Publikation

New Multiple Sclerosis Lesion Detection with Convolutional Neural Registration Networks

Julia Andresen; Hristina Uzunova; Jan Ehrhardt; Heinz Handels

In: Olivier Commowick; Frédéric Cervenansky; François Cotton; Michel Dojat (Hrsg.). MICCAI2021 - MSSEG-2 challenge proceedings: Multiple sclerosis new lesions segmentation challenge using a data management and processing infrastructure. MSSEG-2 challenge: Multiple sclerosis new lesions segmentation challenge, located at MICCAI 2021, September 23, Pages 111-111, HAL, 2021.

Zusammenfassung

The delineation of new multiple sclerosis lesions is an important yet difficult and error-prone task. We present an automatic framework for new lesion segmentation composed of two convolutional neural networks (CNNs). The first CNN registers images from different timepoints and simultaneously segments non-corresponding regions. The second CNN refines the non-correspondence maps into segmentations of new lesions. Segmentation performance is shown to profit from the registration results of the first network.

Weitere Links

Deutsches Forschungszentrum für Künstliche Intelligenz
German Research Center for Artificial Intelligence