Device for automatically detecting lung function variability

Abstract

A device for detecting and quantifying lung function includes a piezoelectric sensor configured for attachment to a mammalian chest wall and to feed signals to an ADC providing digitized signals to a processor. The processor firmware extracts inspiration and expiration times from the digitized signals and generates an I/E ratio from them. The processor has firmware to detect wheezing sounds in the signals. The device has an analog event detector configured to wake-up the processor upon detection of candidate wheeze sounds in the signals. In embodiments, the analog event detector includes bandpass filters coupled to a modeling circuit feeding a correlation circuit, the filters having bandpass adjusted by feedback from a circuit within the analog event detector. In an embodiment, the device uses an FFT with a gated recurrent unit (GRU) with partial reset (GRUPR) neural network to detect wheezes.