5  NRRD - Intro

The NRRD (Nearly Raw Raster Data) file format is used to store and organize scientific and medical imaging data, commonly for 3D and 4D datasets. NRRD files are often used in applications like 3D Slicer for handling multi-dimensional imaging data. The format is flexible, supporting various data types, dimensions, and metadata.

5.1 Overview

5.1.1 Structure of NRRD File

  1. Header: Contains metadata about the image, specifying dimensions, data type, spacing, orientation, and other descriptive information. The header is usually in plain text, making it easy to read and edit.
  2. Data: Contains the raw imaging data (voxel values). The data can be embedded directly in the .nrrd file or stored in a separate file, referred to by the header.

5.1.2 Basic Structure of an NRRD Header

An NRRD header typically looks like this:

NRRD0004
type: float
dimension: 3
sizes: 256 256 128
encoding: gzip
spacings: 1.0 1.0 1.5
axis-mappings: 0 1 2
kinds: space space space
space origin: (0,0,0)
  • NRRD0004: Magic number indicating the NRRD format version.
  • type: Data type of the voxel values (e.g., float, short, int).
  • dimension: Number of dimensions (e.g., 2 for 2D, 3 for 3D).
  • sizes: Number of voxels along each dimension.
  • encoding: Specifies how the data is stored (e.g., raw, gzip for compression).
  • spacings: Physical spacing between voxels along each axis.
  • axis-mappings: Maps data dimensions to spatial dimensions.
  • kinds: Specifies the type of each axis (e.g., space for spatial dimensions).
  • space origin: Coordinates of the origin in physical space.

5.2 pynrrd

Code 
from pyhere import here
import nrrd

The pynrrd Python library is a popular tool for reading, writing, and manipulating NRRD files. Below are the steps to install pynrrd and use it to read NRRD files.

5.2.1 Reading an NRRD File

To read an NRRD file, use the nrrd.read() function, which returns two items:

  • data: A NumPy array containing the image data.
  • header: A dictionary containing the metadata from the NRRD header.

Here’s a basic example:

Code 
# Read the NRRD file
data, header = nrrd.read(here("data/MRI-Brain-Eye/Case1/3 t1 axial Processed_CaPTk.nrrd"))

# Display the shape of the data
print("Data shape:", data.shape)

# Display the header information
print("Header metadata:", header)
Data shape: (192, 256, 192)
Header metadata: OrderedDict([('type', 'unsigned short'), ('dimension', 3), ('space', 'left-posterior-superior'), ('sizes', array([192, 256, 192])), ('space directions', array([[ 0.97656101, -0.00170442,  0.        ],
       [ 0.00170442,  0.97656101,  0.        ],
       [ 0.        ,  0.        ,  1.        ]])), ('kinds', ['domain', 'domain', 'domain']), ('endian', 'little'), ('encoding', 'gzip'), ('space origin', array([ -89.87082095, -145.95727884,  -78.06524467]))])

  • data: This will be a multi-dimensional NumPy array, with dimensions specified by header['sizes']. You can directly manipulate this array or use it for visualization.

  • header: A dictionary containing the metadata extracted from the NRRD header, including keys like type, dimension, spacings, and space origin.

5.2.2 Accessing Specific Metadata

You can access specific metadata fields from the header dictionary:

Code 
# Access voxel spacing
spacing = header.get('spacings', None)
print("Voxel Spacing:", spacing)

# Access data type
data_type = header.get('type', None)
print("Data Type:", data_type)
Voxel Spacing: None
Data Type: unsigned short

5.2.3 Visualizing a Slice of the NRRD Data

To visualize a slice of the 3D NRRD data (for example, slice 64 along the z-axis), you can use matplotlib:

Code 
import matplotlib.pyplot as plt

# Select a slice from the 3D data
slice_data = data[:, :, 64]

# Display the slice
plt.imshow(slice_data, cmap='gray')
plt.title("Slice 64 of the NRRD data")
plt.axis('off')
plt.show()