Jan Rexilius MS, PRESENTER: Nothing to Disclose

Abstract: HTML Purpose: To develop a semi-automatic method for accurate and reproducible MR based volumetry of white matter lesions. Accurate lesion volumetry is crucial, e.g., for the quantitative clinical evaluation of brain MR scans in Multiple Sclerosis (MS) patients. Methods and Materials: We propose a volumetry method that is applicable to any image offering contrast between lesion and surrounding white matter tissue; typically, T2, PD, or Gd-T1 with a slice thickness of 1-5 mm are used. The method consists of three steps. First, a cuboid ROI is selected from a MR volume and resampled in z-direction to an isotropic voxel size using a Lanczos 3-lobed filter. Second, an Interactive Watershed Transform (IWT) serves to separate the lesions from other hyper- or hypointense structures. The IWT results in an over-inclusive lesion segmentation and is entirely driven by include and exclude markers in real-time. Third, an automatic regional histogram analysis is used to calculate lesion volumes. The analysis is based on a bimodal Gaussian model for lesion and white matter tissue as well as an explicit model for partial volume effects (PVE). The method is evaluated on 66 realistic phantom datasets with known ground truth; lesion phantoms are based on a multispectral MR image of a healthy volunteer: Siemens Magnetom Vision 1.5T, coronal, T2/PD/T1-weighted, matrix 512x512, slice thickness 3.0mm, 40 slices. Realistic hyper- and hypointense lesions (black holes) of different shape and size (0.05, 0.1, 0.2, 0.4, 0.7, and 1.0 ml) were embedded into this image at a given intensity and noise level. Results: All datasets have been successfully analyzed by the new method. On average, 5 markers were required (range: 3-10) resulting in a mean total analysis time of 1:12min (range: 37sec-1:51min). Accuracy analysis was performed in comparison to the known ground truth resulting in a mean overestimation of 8.5% (SD 11.6%) for small lesions (<0.3 ml) and -0.03% (SD 2.2%) for intermediate lesions (>0.3 ml). In comparison, mean overestimation for manual evaluation of three experts are 111.3% (SD 52.1%) and 56.9% (SD 19.6%), respectively. Conclusion: Our new semi-automatic method permits a highly accurate and reproducible volumetry of white matter lesions. The method accounts for PVE that represent the main problem in lesion volumetry. Even if a slight overestimation is observed for small lesions, results are clearly superior to manual tracing. The presented method is likely to become important for early diagnosis and decision support in white matter diseases, particularly in MS.       Questions about this event email: hahn@mevis.de

Rexilius MS, J, New Robust Semi-Automatic Method for Accurate Volumetry of White Matter Lesions.  Radiological Society of North America 2003 Scientific Assembly and Annual Meeting, November 30 - December 5, 2003 ,Chicago IL. http://archive.rsna.org/2003/3107692.html