Semi-Automatic quantification of demyelination load in children with metachromatic leukodystrophy on MR-images

Aims: Metachromatic Leukodystrophy (MLD) is a rare lysosomal storage disease, which leads to progressive demyelination. This process can be sensitively detected and visualized by MRI. The volume of demyelination can be measured directly using manual segmentation, but this is quite time-consuming and labour-intensive.

Aims: The aim of this work was therefore the development and validation of a semi-automatic method for determining the „demyelination load“ in MLD on routinely used MR-images.

Methods: The software toolbox consists of an automatic multi-spectral preprocessing of different sequences followed by steps of interaction allowing the user to choose and modify suggested segmentations. Algorithms of iterative intensity-thresholding and cluster-segmentation are implemented. Validation of the method was performed through a comparison against manually segmented results taken from our cohort of approximately 70 children with MLD; validation results from a subgroup of 10 children are available. The target value consisted of the overlap of the segmented demyelination load.

Results: Semi-automated segmentation was found to be less time-consuming than manual segmentation (12±3min vs. 26±9min, p<0.01). The agreement between the semi-automatically and manually segmented „demyelination load„ was very high (dice coefficient 0,80).

Summary: The developed method for quantifying the „demyelination load“ in MLD was shown to be precise and more time-efficient than manual segmentation. It can be applied to MR-images of different technical background. This is especially relevant to evaluate new therapeutic strategies.