Algorithm was tested on real MRI data gained of anonymous
patients acquired in clinical practice. Magnetic resonance
images came from various apparatus and were
scanned with various settings, so they have different intensities.
The images for our dataset were selected from 3D MRI
data witch were scanned with T1 relaxation. For the evaluation
of our method, we have used 150 randomly selected
2D images with various measurements and intensities,
which include tumors of different areas, shapes and
locations.
Tumor segmentation was tested by two ways to detect
advantages and disadvantages of proposed algorithm.
First, it was tested with the algorithm which consists only
of adaptive greyscale morphological reconstruction and
adaptive thresholding without the graph cut algorithm.
Second, it was tested also with the graph cut algorithm.
Algorithm results were compared with manual segmentations
of tumors provided by experts. Verification was
based on the per pixel comparison of the segmentation results
and manual segmentations. Resulting segmentation
was transformed to binary image. It is because manual
segmentation was also saved as binary image.