The most commonly used approach for identifying areas of hypoperfusion on perfusion magnetic resonance images is visualization and manual delineation by the clinician. This approach is both subjective and time-consuming, thus unsuitable for both clinical decision-making based on perfusion-diffusion mismatch and analysis of large datasets or multi-center studies. In the following an innovative semi-automated method is suggested to assess the likelihood that an ischemia in brain results in an infarction by image analysis.
The inventive method is based on an automated search for voxels on perfusion maps whose intensity values exceed the predefined quantitative thresholds and that are connected to each other. These connected objects are then sorted based on the following criteria: anatomical location (especially shared vascular structures), the proximity to previously identified infarcted tissue and their size. Finally a probability is assigned to the identified objects. The top 3 closest and the top 3 largest objects form the “high-probability” region. All other objects within the same vascular territory as the acute ischemic region build the “medium-probability” area. All objects in the vascular territories not associated with the acute ischemic region (i.e. excluding those defined in both preceding paragraphs) represent the “low-probability” section. The generated results are shown in an overlaid image – the likelihood mask.
The patent protected technology is available for in-licensing. Further clinical validation can be realized in cooperation with an experienced university clinic in Germany.
The technology has been clinically evaluated based on 21 patient data sets.
A priority claiming European patent application has been filed in July 2017. Additional international patent filings are possible within the priority year.
“Relationship Between Changes in the Temporal Dynamics of the Blood-Oxygen-Level-Dependent Signal and Hypoperfusion in Acute Ischemic Stroke.” Ahmed A. Khalil, Ann-Christin Ostwaldt, Till Nierhaus, Ramanan Ganeshan, Heinrich J. Audebert, Kersten Villringer, Arno Villringer, Jochen B. Fiebach; Stroke, 2017, 48:925-93