Systems and methods to detect cardiac events

Abstract

In some embodiments, features related to inter-beat intervals (IBI) detected by a PPG sensor of a wearable device are extracted and provided to a cardiovascular classifier in order to detect likely instances of a cardiac condition such as atrial fibrillation. Some embodiments use features related to the entropy of the IBI data to improve the predictions generated by the cardiovascular classifier. In some embodiments, co-information between the IBI data and IBI data gathered from healthy and AF populations is determined in order to derive features that represent the probability that a given sample of IBI data represents AF or a normal sinus rhythm. In response to determining likely instances of AF based on these features, the wearable device may obtain clinically acceptable data, such as an ECG, to be transmitted to a separate device for review by a clinician.

1 - 20 . (canceled) 
     
     
         21 . A wearable device, comprising:
 a first sensor;   a second sensor; and   a controller having logic configured to, in response to execution by the controller, cause the wearable device to perform actions comprising:
 analyzing a signal generated by the first sensor to determine a series of heartbeat characteristic values for a subject; 
 generate a sequence of window values for the subject based on the series of heartbeat characteristic values using a sliding window; 
 determining whether the series of heartbeat characteristic values for the subject is associated with the cardiac condition by comparing entropy in the sequence of window values for the subject to entropy in sequences of window values for a population associated with a cardiac condition; and 
 in response to determining that the series of heartbeat characteristic values for the subject is associated with the cardiac condition, collecting information using the second sensor. 
   
     
     
         22 . The wearable device of  claim 21 , wherein the actions further comprise transmitting the information collected using the second sensor to a separate device. 
     
     
         23 . The wearable device of  claim 21 , wherein the series of heartbeat characteristic values is a series of inter-beat interval (IBI) values, and wherein generating the sequence of window values for the subject based on the series of heartbeat characteristic values using the sliding window includes:
 comparing the series of IBI values to a threshold value to generate a series of binary values that indicate whether or not each IBI value is greater than the threshold value; and   generating the sequence of window values based on a sliding window over the series of binary values.   
     
     
         24 . The wearable device of  claim 23 , wherein the threshold value is an average IBI value for at least a portion of the series of IBI values. 
     
     
         25 . The wearable device of  claim 21 , wherein comparing entropy in the sequence of window values for the subject to entropy in sequences of window values for the population associated with the cardiac condition includes comparing a co-information of the sequence of window values for the subject and at least one co-information of the sequences of window values for the population associated with the cardiac condition, wherein each co-information is a value that indicates an amount of information shared by all random variables in the corresponding sequences of window values. 
     
     
         26 . The wearable device of  claim 21 , wherein determining whether the series of heartbeat characteristic values for the subject is associated with the cardiac condition by comparing entropy in the sequence of window values for the subject to entropy in sequences of window values for a population associated with a cardiac condition includes:
 determining one or more features based on the entropy comparisons, wherein the one or more features represent probabilities that the series of heartbeat characteristic values for the subject is similar to the series of heartbeat characteristic values for the population; and   providing the features to a cardiovascular classifier to determine whether the series of heartbeat characteristic values for the subject is associated with the cardiac condition.   
     
     
         27 . The wearable device of  claim 21 , wherein the population is a first population, wherein the first population is known to experience the cardiac condition, and wherein the actions further comprise:
 comparing entropy in the sequence of window values for the subject to entropy in sequences of window values for a second population, wherein the second population is known to not experience the cardiac condition.   
     
     
         28 . The wearable device of  claim 21 , wherein the cardiac condition is ventricular fibrillation, atrial fibrillation, tachycardia, or bradycardia. 
     
     
         29 . The wearable device of  claim 21 , wherein the first sensor is a photoplethysmographic sensor 
     
     
         30 . The wearable device of  claim 21 , wherein the second sensor is an electrocardiographic sensor. 
     
     
         31 . A non-transitory computer-readable medium having computer-executable instructions stored thereon that, in response to execution by one or more processors of a computing device, cause the computing device to perform actions for detecting a cardiac condition, the actions comprising:
 analyzing a signal generated by a first sensor to determine a series of heartbeat characteristic values for a subject;   generate a sequence of window values for the subject based on the series of heartbeat characteristic values using a sliding window;   determining whether the series of heartbeat characteristic values for the subject is associated with the cardiac condition by comparing entropy in the sequence of window values for the subject to entropy in sequences of window values for a population associated with a cardiac condition; and   in response to determining that the series of heartbeat characteristic values for the subject is associated with the cardiac condition, collecting information using a second sensor.   
     
     
         32 . The non-transitory computer-readable medium of  claim 31 , wherein the actions further comprise transmitting the information collected using the second sensor to a separate device. 
     
     
         33 . The non-transitory computer-readable medium of  claim 31 , wherein the series of heartbeat characteristic values is a series of inter-beat interval (IBI) values, and wherein generating the sequence of window values for the subject based on the series of heartbeat characteristic values using the sliding window includes:
 comparing the series of IBI values to a threshold value to generate a series of binary values that indicate whether or not each IBI value is greater than the threshold value; and   generating the sequence of window values based on a sliding window over the series of binary values.   
     
     
         34 . The non-transitory computer-readable medium of  claim 33 , wherein the threshold value is an average IBI value for at least a portion of the series of IBI values. 
     
     
         35 . The non-transitory computer-readable medium of  claim 31 , wherein comparing entropy in the sequence of window values for the subject to entropy in sequences of window values for the population associated with the cardiac condition includes comparing a co-information of the sequence of window values for the subject and at least one co-information of the sequences of window values for the population associated with the cardiac condition, wherein each co-information is a value that indicates an amount of information shared by all random variables in the corresponding sequences of window values. 
     
     
         36 . The non-transitory computer-readable medium of  claim 31 , wherein determining whether the series of heartbeat characteristic values for the subject is associated with the cardiac condition by comparing entropy in the sequence of window values for the subject to entropy in sequences of window values for a population associated with a cardiac condition includes:
 determining one or more features based on the entropy comparisons, wherein the one or more features represent probabilities that the series of heartbeat characteristic values for the subject is similar to the series of heartbeat characteristic values for the population; and   providing the features to a cardiovascular classifier to determine whether the series of heartbeat characteristic values for the subject is associated with the cardiac condition.   
     
     
         37 . The non-transitory computer-readable medium of  claim 31 , wherein the population is a first population, wherein the first population is known to experience the cardiac condition, and wherein the actions further comprise:
 comparing entropy in the sequence of window values for the subject to entropy in sequences of window values for a second population, wherein the second population is known to not experience the cardiac condition.   
     
     
         38 . The non-transitory computer-readable medium of  claim 31 , wherein the cardiac condition is ventricular fibrillation, atrial fibrillation, tachycardia, or bradycardia. 
     
     
         39 . The non-transitory computer-readable medium of  claim 31 , wherein the first sensor is a photoplethysmographic sensor 
     
     
         40 . The non-transitory computer-readable medium of  claim 31 , wherein the second sensor is an electrocardiographic sensor.