A Method of Computational Image Analysis for Predicting Tissue Infarction After Acute Ischemic Stroke
Better predict brain tissue infarctionProvide new information to guide treatment
Clinical decision supportPredict brain tissue infarction in acute ischemic stroke patients
Researchers at UCLA have designed a specialized deep convolutional neural network (CNN) image analysis algorithm that automatically learns hierarchical spatio-temporal features, which are more predictive than traditional parameters, such as CBV.
State Of Development Algorithm has been developed and successfully tested. Background Ischemic stroke occurs when a blood vessel supplying blood to the brain is blocked by a blood clot or plaque fragment. Sudden loss of blood circulation to an area of the brain results in a corresponding loss of neurologic function. Emergent and accurate brain imaging is essential for excluding hemorrhage, differentiating between irreversibly and reversibly affected brain tissue (dead tissue vs. tissue at risk), identifying stenosis or occlusion of major extra- and intracranial arteries, and allowing time-critical decision-making on selection of patients appropriate for thrombolytic therapy. Current techniques apply single value decomposition (SVD) to deconvolve pre-therapy magnetic resonance (MR) perfusion-weighted images (pre-PWIs). Parameters, such as time-to-maximum (Tmax) and cerebral blood volume (CBV), generated from the deconvolution process are used for tissue at risk (penumbra) and infarct prediction. However, the threshold of these imaging parameters for detecting infarct is still under debate, and there are growing concerns that the parameters obtained from the plot generated via deconvolution are less predictive due to distortions introduced during the deconvolution process. Additional Technologies by these Inventors Tech ID/UC Case 27226/2015-955-0 Related Cases 2015-955-0
USA
