Neural network and artificial immune algorithms for the classification of medical data series.
Wiesław Wajs
Vol. 16, no. 1 (2012), s. 89-96
This paper describes the applicability of artificial immune algorithms. Medical data series classification technique by Artificial Immune Algorithm is used for Neural Network Algorithm input data definitions. Artificial Immune Algorithms is created and trained for the purpose of Arterial Blood Gas parameters classification: pH, PaCO2, PaO2, HCO3. The main goal of this paper is to develop a artificial neural network technique for Arterial Blood Gases short-term prediction. The main question that is considered is how to predict some dynamic parameters that describe blood gases nature. A model of a physical system has an error associated with its predictions due to the dependences of the physical system's output on uncontrollable and unobservable quantities. The use of artificial methods creates the possibilities of obtaining some parameter values on the proper level of probability. This would provide a direct feedback to the clinical staff about the progress of a patient, the success of individual treatments, and quality of care as well as predicting blood gas value.
Dla rozpoznawania przypadków chorobowych, które są opisane numerycznymi danymi wykorzystano metody sztucznej inteligencji. W pracy wykorzystano dwie metody: metodę sztucznych sieci neuronowych oraz metodę sztucznych sieci immunologicznych. Przedstawiono wyniki uzyskane tymi metodami w odniesieniu do przypadków dysplazji oskrzelowo płucnej dla dzieci, których waga była poniżej 1500 g.