Med Phys 2002 Jul;29(7):1536-46
He R, Narayana PA.
Department of Radiology, University of Texas at Houston Medical School, 77030, USA.
A method for automatic identification and delineation of contrast-enhanced multiple sclerosis (MS) lesions on brain magnetic resonance images is described.
This method relies on adaptive local segmentation derived from the morphological "open" and "reconstruction" operations on gray scale images for identification of both lesion and nonlesion enhancements.
Nonlesion enhancements from vasculature and extrameningeal tissues are identified by exploiting their topologic relationship to the brain mask.
Enhancing structures without a blood-brain-barrier, such as choroid plexus, are identified and eliminated by spatially mapping the locations of the MS lesions visualized on dual echo images onto the post-contrast images.
Delineation of enhancements is realized using fuzzy connectivity.
Both the detection and delineation results are validated using statistical methods.