W. Dassen et al., DEVELOPMENT OF ECG CRITERIA TO DIAGNOSE THE NUMBER OF NARROWED CORONARY-ARTERIES IN REST ANGINA USING NEW SELF-LEARNING TECHNIQUES, Journal of electrocardiology, 27, 1994, pp. 156-160
Recently, an evaluation of the value of the resting electrocardiogram
recorded during chest pain for identifying high-risk patients with thr
ee-vessel or left main stem coronary artery disease has resulted in th
e definition of one characteristic pattern: ST-segment depression in l
eads I, II, and V-4-V-6 and elevation in lead aVR. This study evaluate
d the generation of such criteria using two self-learning techniques:
neural networks and induction algorithms. In 113 patients, five variab
les, including the amount of ST elevation, the number of leads with ab
normal ST-segments, and this above-mentioned characteristic sign, were
correlated with the number of narrowed vessels. All patients were ran
domly subdivided into a training (n = 63) and test set (n = 50), strat
ified for both this characteristic sign and for the vessel involved. U
sing the learning set, the neural network and the induction algorithm
were trained separately to identify (1) pure left main stem disease an
d (2) three-vessel disease and left main stem disease. The neural netw
ork was trained for 1,000 runs. The induction algorithm was trained, a
llowing all variables to be used in any order. The experiments were re
peated after adding weight factors to promote the recognition of the m
ore severe cases. Subsequently, the ST elevation in all 12 leads was a
dded to the training and test sets, once with and once without the pol
arity of the ST deviation. Altogether, 18 different combinations were
evaluated. Basically, the neural network and the induction algorithm a
pproach misclassified the same cases in corresponding test combination
s. The application of weight factors either did not influence the clas
sification or improved the results at the cost of the nonsupported cat
egory. The inclusion of the 12 additional parameters did not necessari
ly improve and sometimes dramatically worsened the classification proc
ess.