N.M.Emanuel Institute of Biochemical Physics of Russian Academy of Sciences, Russia, 117997, Moscow, Kosygina, 4
Department of Pirogov Russian National Research Medical University “Clinical and Research Center of Gerontology” of Russian Ministry of Health, Moscow, 1st Leonova St., 16, 129226
Institution of Russian Academy of Sciences Dorodnicyn Computing Centre of RAS, Russia, 119991, Moscow, Vavilova, 40

azfor@narod.ru
hla2222@yandex.ru 
senkoov@mail.ru

Abstract.The paper presents results of studies, aimed to asses relation between a set of clinical and laboratory indices and risk of transient ischemic attacks (TIA) and ischemic stroke in patients from elder age groups with chronic cerebral ischemia (CCI). The studies were based on pattern recognition techniques and data mining method implementing optimal statistically valid partitioning of feature space. The risk factors of transient ischemic attack were found: lipidic spectrum, blood cell counts, biochemical indices, type of hyperlipidemia, II-III stages of CCI, left ventricular hypertrophy by electrocardiogram, ultrasonic scanning features of thyroid diffusive changes. It was found that exactness of TIA prediction with the help of pattern recognition techniques achieves 81.5% if all features forming basic regularities are used. In patients that previously suffered from TIA set of risk factors for further ischemic stroke includes: CHD with ciliary arrhythmia, breaks of rhythm, hyperglycemia. Forecasting ability of pattern recognition technique is 77%.

Key words: risk forecasting factors, transitory ischemic attack, ischemic stroke, lipid metabolism, intellectual data analysis, data mining, recognition.