Stroke is one of the main death causes worldwide. Considering all causes of death in Brazil, stroke was responsible for 10.2% in the year 2009. There are two types of stroke, ischemic, which corresponds to 87% of cases and hemorrhagic, 13% of cases. The early diagnosis avoids irreversible cerebral damage. High-resolution computed tomography (HRCT) and nuclear magnetic resonance imaging (NMR) are the two main techniques of diagnostic imaging used to detect stroke. NMR is a technique that provides better detection of hypodense areas, but is expensive and time consuming. HRCT has lower cost and greater accessibility for the population, so it is still the main method to diagnose stroke. In most cases, the determination of the ischemic area is carried out subjectively through the evaluation of a radiologist or neuroradiologist. This research proposes the implementation of a system for image segmentation, with enhancement of ischemic stroke areas and region quantification in retrospective examinations. Different image processing methods will be used for image enhancement and detection of morphological patterns such as Wavelets filtration and fuzzy C-means clustering. The great advantage of this method is to combine different image processing techniques and optimize them for better results to quantify ischemic stroke areas. We also propose a novel correlation between the objective and subjective analysis performed by the radiologist, and the comparison between the method of HRCT with MRI and CT Perfusion in later periods at the same patient exam. We also intend to develop a computer-aided diagnosis software to assist the radiologist in detecting acute stroke regions in HRCT images. All these factors contribute to the diagnosis of ischemic stroke in HRCT images through accurate and reliable methods.
News published in Agência FAPESP Newsletter about the scholarship: