TY - GEN
T1 - Validation of a federation of collaborative rational agents for the diagnosis of acute coronary syndromes in a population with high probability
AU - Sprockel, John
AU - Diaztagle, Juan
AU - Llanos, Alberto
AU - Castillo, Cristian
AU - Gonzalez, Enrique
N1 - Publisher Copyright:
© 2016 IEEE.
PY - 2017/1/31
Y1 - 2017/1/31
N2 - Acute myocardial infarction is the main cause of death worldwide; it is part of the acute coronary syndromes (ACS) which are characterized by an acute obstruction of the blood flow in the arteries of the heart. ACS diagnosis poses a highly complex problem where the use of intelligent systems represents an opportunity for the optimization of the diagnosis. The objective of the present work is to perform a cross validation of a federation of collaborative rational agents for the diagnosis of ACS in a population with high probability exhibiting chest pain. A study of diagnostic tests was performed, the diagnostic standard criterion was the third redefinition of infarction or some strategy for coronary stratification. The index test was the result of a system based on a federation of collaborative rational agents based on the assembly of neural networks by means of a weighted voting system in accordance with positive likelihood ratios. A sample of 108 patients was calculated and a contingency table was built in order to calculate the operational characteristics. 148 patients were taken into consideration, ACS was discarded in 29,2%, 51,7 exhibited acute infarction, and 19,1% exhibited unstable angina. The federation system reached a precision of 79%, sensibility of 97,1%, specificity of 36,4%, and AUC of 0,672. It is concluded that a multi-agent system based on the assembly of neural networks attained an acceptable performance for the diagnosis of ACS in a population with high probability.
AB - Acute myocardial infarction is the main cause of death worldwide; it is part of the acute coronary syndromes (ACS) which are characterized by an acute obstruction of the blood flow in the arteries of the heart. ACS diagnosis poses a highly complex problem where the use of intelligent systems represents an opportunity for the optimization of the diagnosis. The objective of the present work is to perform a cross validation of a federation of collaborative rational agents for the diagnosis of ACS in a population with high probability exhibiting chest pain. A study of diagnostic tests was performed, the diagnostic standard criterion was the third redefinition of infarction or some strategy for coronary stratification. The index test was the result of a system based on a federation of collaborative rational agents based on the assembly of neural networks by means of a weighted voting system in accordance with positive likelihood ratios. A sample of 108 patients was calculated and a contingency table was built in order to calculate the operational characteristics. 148 patients were taken into consideration, ACS was discarded in 29,2%, 51,7 exhibited acute infarction, and 19,1% exhibited unstable angina. The federation system reached a precision of 79%, sensibility of 97,1%, specificity of 36,4%, and AUC of 0,672. It is concluded that a multi-agent system based on the assembly of neural networks attained an acceptable performance for the diagnosis of ACS in a population with high probability.
KW - Acute coronary syndrome
KW - Acute myocardial infarction
KW - Artificial neural networks
KW - Assembly-based systems
KW - Chest pain
KW - Intelligent systems
KW - Medical diagnosis
KW - Multi-agent systems
KW - Unstable angina
UR - http://www.scopus.com/inward/record.url?scp=85015404999&partnerID=8YFLogxK
U2 - 10.1109/ICMLA.2016.175
DO - 10.1109/ICMLA.2016.175
M3 - Conference contribution
AN - SCOPUS:85015404999
T3 - Proceedings - 2016 15th IEEE International Conference on Machine Learning and Applications, ICMLA 2016
SP - 705
EP - 708
BT - Proceedings - 2016 15th IEEE International Conference on Machine Learning and Applications, ICMLA 2016
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 15th IEEE International Conference on Machine Learning and Applications, ICMLA 2016
Y2 - 18 December 2016 through 20 December 2016
ER -