The majority of images available below are simple test images that can be used in the initial stages of testing segmentation methods. An MRI simulator and simulated brain data sets are available from the McConnell Brain Imaging Centre at the Montreal Neurological Institute and Hospital: BrainWeb: Simulated Brain Database
The following consists of 4 increasingly complex shapes with 3 different levels of signal-to-noise and 3 different levels of contrast-to-noise: phantom slice "2" is circular; "4" has 6 low-amplitude lobes; "6" has 6 high-amplitude lobes; and "8" has 6 high-amplitude, non-symmetric lobes. Each phantom is an individual image, so there are 4 * (3+3) = 24 images (256 pixels * 256 pixels * 2 bytes = 131072 bytes/image, 24*131072= 3,072 Kbytes = 3 Mb). Download phantom slices (.tar.Z, 1.3 Mb).
This image shows three circles with varying signal-to-noise:
This shows the three circles with varying contrast-to-noise:
This shows varying shape complexity (circle, low-amplitude and
high-amplitude "lobes"):
Original scan: 4_8_*.img (60 slices, 7.5 Mb)
Simulated scan with gradient: 4_11_*.img (55 slices, 6.9 Mb)
Simulated scan with gradient and noise: 4_12_*.img (55 slices, 6.9 Mb)
Click for a description of how these gradients were generated.
Download phantom brains (.tar.Z 2.9 Mb). Each image slice is 256x256 by 16-bit.
This shows the original scan, phantom with gradient, and the phantom
with gradient and noise: 
Each of these 3D data sets is an object that is a roughly spherical "gray matter" core surrounded by white matter. (52 slices, 6.5 Mb each data set, 32.5 Mb total)
Sphere1: phantom with noise, no gradient. Download sphere1 (.tar.Z 2 Mb).
Sphere2: phantom with noise, small gradient (few overlapping intensities). Download sphere2 (.tar.Z 2.1 Mb).
Sphere3: phantom with noise, larger gradient (intensities overlap by about half). Download sphere3 (.tar.Z 2.3 Mb).
Sphere4: phantom with noise, large gradient (intensities overlap completely). Download sphere4 (.tar.Z 2.4 Mb).
Sphere5: phantom, no noise, same gradient as sphere2. Download sphere5
(.tar.Z 62 Kb).