Lung function monitoring from heart signals
A breathing monitor has ECG, PPG, or bioimpedance sensors feeding a neural network to provide inspiratory and expiratory phases of breathing and tidal volume (TV), filters for the inspiratory and expiratory phases of breathing and TV; and apparatus configured to provide measurements of breathing rate (RR), and fractional inspiratory time (FIT). In embodiments, the device uses the RR, FIT, and TV to estimate spirometric parameters such as lung obstruction severity, forced expiratory volume in one second (FEV1), forced expiratory volume (FEV), forced vital capacity (FVC), FEV1/FEV ratio, and FEV1/FVC ratio. A method of determining a classification of lung obstruction from heart signals includes feeding heart signals into a neural network to determine TV and inspiratory and expiratory classes used to determine FIT and RR; and using FIT, RR, and TV to determine lung obstruction classification of mild, moderate, severe, or very severe obstructive symptoms.
1 . A method of operating a breathing monitor to determine a classification of lung obstruction or inflammation based upon heart signals from a subject, comprising:
at a processor of the breathing monitor, receiving heart signals acquired from the subject using a heart sensor selected from the group consisting of an electrocardiographic (ECG) sensor, a photoplethysmographic (PPG) sensor, and a bioimpedance sensor;
processing the heart signals using a neural network to generate inspiratory and expiratory class signals based on the digital heart signals and to generate tidal volume signals as a function of peaks and valleys of waveforms corresponding to the inspiratory and expiratory class signals;
determining, via the processor of the breathing monitor, at least one of a force expiratory volume in one second (FEV1) or a forced vital capacity (FVC) using the inspiratory and expiratory class signals and tidal volume signals;
using the FEV1 or the FVC to determine, via the processor of the breathing monitor, a lung obstruction or inflammation classification indicating the subject has mild, moderate, severe, or very severe obstructive or inflammation symptoms;
logging, via the processor of the breathing monitor, the lung obstruction or inflammation classification to create a time progression of the disease; and
outputting, from the breathing monitor, a medical alert based on the determined lung obstruction or inflammation classification.
2 . The method of claim 1 , wherein the weights in the weight memory are previously determined by training a corresponding neural network on a multi-patient database of heart signals and respiration signals.
3 . The method of claim 2 , wherein the multi-patient database further comprises electrocardiographic and respiration signals obtained from the subject.
4 . The method of claim 2 , wherein the heart signals comprise signals from a sensor selected from the group consisting of an electrocardiographic (ECG) sensor, a photoplethysmographic (PPG) sensor, and a bioimpedance sensor.
5 . The method of claim 4 , wherein determining the lung obstruction classification or inflammation is performed with a classifier selected from the group consisting of decision tree classifiers, random forest classifiers, K-nearest-neighbors (KNN) classifiers, neural network classifiers, and support vector machines classifiers.
6 . The method of claim 1 , wherein the neural network is selected from the group consisting of a gated recurrent unit (GRU) neural network, a convolutional neural network, and a long short-term memory (LSTM) neural network.
7 . The method of claim 1 , wherein neural network is further trained on obtained signals to modify the weights in the weight memory.