Combining Neural Network and Wavelet Transformn for Trigger Asynchrony Detection

Trigger asynchrony isa phenomenonwhen the ventilator isoutofsynchronization withbreathes ofrespiratory patients. Theoccurrence oftrigger asynchrony wouldcause discomfort andharmstothepatients. Thusunderstanding the trigger asynchrony situation forbetter setting theventilator parameters tolowerthepossibility ofoccurrence oftrigger asynchrony iscritical torespiratory patient care.Thispaper proposes thecombination ofneuralnetworks andwavelet feature extraction fortrigger asynchrony detection. The performances usingvarioustraining situations are also compared. A breath cycle iscomposed ofinspiratory phaseand expiratory phase. Inthis paperwealsoexplore theperformance differences betweenthesituation whentheneuralnetwork detection isapplied withthesametrained neural network for inspiratory andexpiratory phases andthesituation whenthe detection and isapplied withdifferent neuralnetworkfor inspiratory andexpiratory phases. Itwasfoundthatalthough separating detection withdifferent neuralnetworks forthe inspiratory andexpiratory phases requires slightly moretime, it achieves higher performance thanthatthedetection isapplied withthesameneural network forbothphases. Theresults are alsocompared withtheresults byphysicians' observations for accuracy evaluation.