Regresar

Spontaneous intracerebral hemorrhage image analysis methods: A survey

Abstract:

Spontaneous intracerebral hemorrhages (ICH) account for 10-30% all strokes and are a result of acute bleeding into the brain by rupturing of small penetrating arteries. The societal impact of hemorrhage strokes are magnified by the fact that affected patients typically are a decade younger than those afflicted with ischemic strokes. The ICH continue to kill or disable most of their victims some studies show that those who suffer ICH have a 30-day mortality rate of 35-44% and a 6-month mortality rate approaching 50%. Diagnosis of ICH is based largely on clinical history and corroborative Computer Tomography (CT) scanning of the brain. The heat CT scan has a sensitivity and specificity that approach 100% for acute ICH. The hemorrhage volume is the most important pbkp_redictor of clinical outcome after ICH and it can be approximated rapidly with a head CT. Contrast-enhanced CT scan that may now be readily accomplished on the latest-generation scanners. These images can exclude most gross vascular and tumor causes of hemorrhage rapidly and can have an impact on the therapeutic plan. We survey several available medical image analysis methods, which have been used in CAD systems for segmentation and tracking of ICH. These methods including diverse algorithms and techniques such as: MRI based techniques: susceptibility-weighted imaging (SWI), gradient-recalled echo imaging (GREI) and GRE-type single-shot echo-planar imaging (GRE-EPI), Artificial neural networks training based on the electrical impedance tomography signals, Statistical techniques as frequency histograms and k-means clustering, Labeling approaches based on the combination of maximum a-posteriori (MAP) estimation and Markov random fields (MRF) models, Quantitative measure of side to side of cerebral blood flow (CBF) asymmetry algorithm, Volume region extraction based on digital atlas, Hybrid approaches including the suitable combination of two or more methods such as Unsupervised fuzzy clustering and expert system-based labeling, Mathematical morphology and histogram based intensity analysis, Deformable models and similarity measures Spontaneous ICH segmentation, at present, is not a solved problem. Future work will be focused on the development of better automatic segmentation and tracking methods to gain in accuracy and precision in the ICH volume determination.