Biomedical Applications
The human body is a complex system that contains many different control loops to maintain life. When one of these feedback loops is not functioning properly, medical devices are used to replace their functionality accordingly. In our research projects, we aim to model the nonlinear behavior of the human body, estimate appropriate parameters for the attending physician and by means of adaptive, safe controllers to ultimately improve the health of the patient.
VIEshunt: Toward a SmartShunt for Hydrocephalus
Hydrocephalus is a medical condition in which excessive accumulation of fluid in the brain causes neurological damage. Up to this date no treatment exists and the most common form of therapy still relies on passive mechanical shunt systems to drain the excessive fluid out of the brain into the abdominal area. The project VIEshunt is our commitment to leading hydrocephalus therapy into the 21st century by developing a ventricular intelligent electromechanical shunt system for hydrocephalus patients.
The aim of developing a novel improvement to traditional passive shunt based treatment of hydrocephalus brings together leading experts in the fields of neurosurgery, pathophysiology, and engineering from a variety of institutions. Togehter we are working toward a smart shunt system by approaching the development from three sides.
- Data-driven modelling of the cerebrospinal fluid dynamics. Much of the pathophysiology of hydrocepahlus is still unknown, however, a detailed quantitiative understanding of the dynamics underlying the disease is at the heart of improving patient healthcare. In order to do so we use modern machine learning and system identification techniques to explore extensive in-vivo data sets.
- Development intelligent hardware systems such as of mechatronic shunts and Hardware-in-the-Loop test-benches. Conventional shunts are basically simple passive mechanical valves, which are unable to perform physiologic control. We are therefore developing Download VIEshunt, a ventricular intelligent electromechanical shunt system for hydrocephalus patients. On the other side of the spectrum, sophisticated test benches are constructed for rapid control prototyping. This speeds up testing and development cycles, decreases costs and minimizes the need for animal trials.
- Design of safe learning-based control systems. Having developed quantitative models and mechatronic shunts enables the design of state estimators for the precise and accurate determination of the patient's health status. Moreover, acquired data and human feedback can be leveraged to adapt to the specific patient needs in order to optimize therapy.
Contact Person
Collaborators
University Hospital Zurich: external page Department of Neurosurgery
University of Zurich: external page Center for Surgical Research; external page Interface Group
Massachusetts Institute of Technology: external page Therapeutic Technology Design and Development (TTDD)
Funding
This project is funded by the external page Swiss National Science Foundation (SNSF), the external page Swiss Innovation Agency (Innosuisse) and external page NCCR Automation.
MRI-conditional Extracorporeal Membrane Oxygenation
Congenital heart disease (CHD) describes one of the most common birth defects in humans with a prevalence of around 7.2 per 1’000 births in Europe. Around 7% of newborns with CHD require immediate open-heart surgery within the first 6 months of their life. During these surgical interventions, a cardiopulmonary bypass (CPB) is used to sustain circulation and oxygenation of the blood. Recent studies have shown that in some cases an increase of white matter injury was observed after the surgery in comparison to before. The exact causes of these injuries have not yet been identified. A promising approach to gain more insights into these dynamics would be to use magnetic resonance imaging (MRI) while the CPB is attached.
The long-term goal is to create a mathematical model which can be used as a decision-support tool during surgery. By providing real-time insights through a partial digital twin of the patient, the system could help clinicians better understand how their interventions affect cerebral perfusion and overall brain health.
Medical devices used around MRI scanners must meet standardized safety and material requirements. Traditionally, two workarounds for prototypes of heart-lung machines (HLMs) have been used to remove MR unsafe components from the room: long blood tubes or long drive shafts from the motor to the pump head. However, long blood tubes are not an option for small children, as the amount of blood required to fill the tubes would be greater than their total blood volume of 400 mL or 1400 mL for infants or children aged five years, respectively. Long drive shafts are also fragile and difficult to handle.
To address the challenges of operating a HLM in close proximity of the MR scanner, we first developed and tested three MR-conditional HLMs for their hydraulic characteristics (Hydraulic In Vitro Characterization of MR-Conditional Blood Pumps external page https://doi.org/10.1111/aor.70041) and hemolysis behavior (Preliminary in vitro hemolysis evaluation of MR-conditional blood pumps external page https://doi.org/10.3389/fmedt.2025.1671938). The roller pump (Figure 2A) showed little change in flow due to change in pressure head and very low blood damage, which is why it was then selected for in-vivo studies to collect real-world data.
Currently, we are building a mathematical model from the collected real-world data to better understand how blood flows to the brain and how different controllable factors influence cerebral perfusion.
