Breathing is a vital function intrinsic to the survival of any human being. In preterm infants it is an important indicator of maturation and feeding competency, which is a hallmark for hospital release. The recommended method of measurement of infant respiration is the use of thermistors. Accurate event detection within thermistor generated signals relies heavily upon effective noise reduction, specifically baseline drift removal. Baseline drift originates from several sensor-based factors, including thermistor placement within the sensor and in relation to the infant nares. This work compares four methods for baseline drift removal using the same event detection algorithm. The methods compared were a linear spline subtraction, a cubic spline subtraction, a neural network baseline approximation, and a double differentiation of the thermistor signal. The method yielding the highest event detection rate was shown to be the double differentiation method, which serves to attenuate the baseline drift to zero without approximating and subtracting it.
展开▼
