Author/Editor     Grošelj, C; Kukar, M; Fettich, JJ; Kononenko, I
Title     Impact of machine learning to the diagnostic certainty of the patient's group with low coronary artery disease probability
Type     članek
Source     In: Kononenko I, Urbančič T, editors. CADAM-97. Zbornik Računalniška analiza medicinskih podatkov; 1997 nov 12; Bled. Ljubljana: Inštitut Jožef Stefan,
Publication year     1997
Volume     str. 68-74
Language     eng
Abstract     The stepwise diagnostic process of coronary artery disease (CAD) consists of evaluation of symptoms and signs of the disease and ECG diagnosed at rest, ECG during exercise (exercise ECG), myocardial perfusion scintigraphy (MPS), and coronary angiography. In classical diagnostic decision many diagnostic data can't be included ininterpretation of the test results. Machine Learning (ML) can use all particular data, so better test would be expected using this method. Our goal was to predict in a group of 327 patients the results of coronary angiography obtained by ML method and compare them with the results of exercise ECG and MPS. The Naive Bayesian Classifier as one of the ML methods was applied. The patients, who at any step of diagnostic process, reach the post-test probability of 0.10 or less, are considered as free of disease and do not need any further diagnostic examination. In 18 patients (5.5%) at exercise ECG and in 17 patients (5.2%) at MPS the post-test probability changed to value of under 0.10. In all of these patients results of coronary angiography were negative.
Descriptors     CORONARY DISEASE
ELECTROCARDIOGRAPHY
HEART
CORONARY ANGIOGRAPHY
ARTIFICIAL INTELLIGENCE
DECISION MAKING, COMPUTER-ASSISTED