Commercial heart–lung machines (HLMs) cannot be safely used near an MRI scanner because many of their components are made of materials that are strongly attracted to the scanner’s magnetic field. To overcome this, we developed three new prototypes designed specifically for MRI environments.
Each prototype is powered by a pneumatic motor instead of an electric one. This allows the motor’s control valve to be placed safely outside the MRI room. The pump frames were built using primarily non-metallic materials, and all sensors were enclosed in copper shielding to reduce interference.
To evaluate MRI compatibility, we measured changes in signal-to-noise ratio (SNR) in structural brain scans. When the brain was positioned at the scanner bore, the SNR decreased by an average of only 8.43%, a difference small enough to be visually undetectable in the images. Hydraulic tests also confirmed that all three prototypes functioned as intended. The study results can be found here: external page https://onlinelibrary.wiley.com/doi/full/10.1111/aor.70041
We also assessed potential blood damage. Each pump was tested using bovine blood circulated through a simple loop with a reservoir, two pressure sensors, and a flow sensor at body temperature (37.5 ± 1.0°C) using a water bath. Due to limitations in one prototype, all pumps were tested at a flow rate of 1 L/min and pressure head of 10 mmHg. Under these conditions, the prototypes performed similarly to current systems. The study results can be found here: external page https://www.frontiersin.org/journals/medical-technology/articles/10.3389/fmedt.2025.1671938/full
The roller pump was then selected for an in-vivo study to collect physiological data. Four trials were carried out in which piglets were cannulated and supported on cardiopulmonary bypass (CPB) inside the MRI scanner. During CPB, we performed cooling, rewarming, and unilateral carotid artery clamping while repeatedly acquiring MRI measurements of blood flow, diffusion, perfusion, and metabolism.
Throughout all experiments, key brain regions, such as the basal ganglia, hippocampus, and internal capsule, were clearly visualized with no imaging artifacts. The pump operated smoothly, and all monitoring signals remained stable. MRI data showed flow-related and metabolic changes. The study results are currently in the process of being published.
Contact Person
Collaborators
University Hospital of Zurich: external page Clinic of Cardiac Surgery
University Hospital of Zurich: external page Institute for Diagnostic and Interventional Radiology
University Animal Hospital of Zurich: external page Anesthesiology
University Animal Hospital of Zurich: external page Clinic of Diagnostic Imaging
University Animal Hospital of Zurich: external page Veterinary Clinical Pathology
Funding
This project is funded by the external page Swiss Innovation Agency (Innosuisse)
Physiological Adaptation of Ventricular Assist Devices
The Zurich Heart project is a multidisciplinary and inter-institutional cooperation between the University of Zurich (UZH), the University Hospital Zurich (USZ), the German Heart Center Berlin and ETH Zurich aimed at developing new technologies for left ventricular assist devices (LVADs). The Download Zurich Heart project was accepted as strategic project by Hochschulmedizin Zürich in 2012.
Currently, more than 23 million people worldwide suffer from heart failure. Mechanical circulatory support, i.e. an implantable mechanical pump, has proven a viable solution for such patients, especially given the shortage of donor hearts for transplants.
The existing technology is mainly based on continuous flow pump operating principles. One of the problems of current clinical LVADs is that they operate at a predefined speed, which is chosen by the physician, without being adapted to the physiological requirements. This lack of adaptation causes various problems, which affect the patient’s heart and circulation. For instance, ventricular suction due to high unloading can lead to myocardial damage. Low unloading can also cause under-perfusion, flow stagnation or even regurgitation and lung edema.
In order to solve these issues, our research concentrates on the following areas:
- Physiological control of LVADs
- Sensor development
- Suction detection and prevention
- Investigation of pulsatile LVAD operation
- Physiological control under bi-ventricular support
The above video shows a novel versatile hybrid mock circulation (vHMC) based on the hardware-in-the-loop concept. It is divided into a hardware part and a numerical model of the cardiovascular system. The implant is connected to the hydraulic interface and interacts in real time with the mimicked cardiovascular system. The vHMC enables dynamic testing of active implants, such as ventricular assist devices for left- or biventricular support, testing of total artificial hearts or of novel pump concepts. In addition, it supports testing of passive implants, like valves or grafts. Besides characterizing implants, new control approaches like physiological controllers that allow adjusting the pump flow to the perfusion demand of the patient can be tested.
The ventricular volume sensor project investigates physical principals, which allow the deduction of the ventricular volume of the human heart. A viable way is to use the reflection of ultrasonic waves at the myocardium to estimate the ventricular volume. Furthermore, we have observed the magnitude of the R-wave in the internal ECG signal to change with ventricular volume.
Contact Person
Collaborators
external page Prof. Volkmar Falk
external page Dr. Diane de Zélicourt