--- a
+++ b/scripts/data_exploration.py
@@ -0,0 +1,115 @@
+#==============================================================================#
+#  Author:       Dominik Müller                                                #
+#  Copyright:    2020 IT-Infrastructure for Translational Medical Research,    #
+#                University of Augsburg                                        #
+#                                                                              #
+#  This program is free software: you can redistribute it and/or modify        #
+#  it under the terms of the GNU General Public License as published by        #
+#  the Free Software Foundation, either version 3 of the License, or           #
+#  (at your option) any later version.                                         #
+#                                                                              #
+#  This program is distributed in the hope that it will be useful,             #
+#  but WITHOUT ANY WARRANTY; without even the implied warranty of              #
+#  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the               #
+#  GNU General Public License for more details.                                #
+#                                                                              #
+#  You should have received a copy of the GNU General Public License           #
+#  along with this program.  If not, see <http://www.gnu.org/licenses/>.       #
+#==============================================================================#
+#-----------------------------------------------------#
+#                   Library imports                   #
+#-----------------------------------------------------#
+from miscnn.data_loading.interfaces import NIFTI_interface
+from miscnn import Data_IO
+from tqdm import tqdm
+import os
+import numpy as np
+import pandas as pd
+
+#-----------------------------------------------------#
+#                    Configurations                   #
+#-----------------------------------------------------#
+# Data directory
+path_data = "data"
+
+#-----------------------------------------------------#
+#                   Data Exploration                  #
+#-----------------------------------------------------#
+# Initialize Data IO Interface for NIfTI data
+interface = NIFTI_interface(channels=1, classes=3)
+
+# Create Data IO object to load and write samples in the file structure
+data_io = Data_IO(interface, path_data, delete_batchDir=True)
+
+# Access all available samples in our file structure
+sample_list = data_io.get_indiceslist()
+sample_list.sort()
+
+# Print out the sample list
+print("Sample list:", sample_list)
+
+# Now let's load each sample and obtain collect diverse information from them
+sample_data = {}
+for index in tqdm(sample_list):
+    # Sample loading
+    sample = data_io.sample_loader(index, load_seg=True)
+    # Create an empty list for the current asmple in our data dictionary
+    sample_data[index] = []
+    # Store the volume shape
+    sample_data[index].append(sample.img_data.shape)
+    # Identify minimum and maximum volume intensity
+    sample_data[index].append(sample.img_data.min())
+    sample_data[index].append(sample.img_data.max())
+    # Store voxel spacing
+    sample_data[index].append(sample.details["spacing"])
+    # Identify and store class distribution
+    unique_data, unique_counts = np.unique(sample.seg_data, return_counts=True)
+    class_freq = unique_counts / np.sum(unique_counts)
+    class_freq = np.around(class_freq, decimals=6)
+    sample_data[index].append(tuple(class_freq))
+
+# Transform collected data into a pandas dataframe
+df = pd.DataFrame.from_dict(sample_data, orient="index",
+                            columns=["vol_shape", "vol_minimum",
+                                     "vol_maximum", "voxel_spacing",
+                                     "class_frequency"])
+
+# Print out the dataframe to console
+with pd.option_context('display.max_rows', None, 'display.max_columns', None):
+    print(df)
+
+# Calculate mean and median shape sizes
+shape_list = np.array(df["vol_shape"].tolist())
+for i, a in enumerate(["X", "Y", "Z"]):
+    print(a + "-Axes Mean:", np.mean(shape_list[:,i]))
+    print(a + "-Axes Median:", np.median(shape_list[:,i]))
+
+# Calculate average class frequency
+df_classes = pd.DataFrame(df["class_frequency"].tolist(),
+                          columns=["background", "lung_L",
+                                   "lung_R", "infection"])
+print(df_classes.mean(axis=0))
+
+## The resolution of the volumes are fixed for the x,y axes to 512x512 for the
+## "coronacases" data and 630x630 for the "radiopaedia" volumes.
+## Strangely, a single sample "radiopaedia_14_85914_0" has only 630x401 for x,y.
+## Furthermore, the sample "radiopaedia_10_85902_3" has a ~ 8x times higher
+## z axis than the other radiopaedia samples.
+## Overall, the aim to identify suited resampling coordinates will be tricky.
+##
+## When looking on the voxel spacings from the affinity matrix of the CTs,
+## we see that the x,y spacings or quite similiar whereas the z axis reveal
+## large differences. Resampling to a common voxel spacing is mandatory.
+##
+## We see that the radiopaedia ct volumes have already preprocessed
+## In the description of the data set, it is documentated that they
+## used a clipping to the lung window [-1250, 250] and normalized these values
+## to [0,255]. Therefore we have to perform the identical preprocessing to the
+## other ct volumes as well
+##
+## Also, the class frequency reveal a heavy bias towards the background class
+## as expected in medical image segmentation
+##
+## In COVID-19-CT-Seg dataset, the last 10 cases from Radiopaedia have been
+## adjusted to lung window [-1250,250], and then normalized to [0,255],
+## we recommend to adust the first 10 cases from Coronacases with the same method.