A flexible machine learning image analysis system for high-precision computer-assisted segmentation of multispectral MRI data sets in patients with multiple sclerosis

Automatic brain segmentation is an issue of specific clinical relevance in both diagnosis and therapy control of patients with demyelinating diseases such as multiple sclerosis (MS). We present a complete system for high-precision computer-assisted image analysis of multispectral MRI data based on a flexible machine learning approach. Careful quality evaluation shows that the system outperforms conventional threshold-based techniques w.r.t. inter-observer agreement levels for the quantification of relevant clinical parameters, such as white matter lesion load and brain parenchyma volume