New technique to identify and predict stroke lesions

Strokes cause the closure of vessels leading to a lack of blood supply in the brain. Therefore, some areas are not provided sufficiently with oxygen which causes damage. Those areas, called lesions, are not fully functional anymore and require treatment. Currently, lesions are still hard to locate by only using CT imaging. "Lesions might evolve over time by growing and shrinking," commented Robin Maack, doctoral student in the Computer Graphics and Human Computer Interaction group of Professor Dr Hans Hagen at Technische Universität Kaiserslautern. "The final shape of the lesion is of great importance for the treatment and rehabilitation of a patient." Together with Dr Christina Gillmann from Leipzig University, Maack is working on a new method to quickly identify lesions, including the prediction of their future shape.

To achieve their goal, they use a neural network. "This is a technique of Artificial Intelligence using data to learn about lesion properties in CT images," explains Maack. "Those special algorithms implement the clinicians' experience , drawing the lesion shapes into existing images and thus feeding the network with information." With this procedure, Maack and Gillmann are able to analyse new CT images without the need of direct assistance by a clinician.
This technique is a valuable addition to current stroke treatment to speed up treatment plan creation. The team will be presenting their work at the trade fair.

MEDICA-tradefair.com; Source: Technische Universität Kaiserslautern