A New Traumatic Axonal Injury Classification Scheme based on Clinical and Improved MR Imaging Biomarkers


The global aim is to develop a novel classification for Traumatic Axonal Injury (TAI) using data from multimodal MR Imaging and to determine its clinical value for characterization of injury severity and prediction of outcome. Methods of deep learning techniques will be developed and used for lesion mapping. The joint venture will utilize MRI dataset from CENTER-TBI and two local studies (, 1400 patients), hence be the largest MRI study. All studies have comprehensive collection of acute phase variables and outcome. Several training sets will be used for model selection, and methods for automated image analyses will be refined and validated. Finally, the model for a possible novel MRI classification will be tested in a large data set. With the development an improved MRI-based classification scheme of TAI lesions, we aim to provide a better assessment of injury severity in the acute phase, and to better predict long-term outcome. Hence, we will provide a timely new tool for neuroradiologists, clinicians and researchers for more precise TBI diagnosis allowing for tailored and cost-efficient treatment and rehabilitation. Furthermore, this project will bring the field forward by increasing our understanding of the pathophysiology in TBI, the link between level of consciousness and injury mechanisms, type and location, and long-term outcome. The recent advancement in automated image analyses provide unique opportunities for a classification with great clinical applicability.


Computational neurosciences, Imaging techniques, Clinical trial, "omics" approaches, Traumatic axonal injury, diffuse axonal injury, classification systems, injury severity, prognostication, automated MRI methods, deep learning algorithms

Call topic

External Insults

Proposed runtime

2017 - 2021

Project team

Anne Vik (Coordinator)
Norway (RCN)
Patrick Cras
Belgium (FWO)
Bram van Ginneken
The Netherlands (NWO)
David Menon
United Kingdom (MRC)