dwib0_arr):
+    count = 0
+    for b0 in dwib0_arr:
+        img = nib.load(b0)
+        imgF32 = img.get_fdata().astype(np.float32)
+        ''' Intensity based segmentation of MR images is hampered by radio frerquency field
+            inhomogeneity causing intensity variation. The intensity range is typically
+            scaled between the highest and lowest signal in the Image. Intensity values
+            of the same tissue can vary between scans. The pixel value in images must be
+            scaled prior to providing the images as input to CNN. The data is projected in to
+            a predefined range [0,1] '''
+        p = np.percentile(imgF32, 99)
+        imgF32_sagittal = imgF32 / p 			          # sagittal view
+        imgF32_sagittal[ imgF32_sagittal < 0 ] =  sys.float_info.epsilon	      
+        imgF32_sagittal[ imgF32_sagittal > 1 ] = 1 		  
+        imgF32_coronal = np.swapaxes(imgF32_sagittal,0,1) # coronal view
+        imgF32_axial = np.swapaxes(imgF32_sagittal,0,2)   # Axial view
+
+        # dwi volume data is written to the binary file
+        imgF32_sagittal.tofile(sagittal_f_handle)
+        imgF32_coronal.tofile(coronal_f_handle)
+        imgF32_axial.tofile(axial_f_handle)
+
+        print('Case ' + str(count) + ' done')
+        count = count + 1
+
+    # Closing the binary file
+    sagittal_f_handle.close()
+    axial_f_handle.close()
+    coronal_f_handle.close()
+
+# parser module for input arguments
+SUFFIX_TXT = "txt"
+parser = argparse.ArgumentParser()
+parser.add_argument('-i', action='store', dest='b0', type=str,
+                        help="txt file containing list of /path/to/b0, one path in each line")
+args = parser.parse_args()
+
+try:
+    args = parser.parse_args()
+    if len(sys.argv) == 1:
+        parser.print_help()
+        parser.error('too few arguments')
+        sys.exit(0)
+
+except SystemExit:
+    sys.exit(0)
+
+if args.b0:
+    f = pathlib.Path(args.b0)
+    if f.exists():
+        print ("File exist")
+        filename = args.b0
+    else:
+        print ("File not found")
+        sys.exit(1)
+
+    # Input caselist.txt
+    if filename.endswith(SUFFIX_TXT):
+        with open(filename) as f:
+            dwib0_arr = f.read().splitlines()
+
+storage = path.dirname(dwib0_arr[0])
+# dwi cases will be written to the below binary files
+sagittal_bin_file = storage + '/sagittal-binary-dwi'
+coronal_bin_file = storage + '/coronal-binary-dwi'
+axial_bin_file = storage + '/axial-binary-dwi'
+
+# The above binary files will be converted to 3D numpy array
+sagittal_trainingdata = storage + '/sagittal-traindata-dwi.npy'
+coronal_trainingdata = storage + '/coronal-traindata-dwi.npy'
+axial_trainingdata = storage + '/axial-traindata-dwi.npy'
+
+# Open the binary file for writing
+sagittal_f_handle = open(sagittal_bin_file, 'wb')
+coronal_f_handle = open(coronal_bin_file, 'wb')
+axial_f_handle = open(axial_bin_file, 'wb')
+
+process_trainingdata(dwib0_arr)
+x_dim=len(dwib0_arr)*256
+y_dim=256
+z_dim=256
+
+# Open the binary file and convert it to 3D numpy array
+merge_sagittal = np.memmap(sagittal_bin_file, dtype=np.float32, mode='r+', shape=(x_dim, y_dim, z_dim))
+print("Saving sagittal training data to disk")
+np.save(sagittal_trainingdata, merge_sagittal)
+os.unlink(sagittal_bin_file)
+
+merge_coronal = np.memmap(coronal_bin_file, dtype=np.float32, mode='r+', shape=(x_dim, y_dim, z_dim))
+print("Saving coronal training data to disk")
+np.save(coronal_trainingdata, merge_coronal)
+os.unlink(coronal_bin_file)
+
+merge_axial = np.memmap(axial_bin_file, dtype=np.float32, mode='r+', shape=(x_dim, y_dim, z_dim))
+print("Saving axial training data to disk")
+np.save(axial_trainingdata, merge_axial)
+os.unlink(axial_bin_file)