Scribd
Upload
304 views

Simple It K Tutorial

This tutorial shows how to open and extract features from a medical image (.mhd) file in SimpleITK. It defines functions to load the image, read candidate coordinates from a CSV file, transform coordinates between world and voxel space, extract image patches around candidates, and visualize the patches. These functions are used to open an image and CSV, extract patches around coordinates from the CSV, normalize and visualize the patches.

Uploaded by

MarceloAmancio
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
304 views

Simple It K Tutorial

This tutorial shows how to open and extract features from a medical image (.mhd) file in SimpleITK. It defines functions to load the image, read candidate coordinates from a CSV file, transform coordinates between world and voxel space, extract image patches around candidates, and visualize the patches. These functions are used to open an image and CSV, extract patches around coordinates from the CSV, normalize and visualize the patches.

Uploaded by

MarceloAmancio
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 3

SimpleITK Tutorial(1)

January 29, 2016

1 Tutorial on how to open, visualize and extract some features


from a .mhd Image
This Tutorial will show how to: - Open and read a .mhd image - Visualize a .mhd image - Read a list of
candidates from a .csv file - Transform from world coordinates to voxel coordinates - Extract some features
/ patches of candidates and visualize them To be able to run this tutorial some python libraries / modules
need to be installed: - Simple ITK: a library for handling and processing medical images - Numpy: a
fundamental package for scientific computing with Python - PIL (Python Imaging Library): a library for
adding image processing capabilities to your Python interpreter - Matplotlib: a plotting library for the
Python programming language
We start importing required modules / libraries using the import command from python
In [1]: import SimpleITK as sitk
import numpy as np
import csv
import os
from PIL import Image
import matplotlib.pyplot as plt
%matplotlib inline
We define now a function to: - Open the image - Store it into a numpy array - Extract the following
info: Pixel Spacing, Origin This function takes as input the name of the image and returns: - The array
corresponding to the image (numpyImage) - Origin (numpyOrigin) - PixelSpacing (numpySpacing)
In [2]: def load_itk_image(filename):
itkimage = sitk.ReadImage(filename)
numpyImage = sitk.GetArrayFromImage(itkimage)

numpyOrigin = np.array(list(reversed(itkimage.GetOrigin())))
numpySpacing = np.array(list(reversed(itkimage.GetSpacing())))

return numpyImage, numpyOrigin, numpySpacing


To be able to open and read the list of candidates, we need to use the csv python module. We define
now a function to: - Open a csv file - Read a csv file - Save each line of a csv file This functions takes as
input the name of the csv file and returns: - A list of each line of the csv
In [3]: def readCSV(filename):
lines = []
with open(filename, "rb") as f:
csvreader = csv.reader(f)
for line in csvreader:
lines.append(line)
return lines

1
Since the coordinates of the candidates are given in World Coordinates, we now need to transform
from world coordinates to voxel coordinates. We define now a function to do that. Please note that the
transformation below is only valid if there is no rotation component in the transformation matrix. For all
CT images in our dataset, there is no rotation component so that this formula can be used. This function
takes as inputs: - The world coordinates - The origin - The pixel Spacing This function returns: - Voxel
coordinates (voxelCoord)

In [4]: def worldToVoxelCoord(worldCoord, origin, spacing):

stretchedVoxelCoord = np.absolute(worldCoord - origin)


voxelCoord = stretchedVoxelCoord / spacing
return voxelCoord

We want to extract now some features from the candidates. We define some normalized planes to extract
views from the candidates

In [5]: def normalizePlanes(npzarray):

maxHU = 400.
minHU = -1000.

npzarray = (npzarray - minHU) / (maxHU - minHU)


npzarray[npzarray>1] = 1.
npzarray[npzarray<0] = 0.
return npzarray

After having defined these auxiliary functions, we can now define the main part of our script. First we: -
Specify the path where the image (img path) is - Specificy the path where the file with the list of candidates
is (cand path)

In [6]: img_path = data/1.3.6.1.4.1.14519.5.2.1.6279.6001.148447286464082095534651426689.mhd


cand_path = data/candidates.csv

Using the function defined in line 2 we can: - Load the image - Extract the Origin - Extract the Pixel
Spacing

In [7]: # load image


numpyImage, numpyOrigin, numpySpacing = load_itk_image(img_path)
print numpyImage.shape
print numpyOrigin
print numpySpacing

(220, 512, 512)


[-308.25 180. 145.300003]
[ 1.25 0.703125 0.703125]

Using the function defined in line 3 we can: - Load the csv file - Get the candidates Using the function
defined in line 4 we can: - Transform from world to voxel coordinates

In [8]: # load candidates


cands = readCSV(cand_path)
print cands
# get candidates
for cand in cands[1:]:
worldCoord = np.asarray([float(cand[3]),float(cand[2]),float(cand[1])])
voxelCoord = worldToVoxelCoord(worldCoord, numpyOrigin, numpySpacing)
voxelWidth = 65

2
[[seriesuid, coordX, coordY, coordZ, label], [1.3.6.1.4.1.14519.5.2.1.6279.6001.1484472864640

Using the function defined in line 5 we can: - Extract patch for each candidate in the list - Visualize each
patch - Save each page as image in .tiff format

In [ ]: for cand in cands[1:]:


worldCoord = np.asarray([float(cand[3]),float(cand[2]),float(cand[1])])
voxelCoord = worldToVoxelCoord(worldCoord, numpyOrigin, numpySpacing)
voxelWidth = 65
patch = numpyImage[voxelCoord[0],voxelCoord[1]-voxelWidth/2:voxelCoord[1]+voxelWidth/2,voxel
patch = normalizePlanes(patch)
print data
print worldCoord
print voxelCoord
print patch
outputDir = patches/
plt.imshow(patch, cmap=gray)
plt.show()
Image.fromarray(patch*255).convert(L).save(os.path.join(outputDir, patch_ + str(worldCoo

You might also like