Electronically assisted chemical stimulus for symptom intervention

Abstract

A symptom intervention system monitors data representative of a user's movement, identifies an onset of a symptom of a physical condition, and applies an actuation to intervene with the identified onset. A machine-learned model is trained to identify an onset of a symptom based on the monitored data . The system may use the machine-learned model to determine whether to modify an upcoming administration of a chemical stimulus that is administered to the user to treat their physical condition. The system may determine a modification to a dose or a time associated with the upcoming administration of the stimulus and apply the stimulus to the user based on the determined modification. The system may use the machine-learned model to determine that the user is exhibiting a particular symptom of their physical condition. Depending on the symptom, the system may depolarize or hyperpolarize neurons of the user.

What is claimed is: 
     
         1 . A method comprising:
 identifying a chemical stimulus administered to a user to treat a physical condition of a user, wherein an upcoming administration of the chemical stimulus is characterized by at least one of a dose and a time to administer the chemical stimulus;   monitoring a plurality of movement signals representative of movement of the user;   determining, using a machine-learned model configured to identify an onset of a symptom of the physical condition based on the monitored plurality of movement signals, whether to modify the upcoming administration of the chemical stimulus; and   in response to determining to modify the upcoming administration of the chemical stimulus:
 determining a modification to the dose or the time associated with the upcoming administration of the chemical stimulus; and 
 applying the chemical stimulus to the user based on the determined modification. 
   
     
     
         2 . The method of  claim 1 , further comprising:
 receiving historical activity data collected from a plurality of sensors configured to monitor a given user's activity data, the historical activity data including at least one of historical movement signals, hormone activity, a previous administration of the chemical stimulus, a heart rate, or a respiration rate;   labeling the historical activity data with a given symptom label representative of a corresponding symptom characterized by the historical activity data;   creating a first training set based on the labeled historical activity data; and   training the machine-learned model using the first training set.   
     
     
         3 . The method of  claim 2 , wherein the received historical activity data is collected from sensors monitoring a plurality of users having the physical condition, and further comprising:
 labeling the monitored plurality of movement signals with a symptom label representative of the symptom characterized by the monitored plurality of movement signals;   creating a second training set using the labeled plurality of movement signals; and   retraining the machine-learned model using the second training set such that the machine-learned model is customized to motions of the user.   
     
     
         4 . The method of  claim 2 , further comprising:
 receiving feedback of the determined modification indicating a measure of approval that the user has with the determined modification;   modifying an association between the identified onset of the symptom of the physical condition and the monitored plurality of movement signals; and   retraining the machine-learned model using the modified association.   
     
     
         5 . The method of  claim 1 , further comprising monitoring hormone activity of the user using a plurality of sensors configured to measure at least one of a level of a hormone or a level of a biomolecule regulated by the hormone, wherein the machine-learned model is configured to identify the onset of the symptom of the physical condition further based on the monitored hormone activity. 
     
     
         6 . The method of  claim 1 , wherein determining, using the machine-learned model, whether to modify the upcoming administration of the chemical stimulus comprises:
 generating a feature vector representative of the monitored plurality of movement signals and one or more of a hormone activity of the user, a previous administration of the chemical stimulus, and motor intent data of the user;   applying the machine-learned model to the feature vector, wherein machine-learned model identifies the onset of the symptom with a confidence score as being associated with the feature vector; and   in response to the confidence score exceeding a threshold confidence, determining to modify the upcoming administration of the chemical stimulus.   
     
     
         7 . The method of  claim 1 , further comprising:
 determining an “on” time duration of a previous administration of the chemical stimulus, the “on” time duration starting at a first time to administer the chemical stimulus and ending at an occurrence of the symptom after the first time to administer the chemical stimulus, the first occurrence of the symptom identified using the machine-learned model;   determining an “off” time duration of the previous administration of the chemical stimulus, the “off” time duration starting at the first occurrence of the symptom and ending at a second time to administer the chemical stimulus after the first time; and   wherein determining whether to modify the upcoming administration of the chemical stimulus comprises:
 in response to determining that the “off” time duration is greater than the “on” time duration, determining to modify the upcoming administration. 
   
     
     
         8 . The method of  claim 1 , further comprising:
 causing the client device to render a graphical user interface (GUI) comprising user input fields to approve or reject the determined modification; and   in response to receiving a user input indicating that the determined modification is approved, modifying the dose or the time associated with the upcoming administration of the chemical stimulus.   
     
     
         9 . The method of  claim 1 , wherein the plurality of movement signals is a first plurality of movement signals, further comprising:
 measuring a second plurality of movement signals at a first joint of the user;   measuring a third plurality of movement signals at a second joint of the user, the second joint symmetric about the sagittal plane to the first joint;   determining a first kinematic metric score based on a comparison of the second plurality of movement signals to the third plurality of movement signals, the first kinematic metric score indicative of a measure of symmetry of motion about the sagittal plane;   generating a baseline movement profile of the first joint using historical movement signals collected at the first joint; and   determining a second kinematic metric score based on a comparison of the second plurality of movement signals to the baseline movement profile, the second kinematic metric score indicative of a measure of a variance from an expected movement,   wherein the machine-learned model is configured to identify the onset of the symptom of the physical condition further based on at least one of the first kinematic metric score or the second kinematic metric score.   
     
