Poster 78 - Feedback Control of Mean Arterial Pressure

Saturday, October 26, 2024

10:00 AM - 12:00 PM EST

Location: Exhibit Hall E, F & G

Presenting Author(s)

Benjamin F. Bykov (he/him/his)

High School Research Assistant
Johns Hopkins University, School of Medicine
Baltimore, Maryland, United States

Co-Author(s)

Denis Routkevitch

MD-PhD Candidate
The Johns Hopkins University, School of Medicine, United States

Primary Investigator(s)

Nicholas Theodore, MD

Donlin M. Long Professor of Neurosurgery and Director of the Neurosurgical Spine Center
Johns Hopkins Medicine, United States

Last Author(s)

Nitish Thakor, PhD (he/him/his)

Professor of Biomedical Engineering
Johns Hopkins University, School of Medicine
Baltimore, Maryland, United States

Introduction: Controlling blood pressure is one of the most basic tasks clinicians perform in the hospital, from general anesthesia to stays in the intensive care unit. However, constantly monitoring blood pressure can result in a high cognitive load for doctors, and implementing a feedback controller can alleviate some of this burden. Furthermore, this approach can be more stable in maintaining a blood pressure baseline. Here, we present the systems identification-based design of a feedback controller using norepinephrine, a commonly used blood pressure drug. Technological advancements and age-old medical practices are interwoven, creating contemporary medical solutions.

Materials and

Methods: In 5 rats, a femoral catheterization was performed to measure mean arterial pressure (MAP). Norepinephrine was administered at custom infusion rates (step inputs) using a Kent Genie syringe pump controlled by a MATLAB program. Infusion rate and MAP signals were synchronized, baseline signal values were subtracted and the system transfer function (HPKPD) was estimated from infusion rate (input) to MAP (output) using the TFest function in Matlab. This system was then used to design a proportional-integral-derivative (PID) controller as shown in Figure 1A. Here the error between our target and current MAP was calculated and fed into the controller to output the infusion rate. The PID was tuned using the PID Tuner app in Matlab to a clinically usable response time and transient behavior. We then implemented this controller in vivo in real time.

Results, Conclusions, and Discussions: Prior to our real time implementation, the controller behavior was verified in simulation using a step increase in target MAP (Fig. 1 B, C). In real-time, the target MAP was raised from 99 mm Hg to 108 mm Hg (Fig. 1 D). In response, the infusion rate increased to raise the MAP to above 100 mm Hg. Above a MAP of 108 mm Hg, the infusion rate decreased to not exceed the target MAP (Figure 1 E). Although the system generally followed the target, there was some instability, shown by oscillations around the target. This is likely because of an error between our model and the actual rat, which we plan on addressing through improved model accuracy.

Here we presented a system identification-based, in silico design of PID control of blood pressure. This approach enabled us to bypass tuning the system in vivo. We also implemented this feedback controller in real time, as a proof of concept for further translation. This approach can enhance accuracy and effectiveness in MAP management, leading to better overall care. Future work involves enhancing controller stability by increasing controller accuracy.

Acknowledgements (Optional): I would like to acknowledge Denis Routkevitch for the invaluable opportunity to create this poster submission. His guidance, help, and unwavering support throughout the process have been instrumental in bringing this work to fruition. I also greatly appreciate the preliminary model data he provided, which was essential in developing the content for this project. His expertise and support have had a significant impact.

References (Optional):