Skoči na glavni sadržaj

Izvorni znanstveni članak

Relationship of Plasma Creatine Kinase and Cardiovascular Function in Myocardial Infarction

Nikola Stambuk ; Rudjer Bošković Institute, Bijenička 54, 10000 Zagreb, Croatia
Paško Konjevoda ; Rudjer Bošković Institute, Bijenička 54, 10000 Zagreb, Croatia

Puni tekst: engleski pdf 70 Kb

str. 891-898

preuzimanja: 213



Maximal plasma creatine kinase (CPK) values are often used as an enzymatic test in acute myocardial infarction. The cardiovascular functional ability (FA) is a reliable parameter, which may define the lesion of the cardiovascular system following myocardial infarction. We have applied the Cubist machine learning tool to set up a model that defines the residual cardiovascular functional ability after a myocardial infarction by means of the maximal CPK in the acute phase of the disease. Cubist models numeric data and generates values by means of a collection of rules associated with the linear expression for computing target values. Based on the literature data set of Mirić et al.,1 we have derived and tested a reliable and accurate Cubist model for acute inferior and anteroseptal myocardial infarction consisting of three simple rules. The rules enable simple prediction of the expected cardiovascular functional ability in myocardial infarction following recovery. The model is based on two clinical parameters related to the disease, maximal CPK in the acute phase of the myocardial lesion, and anteroseptal or inferior cardial localisation of the infarction. Strong correlation (r anteroseptal = 1.00, r inferior = 0.94, n = 56), insignificant differences of real and predicted values, and low average error of the leave-one-out test (anteroseptal 1.72, inferior 3.54) confirm the accuracy of the method and its applicability in clinical medicine. In addition to the prognostic and diagnostic application, the extracted rules enable more efficient evaluation of new drugs and therapeutic procedures in Cardiology.

Ključne riječi

CPK, myocardial infarction, localisation, cardiovascular function, Cubist, machine learning, model

Hrčak ID:



Posjeta: 379 *