# Trial‐Based Hemolysis Modeling to Investigate Operating Modes of Continuous‐Flow LVADs

## TL;DR

This paper introduces a new method for predicting blood cell damage in heart pumps using trial data, enabling better treatment strategies for patients.

## Contribution

A CFD-free, trial-based hemolysis modeling approach for online prediction in continuous-flow LVADs.

## Key findings

- The modified power law model achieved an R² of 0.69 for hemolysis prediction.
- Lower pump speed and systemic resistance reduce hemolysis in the Sputnik1 LVAD.
- Speed modulation increases hemolysis in most patient profiles.

## Abstract

The influence of operating modes on pump‐induced hemolysis in continuous‐flow left ventricular assist devices (LVADs) can be assessed using computational fluid dynamics (CFD) simulations alongside power law models derived from shearing device experiments. However, this conventional method incurs high computational costs, limiting the exploration of diverse operating conditions and hindering online hemolysis prediction. This work presents a CFD‐free and trial‐based methodology for determining online‐capable hemolysis models for continuous‐flow LVADs.

The trial‐based hemolysis model is based on a modified power law model, with parameters identified from LVAD hemolysis trials. The dynamic behavior is modeled using the Lagrangian approach. Specifically, this model was determined for the Sputnik1 LVAD and integrated with a lumped‐parameter model of the LVAD‐supported cardiovascular system. Subsequently, hemolysis was predicted across various operating modes and patient conditions.

The RMSE and the R
2 of the modified power law fit were 18.4 [%·mL/h] and 0.69, respectively. The relative error introduced by the Lagrangian approach was below 0.7%. For the Sputnik1, hemolysis decreased with reduced speed. Additionally, lower systemic resistance and diminished left ventricular contractility were associated with lower hemolysis, whereas speed modulation increased hemolysis across most profiles.

The proposed hemolysis model allows to assess various LVAD operating modes and patient conditions, assisting in the selection of low‐hemolysis treatment strategies. For Sputnik1 patients, it is advisable to maintain low pump speed and systemic resistance, while speed modulation should be reserved for those with low hemolysis markers. Integrating this model with online flow sensing would enable online hemolysis prediction.

This work presents a methodology for determining online‐capable hemolysis models for continuous‐flow LVADs, derived from in vitro trials. The models are based on fitting a modified power law and applying the Lagrangian approach. They assist clinicians in selecting treatment strategies that minimize hemolysis, as demonstrated with the Sputnik1 LVAD.