N-dimensional images

Biomedical Image Analysis in Python

Stephen Bailey

Instructor

Images of all shapes and sizes

im[row, col]

foot_x-ray

Images of all shapes and sizes

vol[pln, row, col]

body_gif

Images of all shapes and sizes

im[row, col, ch]

cell_rgb

Images of all shapes and sizes

im_ts[time, row, col, ch]

cell_video

N-dimensional images are stacks of arrays

stack-of-images-shifted-slightly

import imageio
import numpy as np

im1=imageio.imread('chest-000.dcm')
im2=imageio.imread('chest-001.dcm')
im3=imageio.imread('chest-002.dcm')

im1.shape

(512, 512)

vol = np.stack([im1, im2, im3])

vol.shape

(3, 512, 512)

Loading volumes directly

imageio.volread():

read multi-dimensional data directly
assemble a volume from multiple images

import os
os.listdir('chest-data')

['chest-000.dcm', 
 'chest-001.dcm', 
 'chest-002.dcm',
 ..., 
 'chest-049.dcm']

import imageio
vol = imageio.volread('chest-data')

vol.shape

(50, 512, 512)

Shape, sampling, and field of view

Image shape: number of elements along each axis

import imageio
vol = imageio.volread(
            'chest-data')
# Image shape (in voxels)
n0, n1, n2 = vol.shape
n0, n1, n2

(50, 512, 512)

Field of view: physical space covered along each axis

Sampling rate: physical space covered by each element

# Sampling rate (in mm)
d0, d1, d2 = vol.meta['sampling']
d0, d1, d2

(2, 0.5, 0.5)

# Field of view (in mm)
n0 * d0, n1 * d1, n2 * d2

(100, 256, 256)

Let's practice!

Biomedical Image Analysis in Python