Semi-automated thresholding technique for measuring lesion volumes in multiple sclerosis: Effects of the change of the threshold on the computed lesion loads

M. Filippi, M. Rovaris, A. Campi, C. Pereira, G. Comi

Research output: Contribution to journalArticle


Quantitative evaluation of lesion load in multiple sclerosis (MS) from magnetic resonance imaging (MRI) scans is becoming important in understanding and monitoring the progression of the disease. Methods of MS lesion segmentation based on intensity thresholding offer one of the most robust and easily-implemented means of computing the total lesion volume. This study evaluated the effects of slight changes in the choice of intensity threshold on computed lesion volumes in 20 patients with MS using such a technique. After judging the optimum choice of threshold value, the threshold value was increased and decreased by 3% in 1% steps around this value; we observed a mean change of 15% in computed lesion volumes for 1% changes of threshold value. Larger changes in lesion volume were found when the threshold was changed by larger amounts. On the other hand, the amount of time required for manual review decreased, and the confidence with which manual review could be performed increased when using lower thresholds. This study shows that the choice of threshold is a crucial factor in measuring lesion volumes in MS when using intensity-based techniques. It also suggests that in multicenter and/or longitudinal studies, criteria for choosing the threshold should be developed whereby the threshold level should be set such that all MR visible lesions are above it, in order to minimise the human interaction and, consequently, the reproducibility of the results.

Original languageEnglish
Pages (from-to)30-34
Number of pages5
JournalActa Neurologica Scandinavica
Issue number1
Publication statusPublished - 1996



  • Brain magnetic resonance imaging
  • Lesion volume
  • Multiple sclerosis
  • Semi-automated thresholding technique

ASJC Scopus subject areas

  • Clinical Neurology
  • Neuroscience(all)

Cite this