All About Multiple Sclerosis

More MS news articles for February 2003

Brain MRI lesion load quantification in multiple sclerosis: a comparison between automated multispectral and semi-automated thresholding computer-assisted techniques

Magn Reson Imaging 2002 Dec;20(10):713-20
Achiron A, Gicquel S, Miron S, Faibel M.
Multiple Sclerosis Center, Sheba Medical Center, Tel-Hashomer, Israel

Brain magnetic resonance imaging (MRI) lesion volume measurement is an advantageous tool for assessing disease burden in multiple sclerosis (MS).

We have evaluated two computer-assisted techniques: MSA multispectral automatic technique that is based on bayesian classification of brain tissue and NIH image analysis technique that is based on local (lesion by lesion) thresholding, to establish reliability and repeatability values for each technique.

Brain MRIs were obtained for 30 clinically definite relapsing-remitting MS patients using a 2.0 Tesla MR scanner with contiguous, 3 mm thick axial, T1, T2 and PD weighted modalities.

Digital (Dicom 3) images were analyzed independently by three observers; each analyzed the images twice, using the two different techniques (Total 360 analyses).

Accuracy of lesion load measurements using phantom images of known volumes showed significantly better results for the MSA multispectral technique (p < 0.001).

The mean intra-and inter-observer variances were, respectively, 0.04 +/- 0.4 (range 0.04-0.13), and 0.09 +/- 0.6 (range 0.01-0.26) for the multispectral MSA analysis technique, 0.24 +/- 2.27 (range 0.23-0.72) and 0.33 +/- 3.8 (range 0.47-1.36) for the NIH threshold technique.

These data show that the MSA multispectral technique is significantly more accurate in lesion volume measurements, with better results of within and between observers' assessments, and the lesion load measurements are not influenced by increased disease burden.

Measurements by the MSA multispectral technique were also faster and decreased analysis time by 43%.

The MSA multispectral technique is a promising tool for evaluating MS patients.

Non-biased recognition and delineation algorithms enable high accuracy, low intra-and inter-observer variances and fast assessment of MS related lesion load.