Predicting arterial blood gas values from venous samples in patients with acute exacerbation chronic obstructive pulmonary disease using artificial neural network.

Arterial blood gas (ABG) has an important role in the clinical assessment of patients with acute exacerbations of chronic obstructive pulmonary disease (AECOPD). Because of ABG complications, an alternative method is beneficial. We have trained and tested five artificial neural networks (ANNs) with venous blood gas (VBG) values (pH, PCO(2), HCO(3), PO(2), and O(2) saturation) as inputs, to predict ABG values in patients with AECOPD. Venous and arterial blood samples were collected from 132 patients. Using the data of 106 patients, the ANNs were trained and validated by back-propagation algorithm. Subsequently, data from the remainder 26 patients was used for testing the networks. The ability of ANNs to predict ABG values and to detect significant hypercarbia was assessed and the results were compared with a linear regression model. Our results indicate that the ANNs provide an accurate method for predicting ABG values from VBG values and detecting hypercarbia in AECOPD.