     
         10 . The method of  claim 1 , further comprising:
 determining a movement frequency response of the plurality of movement signals,   wherein the machine-learned model is configured to identify the onset of the symptom of the physical condition further based on the movement frequency response.   
     
     
         11 . The method of  claim 1 , wherein the plurality of movement signals is a first plurality of movement signals, further comprising:
 measuring a second plurality of movement signals at a muscle group of a foot, a shank, or a thigh of the user, the second plurality of movement signals representative of a phase in a gait cycle;   creating a baseline gait profile using historical movement signals measured at the muscle group; and   determining a gait report score based on a comparison of the second plurality of movement signals to the baseline gait profile,   wherein the machine-learned model is configured to identify the onset of the symptom of the physical condition further based on the gait report score.   
     
     
         12 . The method of  claim 1 , wherein the plurality of movement signals is a first plurality of movement signals, further comprising:
 monitoring a second plurality of movement signals representative of a symptom-affected movement of the user; and   comparing the second plurality of movement signals to a symptom profile, wherein the symptom profile is created using historical movement data representative of movement while a given user is experiencing the symptom without assistance from chemical stimulus,   wherein determining the modification to the dose or the time associated with the upcoming administration of the chemical stimulus is based on the comparison of the second plurality of movement signals to the symptom profile.   
     
     
         13 . The method of  claim 1 , further comprising:
 receiving a plurality of images from a camera, wherein the user is depicted in the plurality of images; and   determining a change in user posture depicted in the plurality of images,   wherein the machine-learned model is configured to identify the onset of the symptom of the physical condition further based on the change in user posture.   
     
     
         14 . The method of  claim 1 , further comprising:
 monitoring motor intent data of the user, the motor intent data including electromyography (EMG) signals; and   determining a frequency response of the motor intent data, the frequency response indicative of an energy of muscle activity of the user;   determining a measure of fatigue based on a comparison of the frequency response and a rested frequency response profile determined using historical EMG signals,   wherein the machine-learned model is configured to identify the onset of the symptom of the physical condition further based on the measure of fatigue.   
     
     
         15 . The method of  claim 1 , further comprising, in response to determining to modify the upcoming administration of the chemical stimulus, providing a biofeedback to the user, the biofeedback including one or more of a sensory cue to promote a neurotypical movement in the user. 
     
     
         16 . The method of  claim 1 , wherein the physical condition is Parkinson's disease and wherein the symptom is one of a gait freeze or a tremor. 
     
     
         17 . The method of  claim 1 , wherein the chemical stimulus is one of levodopa, carbidopa, or baclofen. 
     
     
         18 . The method of  claim 1 , wherein the time to administer the chemical stimulus is an administration time, further comprising:
 identify a stimulus metabolism period indicating a time period between the intake of the chemical stimulus and a peak efficacy of the chemical stimulus,   wherein determining the modification to the administration time comprises:
 determining a time at which an “off” time duration of the chemical stimulus will begin; and 
 updating the administration time to be earlier, by the stimulus metabolism period, than the time at which an “off” time duration of the chemical stimulus will begin. 
   
     
     
         19 . A system comprising a non-transitory computer-readable storage medium storing instructions for execution and a hardware processor configured to execute the instructions, the instructions, when executed, cause the hardware processor to perform steps comprising:
 identifying a chemical stimulus administered to a user to treat a physical condition of a user, wherein an upcoming administration of the chemical stimulus is characterized by at least one of a dose and a time to administer the chemical stimulus;   monitoring a plurality of movement signals representative of movement of the user;   determining, using a machine-learned model configured to identify an onset of a symptom of the physical condition based on the monitored plurality of movement signals, whether to modify the upcoming administration of the chemical stimulus; and   in response to determining to modify the upcoming administration of the chemical stimulus:
 determining a modification to the dose or the time associated with the upcoming administration of the chemical stimulus; and 
 transmitting the determined modification to a client device of the user. 
   
     
     
         20 . A non-transitory computer readable storage medium storing executable instructions that, when executed by one or more processors, cause the one or more processors to perform steps comprising:
 identifying a chemical stimulus administered to a user to treat a physical condition of a user, wherein an upcoming administration of the chemical stimulus is characterized by at least one of a dose and a time to administer the chemical stimulus;   monitoring a plurality of movement signals representative of movement of the user;   determining, using a machine-learned model configured to identify an onset of a symptom of the physical condition based on the monitored plurality of movement signals, whether to modify the upcoming administration of the chemical stimulus; and   in response to determining to modify the upcoming administration of the chemical stimulus:
 determining a modification to the dose or the time associated with the upcoming administration of the chemical stimulus; and 
 transmitting the determined modification to a client device of the user.