This paper proposes a new recognition method of coronary arterial sten
osis in DSA image, because conventional evaluation algorithms have not
been satisfactory to medical doctors. To solve this problem, the neur
al network which takes into account the experience of the medical doct
or is employed. This method consists of the extraction of characterist
ic parameters which are highly correlated with stenosis and the recogn
ition of stenosis by the neural network. By means of phantom experimen
ts and clinical evaluation, the efficiency of the proposed method has
been shown.