A Semi-Automatic Algorithm for Determining the Demyelination Load in Metachromatic Leukodystrophy

Abstract

Metachromatic leukodystrophy is a lysosomal storage disorder leading to progressive demyelination of brain white matter. This is sensitively detected using magnetic resonance imaging. The volume of demyelination, the "demyelination load," could serve as a useful parameter for assessing both the natural course of the disease and treatment effects. The aim of this study was to develop and validate a semiautomated approach for determining the demyelination load to achieve reliable and time-efficient segmentation results. The demyelination load was determined in 77 magnetic resonance imaging data sets from 35 patients both manually and semiautomatically. For manual segmentation, regarded as the gold standard, the software ITK-Snap was used. For semiautomatic segmentation, a new algorithm called Clusterize was developed and implemented in MATLAB, consisting of automatic iterative region growing followed by the interactive selection of clusters. Results were compared in terms of the obtained volumes, spatial overlap, and time taken to conduct the segmentation. Performance of the semiautomatic algorithm was excellent, with the volumes generated by the new algorithm showing good agreement with the ones generated by the gold standard (93.4 ± 45.5 vs 96.1 ± 49.0 mL, P = NS) with high spatial overlap (Dice's similarity coefficient = 0.7861 ± 0.0697). The semiautomatic algorithm was significantly faster than the gold standard (8.2 vs 27.0 min, P < .001). Intrarater and interrater reliability determined high reproducibility of the method. The demyelination load in metachromatic leukodystrophy can be determined in a time-efficient manner using a semiautomatic algorithm, showing high agreement with the current gold standard.