SYSTEMS AND METHODS FOR DETERMINATION OF BLOOD POTASSIUM LEVELS THROUGH MACHINE LEARNING
Systems and methods disclosed herein are directed to determining serum potassium levels of a patient wearing a biosensing device. A first such method includes operations of capturing a first energy measurement reading by an energy detecting element of an optical sensor, wherein the optical sensor is a component of a biosensing device, and wherein the biosensing device is disposed on a skin surface of the patient, performing a feature extraction on the first energy measurement reading resulting in a feature vector representative of volumetric variations in blood flow of the patient, deploying a trained machine learning model configured to take the feature vector as input and determine a serum potassium level classification of the patient at a time corresponding to when the first energy measurement reading was captured, and generating a graphic user interface (GUI) that displays the serum potassium level classification.
1 . A computerized method comprising:
capturing a first energy measurement reading by an energy detecting element of an optical sensor, wherein the optical sensor is a component of a biosensing device, and wherein the biosensing device is disposed on a skin surface of the patient;
performing a feature extraction on the first energy measurement reading resulting in a feature vector representative of volumetric variations in blood flow of the patient;
deploying a trained machine learning model configured to take the feature vector as input and determine a serum potassium level classification of the patient at a time corresponding to when the first energy measurement reading was captured; and
generating a graphic user interface (GUI) that displays the serum potassium level classification.
2 . The computerized method of claim 1 , wherein the optical sensor further includes a light source configured to emit light, wherein the first energy measurement is reflected or refracted light that corresponds to the light emitted from the light source, and wherein reflection or refraction of the light occurs as the light is traveling to or through a vessel or homogeneously perfused tissue site.
3 . The computerized method of claim 1 , further comprising:
prior to deployment of the machine learning model, capturing, by a temperature sensor of the optical sensor, a skin temperature reading of the patient at the time that the first energy measurement reading was captured; and
determining whether the skin temperature reading satisfies a temperature threshold comparison.
4 . The computerized method of claim 3 , further comprising:
when the skin temperature reading satisfies the temperature threshold comparison, continuing with analysis of the first energy measurement reading; and
when the skin temperature reading does not satisfy the temperature threshold comparison, rejecting the first energy measurement reading and excluding the first energy measurement reading from analysis by the machine learning model.
5 . The computerized method of claim 1 , further comprising:
prior to deployment of the machine learning model, capturing, by an accelerometer component of the biosensing device, accelerometer reading indicative of movement of the patient at the time that the first energy measurement reading was captured; and
determining whether the accelerometer reading satisfies a motion threshold comparison.
6 . The computerized method of claim 5 , further comprising:
when the accelerometer reading satisfies the motion threshold comparison, continuing with analysis of the first energy measurement reading; and
when the accelerometer reading does not satisfy the motion threshold comparison, rejecting the first energy measurement reading and excluding the first energy measurement reading from analysis by the machine learning model.
7 . The computerized method of claim 1 , wherein the feature vector includes features extracted from a PPG waveform generated from the reflected or refracted light, wherein the features include one or more of amplitude, area, width, maximum or minimum slope, or location of other fiducial points related to events during the cardiac cycle.
8 . The computerized method of claim 1 , wherein the trained machine learning model is trained through a training process that includes:
a forward pass that includes passing historical data through a machine learning algorithm, wherein the historical data includes features and target values of serum potassium,
a loss calculation operation that includes determining a difference between predicted values and the target values,
a backward propagation step during which includes computing how each parameter contributed to an error in a prediction determined in the forward pass, and
a parameter revision step that includes revising of initial internal variables such that the trained machine learning model is configured to determine a serum potassium level, and wherein the serum potassium level is classified according to one or more threshold comparisons.
9 . A computing device, comprising:
a processor; and
a non-transitory computer-readable medium having stored thereon instructions that, when executed by the processor, cause the processor to perform operations including:
capturing a first energy measurement reading by an energy detecting element of an optical sensor, wherein the optical sensor is a component of a biosensing device, and wherein the biosensing device is disposed on a skin surface of the patient;
performing a feature extraction on the first energy measurement reading resulting in a feature vector representative of volumetric variations in blood flow of the patient;
deploying a trained machine learning model configured to take the feature vector as input and determine a serum potassium level classification of the patient at a time corresponding to when the first energy measurement reading was captured; and
generating a graphic user interface (GUI) that displays the serum potassium level classification.
10 . The computing device of claim 9 , wherein the optical sensor further includes a light source configured to emit light, wherein the first energy measurement is reflected or refracted light that corresponds to the light emitted from the light source, and wherein reflection or refraction of the light occurs as the light is traveling to or through a vessel or homogeneously perfused tissue site.
11 . The computing device of claim 9 , wherein the operations further include:
prior to deployment of the machine learning model, capturing, by a temperature sensor of the optical sensor, a skin temperature reading of the patient at the time that the first energy measurement reading was captured; and
determining whether the skin temperature reading satisfies a temperature threshold comparison.
