Neural network measures MS disease progression

May 1, 2006

The percentage of brain volume is an important marker of disease progression in multiple sclerosis. Researchers have developed a prototype neural network-based quantification system to measure this important benchmark by computer-assisted segmentation of multispectral MR imaging data.

The percentage of brain volume is an important marker of disease progression in multiple sclerosis. Researchers have developed a prototype neural network-based quantification system to measure this important benchmark by computer-assisted segmentation of multispectral MR imaging data.

Dr. Axel Wismueller of Ludwig Maximilians University in Munich performed MR exams in six women with relapsing-remitting MS. The neural network computed the percentage of brain volume by automatic cerebrospinal fluid segmentation. The voxel-specific gray-level intensity spectrum forms a seven-dimensional feature vector, which is classified by the neural network as either belonging to CSF or not. Findings were reported at the 2005 European Congress of Radiology.

The neural network-based computation significantly outperformed the conventional angle-image method. Specifically, the neural network performed better by retrieving only T2-weighted and perfusion/diffusion-weighted signals, thereby avoiding misclassifications in white matter lesions that are difficult to distinguish from CSF.