Digital stethoscope for counting coughs, and applications thereof
Embodiments disclosed herein improve digital stethoscopes and their application and operation. A first method detects of a respiratory abnormality using a convolution. A second method counts coughs for a patient. A third method predicts a respiratory event based on a detected trend. A fourth method forecasts characteristics of a future respiratory event. In a fifth embodiment, a base station is provided for a digital stethoscope.
1. A computer-implemented method of counting coughs for a patient comprising:
receiving, over a period of time from a microphone of a digital stethoscope, an auditory signal;
generating, with a control unit of the digital stethoscope, an auditory spectrogram based on the auditory signal, wherein the auditory spectrogram comprises a representation of a spectrum of frequencies of the auditory signal filtered by frequency domain functions;
analyzing, with the control unit of the digital stethoscope, the auditory spectrogram to determine that the auditory signal represents a cough;
tracking, with the control unit of the digital stethoscope, the cough in a cough log; and
transmitting based on the cough log, with a communication unit of the digital stethoscope to a cloud-based service to detect a respiratory abnormality, a message indicating a number of coughs tracked over the period of time for storage on a remote server.
2. The method of claim 1 further comprising:
receiving a further auditory signal collected contemporaneously with the auditory signal, from a further microphone of the digital stethoscope, wherein the further auditory signal represents a background noise; and
adjusting, with the control unit, before analyzing the auditory spectrogram, the auditory signal based on the further auditory signal to reduce the background noise.
3. The method of claim 1 further comprising filtering, using one or more cochlear filters, the auditory signal according to a cochlear model before generating the auditory spectrogram, wherein the cochlear model is a mathematical model representing one or more frequency ranges a mammalian cochlea can hear.
4. The method of claim 3 further comprising:
dividing, with the control unit, the auditory signal into one or more frequency bands; and wherein
filtering, using the one or more cochlear filters, the auditory signal according to the cochlear model is based on filtering the one or more frequency bands.
5. The method of claim 1 wherein analyzing the auditory spectrogram includes applying a machine learning process trained using a set of respiratory noises annotated to identify the cough.
6. The method of claim 5 wherein analyzing the auditory spectrogram includes:
performing a convolution procedure on the auditory spectrogram to generate a plurality of convolution values, wherein the convolution values represent a compressed representation of a portion of the auditory spectrogram, and wherein the convolution procedure is trained to generate one or more features for detecting the cough; applying one or more weights to the convolution values, wherein the weights are predetermined values for detection of the cough; and generating a classification value based on applying the weights to the convolution values, wherein the classification value indicates that the auditory signal includes the cough.
7. The method of claim 6 wherein the convolution procedure is performed repeatedly to successively compress the auditory spectrogram into a convolutional matrix to generate the convolution values.
8. The method of claim 1 further comprising predicting a future respiratory event based on the number of coughs tracked over the period of time.
9. A non-transitory computer readable medium including instructions for counting coughs for a patient comprising:
receiving, over a period of time from a microphone of a digital stethoscope, an auditory signal;
generating, with a control unit of the digital stethoscope, an auditory spectrogram based on the auditory signal, wherein the auditory spectrogram comprises a representation of a spectrum of frequencies of the auditory signal filtered by frequency domain functions;
analyzing, with the control unit of the digital stethoscope, the auditory spectrogram to determine that the auditory signal represents a cough;
tracking, with the control unit of the digital stethoscope, the cough in a cough log; and
transmitting based on the cough log, with a communication unit of the digital stethoscope to a cloud-based service to detect a respiratory abnormality, a message indicating a number of coughs tracked over the period of time for storage on a remote server.
10. The non-transitory computer readable medium of claim 9 with instructions further comprising:
receiving a further auditory signal collected contemporaneously with the auditory signal, from a further microphone of the digital stethoscope, wherein the further auditory signal represents a background noise; and
adjusting, with the control unit, before analyzing the auditory spectrogram, the auditory signal based on the further auditory signal to reduce the background noise.
11. The non-transitory computer readable medium of claim 9 with instructions further comprising filtering, using one or more cochlear filters, the auditory signal according to a cochlear model before generating the auditory spectrogram, wherein the cochlear model is a mathematical model representing one or more frequency ranges a mammalian cochlea can hear.
12. The non-transitory computer readable medium of claim 11 with instructions further comprising:
dividing, with the control unit, the auditory signal into one or more frequency bands; and wherein
filtering, using the one or more cochlear filters, the auditory signal according to the cochlear model is based on filtering the one or more frequency bands.
13. The non-transitory computer readable medium of claim 9 with instructions analyzing the auditory spectrogram includes applying a machine learning process trained using a set of respiratory noises annotated to identify the cough.
14. The non-transitory computer readable medium of claim 13 with instructions further comprising:
performing a convolution procedure on the auditory spectrogram to generate a plurality of convolution values, wherein the convolution values represent a compressed representation of a portion of the auditory spectrogram, and wherein the convolution procedure is trained to generate one or more features for detecting the cough; applying one or more weights to the convolution values, wherein the weights are predetermined values for detection of the cough; and generating a classification value based on applying the weights to the convolution values, wherein the classification value indicates that the auditory signal includes the cough.
15. The non-transitory computer readable medium of claim 14 with instructions wherein the convolution procedure is performed repeatedly to successively compress the auditory spectrogram into a convolutional matrix to generate the convolution values.
16. The non-transitory computer readable medium of claim 9 with instructions further comprising predicting a future respiratory event based on the number of coughs tracked over the period of time.
17. A digital stethoscope for counting coughs for a patient comprising:
a microphone configured to receive an auditory signal over a period of time; a control unit, coupled to the microphone, configured to:
analyze the auditory signal to determine that the auditory signal represents a cough indicative of a respiratory abnormality: generate an auditory spectrogram based on the auditory signal, wherein the auditory spectrogram comprises a representation of a spectrum of frequencies of the auditory signal filtered by frequency domain functions: track the cough in a cough log; and
a communication unit, coupled to the control unit, configured to transmit, based on the cough log, a message to a cloud-based service indicating a number of coughs tracked over the period of time for storage on a remote server to detect the respiratory abnormality.