12 . The computing device of claim 11 , wherein the operations further include:
when the skin temperature reading satisfies the temperature threshold comparison, continuing with analysis of the first energy measurement reading; and
when the skin temperature reading does not satisfy the temperature threshold comparison, rejecting the first energy measurement reading and excluding the first energy measurement reading from analysis by the machine learning model.
13 . The computing device of claim 9 , wherein the operations further include:
prior to deployment of the machine learning model, capturing, by an accelerometer component of the biosensing device, accelerometer reading indicative of movement of the patient at the time that the first energy measurement reading was captured; and
determining whether the accelerometer reading satisfies a motion threshold comparison.
14 . The computing device of claim 13 , wherein the operations further include:
when the accelerometer reading satisfies the motion threshold comparison, continuing with analysis of the first energy measurement reading; and
when the accelerometer reading does not satisfy the motion threshold comparison, rejecting the first energy measurement reading and excluding the first energy measurement reading from analysis by the machine learning model.
15 . The computing device of claim 9 , wherein the feature vector includes features extracted from a PPG waveform generated from the reflected or refracted light, wherein the features include one or more of amplitude, height, area, width, maximum or minimum slope, or location of other fiducial points related to events during the cardiac cycle.
16 . The computing device of claim 9 , wherein the trained machine learning model is trained through a training process that includes:
a forward pass that includes passing historical data through a machine learning algorithm, wherein the historical data includes features and target values of serum potassium,
a loss calculation operation that includes determining a difference between predicted values and the target values,
a backward propagation step during which includes computing how each parameter contributed to an error in a prediction determined in the forward pass, and
a parameter revision step that includes revising of initial internal variables such that the trained machine learning model is configured to determine a serum potassium level, and wherein the serum potassium level is classified according to one or more threshold comparisons.
17 . A non-transitory computer-readable medium having stored thereon instructions that, when executed by one or more processors, cause the one or more processor to perform operations including:
capturing a first energy measurement reading by an energy detecting element of an optical sensor, wherein the optical sensor is a component of a biosensing device, and wherein the biosensing device is disposed on a skin surface of the patient;
performing a feature extraction on the first energy measurement reading resulting in a feature vector representative of volumetric variations in blood flow of the patient;
deploying a trained machine learning model configured to take the feature vector as input and determine a serum potassium level classification of the patient at a time corresponding to when the first energy measurement reading was captured; and
generating a graphic user interface (GUI) that displays the serum potassium level classification.
18 . The non-transitory computer-readable medium of claim 17 , wherein the optical sensor further includes a light source configured to emit light, wherein the first energy measurement is reflected or refracted light that corresponds to the light emitted from the light source, and wherein reflection or refraction of the light occurs as the light is traveling to or through a vessel or homogeneously perfused tissue site.
19 . The non-transitory computer-readable medium of claim 17 , wherein the operations further include:
prior to deployment of the machine learning model, capturing, by a temperature sensor of the optical sensor, a skin temperature reading of the patient at the time that the first energy measurement reading was captured; and
determining whether the skin temperature reading satisfies a temperature threshold comparison.
20 . The non-transitory computer-readable medium of claim 19 , wherein the operations further include:
when the skin temperature reading satisfies the temperature threshold comparison, continuing with analysis of the first energy measurement reading; and
when the skin temperature reading does not satisfy the temperature threshold comparison, rejecting the first energy measurement reading and excluding the first energy measurement reading from analysis by the machine learning model.
21 . The non-transitory computer-readable medium of claim 17 , wherein the operations further include:
prior to deployment of the machine learning model, capturing, by an accelerometer component of the biosensing device, accelerometer reading indicative of movement of the patient at the time that the first energy measurement reading was captured; and
determining whether the accelerometer reading satisfies a motion threshold comparison.
22 . The non-transitory computer-readable medium of claim 21 , wherein the operations further include:
when the accelerometer reading satisfies the motion threshold comparison, continuing with analysis of the first energy measurement reading; and
when the accelerometer reading does not satisfy the motion threshold comparison, rejecting the first energy measurement reading and excluding the first energy measurement reading from analysis by the machine learning model.
23 . The non-transitory computer-readable medium of claim 17 , wherein the feature vector includes features extracted from a PPG waveform generated from the reflected or refracted light, wherein the features include one or more of amplitude, height, area, width, maximum or minimum slope, or location of other fiducial points related to events during the cardiac cycle.
24 . The non-transitory computer-readable medium of claim 17 , wherein the trained machine learning model is trained through a training process that includes:
a forward pass that includes passing historical data through a machine learning algorithm, wherein the historical data includes features and target values of serum potassium,
a loss calculation operation that includes determining a difference between predicted values and the target values,
a backward propagation step during which includes computing how each parameter contributed to an error in a prediction determined in the forward pass, and
a parameter revision step that includes revising of initial internal variables such that the trained machine learning model is configured to determine a serum potassium level, and wherein the serum potassium level is classified according to one or more threshold comparisons.