Background. The aim of this study was to compare the precision and sensitivity of ultrasound picture of carotid plaque in relation to two computer programs: a GSM histogram analysis and a detailed pixel distribution analysis.
Material and methods. The material consisted of 130 carotid plaques acquired from patients treated and operated in the Department of Vascular Surgery and Angiology in the period 2002-2003. The analysis was performed with the use of two computer programs: the widely applied assessment of grey scale median (GSM) and detailed pixel distribution analysis. The plaques were evaluated intraoperatively (visual assessment) and subjected to comparative histopatological examination. The microembolisation (TCD) was monitored in each patient. All possible perioperative complications were monitored and documented with a CT or NMR examination. In statistics, a multivariate statistical analysis was applied, supplemented by a logit regression model. The propriety of the model was verified with the help of the Hosmer-Lemeshow test. Its predictive value was checked using an ROC curve. The presence of linear trends was assessed with the help of linear regression models.
Results. Analysis of the components of risk of the appearance of events showed a decisive superiority of pixel analysis with the use of Image-Pro program over the hitherto applied technique based on analysis of grey scale median (GSM).
Conclusions. Computer analysis of ultrasound picture remains a valuable tool in the preoperative assessment of carotid plaque character and in foreseeing the degree of perioperative risk. Multivariate statistical analysis proved to be superior to the detailed pixel distribution analysis program in the computer assessment of ultrasound carotid plaque picture. Plaques, which, in the detailed pixel distribution, contain more than 72% of tissue with features of blood, fat and muscle in their surface, augment the risk of perioperative complications.