Motivation

Deadline has passed. No further submissions will be considered.

It has been estimated that annually 1.4 and 2.5 million traumatic brain injury (TBI) incidents occur alone in the US and Europe, respectively. With respect to fatality and long term disability, TBI is one of the most severe types of injury. Mainly caused by traffic accidents and falls, it can affect everyone regardless of their gender, age and ethnicity. This epidemic characteristic makes TBI to a leading cause of death and disability across demographics [1234].


Clinical Motivation

Predicting the outcome of patients suffering from TBI could both facilitate clinical decision making and support development of new therapeutic concepts. However, prognostics are often based on clinical or cognitive symptoms, which might be biased by personal perception. Therefore, there is an urge for more objective methods, that detect radiological evidence after TBI. However, the strong heterogeneity of the injury pattern and the complex change of pathology over time pose a persistent challenge. This especially holds true for mild TBI (mTBI), where lesions are non-prevalent and conventional MRI often appears normal, but injury can cause post-concussional symptoms and neurocognitive dysfunction.


Computational Motivation

Despite its valuable potential for outcome prediction, MRI based features which describe mTBI pathology are not fully understood. Diffusion MRI could be the key to reveal biological factors of strong predictive power. Indeed, common techniques such as local region or whole brain analysis are highly dependent on adequate image registration and brain parcellation and can be restricted by the high dimensionality of the data.

This challenge asks for methods that focus on finding differences between healthy subjects and TBI patients and sort the given data in distinct categories in an unsupervised manner.

 

Advertisements