Automatic change in condition monitoring by passive sensor monitoring and machine learning
A machine learning system passively monitors sensor data from the living space of a patient. The sensor data may include audio data. Audio features are generated. A trained machine learning model is used to detect a change in condition. In some implementations, the machine learning model is trained in a learning phase based on training data that includes questionnaires completed by caregivers and identified audio features.
1. A system, comprising:
a processor; and
a memory storing computer instructions that, when executed, cause the processor to implement operations including:
monitoring an output of a passive sensor monitor of a living space of a patient;
providing the monitored passive sensor output to a machine learning system having a machine learning model including one or more classifiers trained to identify changes in at least one health condition of the patient based at least in part on the monitored passive sensor output including an audio output, wherein the one or more classifiers are trained by a questionnaire, answered by a caregiver of the patient, having a set of condition questions associated with one or more audio features of the audio output of the monitored passive sensor output, to identify audio features that include at least one of speech utterances, pain sounds, bathroom sounds, kitchen sounds, falling sounds, walking sounds, and sleeping sounds; and
detecting, by the machine learning system, a change in at least one health condition of the patient based at least in part on an audio feature classification of the audio output.
2. A system, comprising:
a machine learning system having at least one processor, a memory, and machine learning model, the machine learning model configured to:
monitor an output of a passive sensor monitor of a living space of a patient;
provide the monitored passive sensor output to a portion of the machine learning system having one or more classifiers trained to identify changes in at least one health condition of the patient based at least in part on the monitored passive sensor output including an audio output, wherein the one or more classifiers are trained by a questionnaire, answered by a caregiver of the patient, having a set of condition questions associated with one or more audio features of the audio output of the monitored passive sensor output, to identify audio features that include at least one of speech utterances, pain sounds, bathroom sounds, kitchen sounds, falling sounds, walking sounds, and sleeping sounds; and
detect a change in at least one health condition of the patient based at least in part on an audio feature classification of the audio output.
3. A method comprising:
monitoring an output of a passive sensor monitor of a living space of a patient;
providing the monitored passive sensor output to a machine learning system having a machine learning model including one or more classifiers trained to identify changes in at least one health condition of the patient based at least in part on the monitored passive sensor output including an audio output, wherein the one or more classifiers are trained by a questionnaire, answered by a caregiver of the patient, having a set of condition questions associated with one or more audio features of the audio output of the monitored passive sensor output, wherein the audio features include at least one of pain sounds, bathroom sounds, kitchen sounds, falling sounds, walking sounds, and sleeping sounds; and
detecting, by the machine learning system, a change in at least one health condition of the patient based at least in part on an audio feature classification of the audio output.
4. A method comprising:
in a training phase, monitoring an output of a passive audio monitor of a living space of a patient;
generating labels specific to particular types of sounds in a patient's living area by having a caregiver answer a health questionnaire to associate audio features of the particular types of sounds in the patient's living area with specific health condition events, wherein the audio features include at least one of speech utterances, pain sounds, bathroom sounds, kitchen sounds, falling sounds, walking sounds, and sleeping sounds; and
training, based at least in part on the generated labels, a machine learning classification model to generate a trained machine model, the trained machine learning classification model automatically determining answers to a health condition questionnaire based on audio feature vectors in the patient's living area.
5. The method of claim 4 , further comprising:
in a deployment phase, monitoring an output of a passive audio monitor of a living space of a patient;
generating audio feature vectors from the monitored output, classifying the audio feature vectors, and generating an input for the trained machine learning classification model of a machine learning system to identify changes in at least one health condition of the patient; and
detecting, by the machine learning system, a change in at least one health condition of the patient.
6. A method, comprising:
training a machine learning model to automatically identify a change in condition of a patient based at least in part on monitored audio feature data from a living area of patient and information provided by a caregiver, wherein the training comprises using training data generated from monitored audio data and responses of a caregiver to a questionnaire to annotate the audio feature data; and
using the trained machine learning model to detect a change in condition of the patient based on monitored audio data from the living area of the patient.
7. A system, comprising:
a processor; and
a memory storing computer instructions that, when executed, cause the processor to implement operations including:
monitoring an output of a passive sensor monitor of a living space of a patient;
providing the monitored passive sensor output to a machine learning system having a machine learning model including one or more classifiers trained to identify changes in at least one health condition of the patient based at least in part on the monitored passive sensor output including an audio output, wherein the one or more classifiers are trained by a questionnaire, answered by a caregiver of the patient, having a set of condition questions associated with one or more audio features of the audio output of the monitored passive sensor output, to identify audio features that include at least one of pain sounds, bathroom sounds, kitchen sounds, walking sounds, and sleeping sounds; and
detecting, by the machine learning system, a change in at least one health condition of the patient based at least in part on an audio feature classification of the audio output.
8. A system, comprising:
a machine learning system having at least one processor, a memory, and machine learning model, the machine learning model configured to:
monitor an output of a passive sensor monitor of a living space of a patient;
provide the monitored passive sensor output to a portion of the machine learning system having one or more classifiers trained to identify changes in at least one health condition of the patient based at least in part on the monitored passive sensor output including an audio output, wherein the one or more classifiers are trained by a questionnaire, answered by a caregiver of the patient, having a set of condition questions associated with one or more audio features of the audio output of the monitored passive sensor output, to identify audio features that at least one of pain sounds, bathroom sounds, kitchen sounds, walking sounds, and sleeping sounds; and
detect a change in at least one health condition of the patient based at least in part on an audio feature classification of the audio output.