Intra-aortic pressure forecasting

A method, comprising:

• receiving a data set containing intra-aortic pressure waveforms of a plurality of patients, each intra-aortic pressure waveform acquired while a respective patient of the plurality of patients received hemodynamic support from a respective transvalvular micro-axial pump, the data set further comprising, for each of the intra-aortic pressure waveforms, one or more corresponding feature waveforms;
• generating a plurality of sequences for each of the intra-aortic pressure waveforms and corresponding feature waveforms of the data set;
• identifying a subset of the plurality of sequences in which a significant change in intra-aortic pressure was present, each significant change including a change in intra-aortic pressure of greater than a predetermined threshold;
• and
• training a machine learning model based at least in part on (i) the intra-aortic pressure waveforms and (ii) the one or more corresponding feature waveforms of each sequence of the identified subset of the plurality of sequences, the machine learning model being usable to predict a future intra-aortic pressure for a given patient based at least in part on a current intra-aortic pressure and one or more current feature measurements associated with a transvalvular micro-axial pump of the given patient.

A non-transitory computer readable storage medium having instructions stored thereon that, when executed by one or more processors, cause the one or more processors to perform operations including:

• receiving a data set containing intra-aortic pressure waveforms of a plurality of patients, each intra-aortic pressure waveform acquired while a respective patient of the plurality of patients received hemodynamic support from a respective transvalvular micro-axial pump, the data set further comprising, for each of the intra-aortic pressure waveforms, one or more corresponding feature waveforms;
• generating a plurality of sequences for each of the intra-aortic pressure waveforms and corresponding feature waveforms of the data set;
• identifying a subset of the plurality of sequences in which a significant change in intra-aortic pressure was present, each significant change including a change in intra-aortic pressure of greater than a predetermined threshold;
• and
• training a machine learning model based at least in part on (i) the intra-aortic pressure waveforms and (ii) the one or more corresponding feature waveforms of each sequence of the identified subset of the plurality of sequences, the machine learning model being usable to predict a future intra-aortic pressure for a given patient based at least in part on a current intra-aortic pressure and one or more current feature measurements associated with a transvalvular micro-axial pump of the given patient.