Sensor systems and methods for characterizing health conditions
A sensing system comprising a hand-held sensing device with a vibracoustic sensor module (VSM). The VSM comprises a voice coil component comprising a coil holder supporting wire windings; a magnet component comprising a magnet supported by a frame, a magnet gap configured to receive at least a portion of the voice coil component in a spaced and moveable manner; a connector connecting the voice coil component to the magnet component, the connector being compliant and permitting relative movement of the voice coil component and the magnet component; a diaphragm configured to induce a movement of the voice coil component in the magnet gap responsive to incident acoustic waves; a housing for retaining the vibroacoustic sensor module having a handle end and a sensor end, the sensor end having an opening, the VSM positioned such that at least a portion of the diaphragm extends across the opening.
1 . A method comprising:
receiving vibroacoustic data corresponding to a first training set of subjects having a bodily condition and a second training set of subjects having an absence of the bodily condition, wherein the vibroacoustic data was recorded by sensing devices, and wherein each of the sensing devices comprises a vibroacoustic sensor module comprising a voice coil component, a magnet component, a connector, and a diaphragm;
segmenting the vibroacoustic data in the time domain into overlapping time windows;
splitting the overlapping time windows in the frequency domain into frequency ranges;
extracting feature sequences from the split windows;
training a machine learning model, using the feature sequences, to compute a biosignature corresponding to the bodily condition;
determining, by the trained machine learning model, and based on the biosignature, a bodily condition of a subject not part of the first or second training set; and
outputting an indication of the bodily condition of the subject.
2 . The method of claim 1 , further comprising:
receiving vibroacoustic data from the subject while the subject is in two or more different positions and while the subject is performing two or more different actions in each of the two or more positions.
3 . The method of claim 2 , wherein the two or more different positions comprise sit, neutral stand, supine, or left lateral decubitus position, and wherein the two or more different actions comprise neutral cough, hand squeeze, or vocalizations that reveal pulmonary consolidation.
4 . The method of claim 1 , further comprising:
receiving, from a volatile organic compound sensor, exhaled breath data corresponding to the subject; and
extracting, from the exhaled breath data, a biological signal component of the exhaled breath data.
5 . The method of claim 1 , further comprising:
receiving, from a machine vision sensor module, movement data corresponding to the subject;
extracting, from the movement data, movement patterns of chest, nostrils and ribs; and
determining, based on the movement patterns of chest, nostrils and ribs, at least a portion of the biosignature.
6 . The method of claim 1 , further comprising:
capturing, by a machine vision sensor module, skin motion amplification of movements of the subject; and
determining one or more physiological parameters of the subject based on the skin motion amplification of movements, wherein the one or more physiological parameters comprise blood pressure, heart rate, respiratory rate, or hemoglobin.
7 . The method of claim 1 , further comprising:
receiving input from one or more additional sensors; and
determining and differentiating, using the input from the one or more additional sensors, between multiple disease biosignatures.
8 . The method of claim 1 , wherein:
the voice coil component comprises a coil holder supporting wire windings;
the magnet component a magnet supported by a frame, the magnet having a magnet gap configured to receive at least a portion of the voice coil component in a spaced and moveable manner;
the connector connects the voice coil component to the magnet component, the connector being compliant and permitting relative movement of the voice coil component and the magnet component;
the diaphragm is configured to induce a movement of the voice coil component in the magnet gap responsive to incident acoustic waves;
the sensing devices further comprise an Inertial Measurement Unit (IMU) mounted to the diaphragm or to the connector; and
at least a portion of the vibroacoustic data is recorded by the IMU.
9 . The method of claim 8 , wherein the IMU and the voice coil component are attached centrally in relation to the frame.
10 . The method of claim 8 , wherein the IMU and the voice coil component are attached asymetrically in relation to the frame.
11 . The method of claim 1 , wherein each vibroacoustic sensor module of the sensing devices has a bandwidth that at least partially overlaps with a bandwidth of another vibroacoustic sensor module of the sensing devices.
12 . The method of claim 11 , wherein each vibroacoustic sensor module of the sensing devices is optimized for low noise and high signal-to-noise ratio signal acquisition, and wherein each vibroacoustic sensor module of the sensing devices is a multi-layered circuit comprising a ground layer plane configured to prevent ground loops, provide a low resistance ground, or act as a shield between signal lines, and wherein the method further comprises applying second to fourth order low pass filtering to the vibroacoustic data.
13 . The method of claim 1 , further comprising receiving environmental data relating to an environment of the subject and calibrating vibroacoustic data corresponding to the subject with the environmental data.
14 . The method of claim 1 , wherein the frequency ranges correspond to less than 20 Hz, 20-20,000 Hz, and greater than 20,000 Hz.
15 . The method of claim 1 , wherein the biosignature is computed from subbands covering cross-frequency couplets in bark-scale.
16 . The method of claim 1 , wherein the vibroacoustic data lower frequency is as low as 0.01 Hz.
17 . The method of claim 1 , wherein the vibroacoustic data is received from a plurality of body positions or from a subject while the subject is performing a plurality of activities.
18 . The method of claim 1 , wherein the vibroacoustic data is received from a plurality of body locations.
19 . A method comprising:
receiving vibroacoustic data corresponding to a subject, wherein the vibroacoustic data was recorded by a sensing device comprising a vibroacoustic sensor module comprising a voice coil component, a magnet component, a connector, and a diaphragm;
determining, by a machine learning model and based on the vibroacoustic data, a bodily condition of the subject, wherein the machine learning model was trained by:
receiving vibroacoustic data corresponding to a first training set of subjects having the bodily condition and a second training set of subjects having an absence of the bodily condition,
segmenting the vibroacoustic data in the time domain into overlapping time windows, splitting the overlapping time windows in the frequency domain into frequency ranges, extracting feature sequences from the split windows, and
training the machine learning model, using the feature sequences, to compute a biosignature corresponding to the bodily condition; and
outputting an indication of the bodily condition of the subject.