The use of EIT in the detection of regional lung dysfunction in prematurely born neonates

Abstract

This research describes the progress of work in developing a system for automated detection of regional lung dysfunction in prematurely born neonates. EIT boundary measurements, observed at each lung region, are treated as a time series. The SPIRIT algorithm is used to extract local (regional) and global patterns from the datasets of healthy and ill neonates. The SAX technique is used to derive a symbolic representation of the global pattern signal. Current results are promising and demonstrate the possibility of characterise EIT boundary signals by 'words'. Such a representation can then be used to train a discrete Hidden Markov Model (HMM) to automatically detect and characterise regional lung function.