Data preprocessing

truenet_tumseg/truenet_tumorseg/truenet_tumseg_data_preprocessing.py

+from __future__ import absolute_import
+from __future__ import division
+from __future__ import print_function
+
+import numpy as np
+
+#=========================================================================================
+# Truenet data preprocessing function
+# Vaanathi Sundaresan
+# 10-03-2021, Oxford
+#=========================================================================================
+
+def preprocess_data(data):
+    '''
+    Min-max intensity normalisation of data
+    :param data: input data
+    :return: normalised data
+    '''
+    return (2 * data / np.amax(np.reshape(data,[-1,1]))) - 1
+
+def preprocess_data_gauss(data):
+    '''
+    Gaussian intensity normalisation of data
+    :param data: input data
+    :return: Gaussian normalised data
+    '''
+    brain1 = data > 0
+    brain = brain1 > 0
+    data = data - np.mean(data[brain])      
+    den = np.std(data[brain])
+    if den == 0:
+        den = 1
+    data = data/den
+    data[brain==0] = np.min(data)
+    return data
+
+def cut_zeros1d(im_array):
+    '''
+    Find the window for cropping the data closer to the brain
+    :param im_array: input array
+    :return: starting and end indices, and length of non-zero intensity values
+    '''
+    im_list = list(im_array > 0)
+    start_index = im_list.index(1)
+    end_index = im_list[::-1].index(1)
+    length = len(im_array[start_index:])-end_index
+    return start_index, end_index, length
+
+def tight_crop_data(img_data):
+    '''
+    Crop the data tighter to the brain
+    :param img_data: input array
+    :return: cropped image and the bounding box coordinates and dimensions.
+    '''
+    row_sum = np.sum(np.sum(img_data,axis=1),axis=1)
+    col_sum = np.sum(np.sum(img_data,axis=0),axis=1)
+    stack_sum = np.sum(np.sum(img_data,axis=1),axis=0)
+    rsid, reid, rlen = cut_zeros1d(row_sum)
+    csid, ceid, clen = cut_zeros1d(col_sum)
+    ssid, seid, slen = cut_zeros1d(stack_sum)
+    return img_data[rsid:rsid+rlen, csid:csid+clen, ssid:ssid+slen], [rsid, rlen, csid, clen, ssid, slen]
+