Estimating and Predicting Hemodynamic Changes During Hemodialysis

Hemodialysis

Hemodialysis, Hemodynamics

Machine learning techniques and algorithms originally developed for use in the field of robotics can be applied to continuous, noninvasive physiological waveform data to discover hidden, hemodynamic relationships. Newly developed algorithms can, in real-time: 1) estimate acute blood loss volume, 2) monitor and estimate fluid resuscitation needs, 3) predict cardiovascular collapse well ahead of any clinically significant changes in standard vital signs, and 4) estimate intracranial pressure. We hypothesize that these same methods can be used to monitor volume loss during hemodialysis, as well as predict intradialytic hypotension, well before it occurs.

1. Collect physiological waveform data from patients undergoing hemodialysis at the University of Colorado Hospital, Children's Hospital Colorado, and Fresenius Medical Centers using non-invasive monitoring techniques. 2. Combine the physiological data from patient monitors with clinical and demographic data, including age, gender, race, problem list, reason for dialysis, estimated dry weight, volume removed, arterial and venous pressures, etc. for use in developing mathematical models of hemodialysis. 3. Develop robust, real-time, computational models for: * estimating acute intravascular volume loss during hemodialysis * predicting an optimal, individual specific, intravascular volume to be removed during a hemodialysis session * predicting intradialytic hypotension 4. Determine: * which non-invasive signals are relevant to each model type * which features extracted from these signals are relevant * which algorithms are capable of using the extracted features for each decision type

No related papers found yet.

None listed

United States