Noninvasive cardiovascular event detection

A non-transitory computer-readable medium storing computer program instructions that, when executed by one or more processors of a computing system, effectuate operations comprising:

• retrieving, from a database, a plurality of datasets comprising waveforms representing (i) cardiac cycles of a first plurality of patients that have experienced at least one of a plurality of cardiovascular events and (ii) cardiac cycles of a second plurality of patients that have not experienced any of the plurality of cardiovascular events;
• determining, for each of the first plurality of patients and the second plurality of patients, based on the waveforms representing the cardiac cycles of the first plurality of patients that have experienced at least one of the plurality of cardiovascular events and the waveforms of the second plurality of patients that have not experienced any of the plurality of cardiovascular events, a respective first intrinsic frequency associated with a systolic phase of a cardiac cycle of a respective patient, a respective second intrinsic frequency associated with a diastolic phase of the cardiac cycle, a respective systolic intrinsic phase angle associated with the systolic phase of the cardiac cycle, and a respective diastolic intrinsic phase angle associated with the diastolic phase of the cardiac cycle to obtain a plurality of first intrinsic frequencies, a plurality of second intrinsic frequencies, a plurality of systolic intrinsic phase angles respectively corresponding to the plurality of first intrinsic frequencies, and a plurality of diastolic intrinsic phase angles respectively corresponding to the plurality of second intrinsic frequencies, respectively;
• generating and storing training data based on one or more of the plurality of first intrinsic frequencies, the plurality of second intrinsic frequencies, the plurality of systolic intrinsic phase angles, and the plurality of diastolic intrinsic phase angles corresponding to one of the first plurality of patients or the second plurality of patients;
• training and testing a neural network implemented on the computing system based on the training data to obtain the trained neural network;
• obtaining, with the computing system, data comprising an arterial pressure waveform representing measurements of a cardiac cycle of a first patient, wherein the arterial pressure waveform is a time-varying physiological signal corresponding to the cardiac cycle of the first patient;
• segmenting the arterial pressure waveform into a systolic portion and a diastolic portion based on a dicrotic notch of the arterial pressure waveform;
• computing, based on the data comprising the arterial pressure waveform and using an intrinsic frequency analysis constrained by continuity and periodicity conditions of the arterial pressure waveform, a first intrinsic frequency of a systolic phase of the cardiac cycle of the first patient, and a second intrinsic frequency of a diastolic phase of the cardiac cycle of the first patient,
• wherein the intrinsic frequency analysis comprises fitting a piecewise sinusoidal model to the arterial pressure waveform across the cardiac cycle, the piecewise sinusoidal model including a first component for the systolic phase and a second component for the diastolic phase, and determining the first intrinsic frequency and the second intrinsic frequency by minimizing a difference between the arterial pressure waveform and the piecewise sinusoidal model;
• determining, based on the first intrinsic frequency and the second intrinsic frequency, respectively, from the intrinsic frequency analysis, a systolic intrinsic phase angle describing the systolic phase of the cardiac cycle of the first patient and a diastolic intrinsic phase angle describing the diastolic phase of the cardiac cycle of the first patient,
• wherein the systolic intrinsic phase angle is determined as a phase parameter of the first component associated with the first intrinsic frequency, and the diastolic intrinsic phase angle is determined as a phase parameter of the second component associated with the second intrinsic frequency;
• providing a set of derived cardiac parameters including the first intrinsic frequency, the second intrinsic frequency, the systolic intrinsic phase angle, and the diastolic intrinsic phase angle to the trained neural network configured to determine whether the first patient has suffered at least one of the plurality of cardiovascular events;
• obtaining, from the trained neural network, a set of values each representing a likelihood that the first patient has suffered at least one of the plurality of cardiovascular events;
• determining whether one or more values from the set of values satisfy a threshold condition;
• and
• storing a result of the determination of whether one or more values from the set of values satisfy the threshold condition in memory.

A method implemented by a computing system executing computer program instructions, the method comprising:

• retrieving, from a database, a plurality of datasets comprising waveforms representing (i) cardiac cycles of a first plurality of patients that have experienced at least one of a plurality of cardiovascular events and (ii) cardiac cycles of a second plurality of patients that have not experienced any of the plurality of cardiovascular events;
• determining, for each of the first plurality of patients and the second plurality of patients, based on the waveforms representing the cardiac cycles of the first plurality of patients that have experienced at least one of the plurality of cardiovascular events and the waveforms of the second plurality of patients that have not experienced any of the plurality of cardiovascular events, a respective first intrinsic frequency associated with a systolic phase of a cardiac cycle of a respective patient, a respective second intrinsic frequency associated with a diastolic phase of the cardiac cycle, a respective systolic intrinsic phase angle associated with the systolic phase of the cardiac cycle, and a respective diastolic intrinsic phase angle associated with the diastolic phase of the cardiac cycle to obtain a plurality of first intrinsic frequencies, a plurality of second intrinsic frequencies, a plurality of systolic intrinsic phase angles respectively corresponding to the plurality of first intrinsic frequencies, and a plurality of diastolic intrinsic phase angles respectively corresponding to the plurality of second intrinsic frequencies, respectively;
• generating and storing training data based on one or more of the plurality of first intrinsic frequencies, the plurality of second intrinsic frequencies, the plurality of systolic intrinsic phase angles, and the plurality of diastolic intrinsic phase angles corresponding to one of the first plurality of patients or the second plurality of patients;
• training and testing a neural network implemented on the computing system based on the training data to obtain the trained neural network;
• obtaining, with the computing system, data comprising an arterial pressure waveform representing measurements of a cardiac cycle of a first patient, wherein the arterial pressure waveform is a time-varying physiological signal corresponding to the cardiac cycle of the first patient;
• segmenting the arterial pressure waveform into a systolic portion and a diastolic portion based on a dicrotic notch of the arterial pressure waveform;
• computing, based on the data comprising the arterial pressure waveform and using an intrinsic frequency analysis constrained by continuity and periodicity conditions of the arterial pressure waveform, a first intrinsic frequency of a systolic phase of the cardiac cycle of the first patient, and a second intrinsic frequency of a diastolic phase of the cardiac cycle of the first patient,
• wherein the intrinsic frequency analysis comprises fitting a piecewise sinusoidal model to the arterial pressure waveform across the cardiac cycle, the piecewise sinusoidal model including a first component for the systolic phase and a second component for the diastolic phase, and determining the first intrinsic frequency and the second intrinsic frequency by minimizing a difference between the arterial pressure waveform and the piecewise sinusoidal model;
• determining, based on the first intrinsic frequency and the second intrinsic frequency, respectively, from the intrinsic frequency analysis, a systolic intrinsic phase angle describing the systolic phase of the cardiac cycle of the first patient and a diastolic intrinsic phase angle describing the diastolic phase of the cardiac cycle of the first patient,
• wherein the systolic intrinsic phase angle is determined as a phase parameter of the first component associated with the first intrinsic frequency, and the diastolic intrinsic phase angle is determined as a phase parameter of the second component associated with the second intrinsic frequency;
• providing, with the computing system, the first intrinsic frequency, the second intrinsic frequency, the systolic intrinsic phase angle, and the diastolic intrinsic phase angle to the trained neural network configured to determine whether the first patient has suffered at least one of the plurality of cardiovascular events;
• obtaining, with the computing system, from the trained neural network, a set of values each representing a likelihood that the first patient has suffered at least one of the plurality of cardiovascular events;
• determining, with the computing system, whether one or more values from the set of values satisfy a threshold condition;
• and
• storing, with the computing system, a result of the determination of whether one or more values from the set of values satisfy the threshold condition in memory.