DICOM and NIFTI

In this section, we're going to look at two Python libraries for interacting with common neuroimaging file formats pydicom and nibabel.

Use a virtual environment

To have full control over installing the packages described in this section, it may be beneficial to create a virtual environment.

DICOM

DICOM is perhaps the most popular imaging file format used within the medical imaging community. One of the more popular Python packages for reading and writing DICOM files is pydicom.

You'll need to install numpy for more advanced functionality.

pip install pydicom numpy

reading a file

To read a DICOM file into variable named ds, use dcmread()

import pydicom

ds = pydicom.dcmread('file.dcm')

Lazy loading

By default, pydicom.dcmread() will only read the DICOM file headers until you attempt to access the pixel data. This design choice saves time and resources.

reading headers

DICOM headers are identified by their group and element. Some example headers include

Name	group	element
Study Date	0008	0020
Study Time	0008	0030
Patient Name	0010	0010
Series Number	0020	0011
Accession Number	0008	0050
Instance ID	0020	0013
Study UID	0020	000D
Series UID	0020	000E

You can access these headers by indexing into the ds object using a tuple of (group, element). To access the Patient Name for example, you would use

element = ds[('0010', '0010')]

In the example above, the returned element will be an instance of pydicom.dataelem.DataElement. Use the value property to access the value of this element

patientname = element.value

reading pixel data

To read the raw DICOM pixel data as a series of bytes, you can access the PixelData property

pixels = ds.PixelData

To receive the pixel data in a more useful numpy array format, use the pixel_array property

pixels = ds.pixel_array

visualizing pixel data

Let's take a look at matplotlib to visualize DICOM pixel data. First, you need to install matplotlib

pip install matplotlib

Now, import matplotlib.pyplot and use imshow() to render the pixel data (a NumPy ndarray) as an image

import matplotlib.pyplot as plt

plt.imshow(ds.pixel_array)

plt.show()

NIfTI

For reading and writing NIFTI files, use the nibabel package

pip install nibabel

reading a file

To read a NIFTI file, use nibabel.load()

import nibabel

ds = nibabel.load('file.nii.gz')

reading headers

The NIFTI file headers are stored in a header property on the ds object. Since the header property behaves like a dictionary, you can use a for loop to iterate over all the headers

formatter = f'header = {header} value = {value}'

for header,value in ds.header.items():
    print(formatter)

reading pixel data

The .get_fdata() method will return the NIfTI pixel data as a NumPy array

pixels = ds.get_fdata()

Memory efficiency

The ds.get_fdata() method has many excellent features for handling large images efficiently. For more details, read the official documentation here.

visualizing pixel data

Let's take a look at matplotlib to visualize NIfTI pixel data. First, you need to install matplotlib

pip install matplotlib

Now, import matplotlib.pyplot and use imshow() to render image 50 along the z-axis

import matplotlib.pyplot as plt

fdata = ds.get_fdata()

plt.imshow(fdata[:,:,50])

plt.show()