WSEAS Transactions on Biology and Biomedicine
Print ISSN: 1109-9518, E-ISSN: 2224-2902
Volume 11, 2014
Performance Assessment of Optimized Extreme Learning Machine Based on Evolutionary Computing for Spirometric Data Classification
Authors: ,
Abstract: Spirometry is the frequently performed clinical pulmonary function test to assess the respiratory dynamics. It measures changes in lung volumes and airflows during the forced expiratory maneuver. These investigations are widely used in the diagnosis and management of lung diseases like asthma and chronic obstructive pulmonary disease. However the test requires considerable patient effort and cooperation and is also sometimes prone to misclassification due to interdependency of data. In this work an attempt has been made to differentiate pulmonary obstructive abnormality using neural computing and spirometric data. A fast Extreme Learning Machine (ELM) and evolutionary algorithm based optimized ELM networks are employed for classification and their performance is analyzed. The performance of Extreme Learning Machine (ELM), achieved a generalization accuracy of 91.03% in 0.0019secs. The evolutionary based optimization technique achieved a classification accuracy of 100% yielding a sensitivity and specificity of 100% with a much compact and less complex network. Hence the results indicate that an optimized ELM network is superior in performance but takes longer processing time due to the iterations performed in the optimization of the network weights. Hence it is concluded that the EELM based computerized model is useful in enhancing diagnostic relevance of spirometric investigations and could provide assistance to clinicians in characterizing pulmonary abnormalities.
Search Articles
Keywords: Pulmonary Function Test, Spirometer, COPD, FEV1, Extreme Learning Machine, Evolutionary extreme learning machine, Obstructive lung disease
Pages: 147-156
WSEAS Transactions on Biology and Biomedicine, ISSN / E-ISSN: 1109-9518 / 2224-2902, Volume 11, 2014, Art. #20