As the population ratio of the aged people increases, the needs for healthcare concentrated on the medical treatment also increase. And there occurs high pressure to provide high quality and patient-oriented medical services to the consumers in consequence. But the services cannot follow up the huge amount of such demands because of the rapidly increasing medical data.
Since PC has been widely available, many researchers have proposed and implemented clinical decision support systems especially for the heart diseases that have high portion of the sudden death. But the previous works have not shown extremely good performance with respect to their long history because of the special characteristics and difficulties of the medical decision-making. They also have considered general systems that are fitted to the all patients rather than to one or specific patient groups. So, the personalized summarization of each patient's medical data is a critical issue of the user-oriented clinical decision support system.
For this purpose, the layered fuzzy decision tree concept that is a kind of decision trees is proposed. The layered fuzzy decision tree classifies certain ECG waveforms using the features extracted by discrete wavelet transform that has good performance both in the time and the frequency domain. And by expansion of the feature space dimension, the layered fuzzy decision tree constructs class-partitioning hypersurfaces with reasonably simple and general methodology. The layer concept can provide a user-oriented clinical decision-supporting algorithm. Several simulation results show the validity and the generality of the layered fuzzy decision tree algorithm.