Jan Kühn

Model-based supervision of a blood pump

Abstract In this paper, we present a novel method to supervise several discrete events and continuous processes causing failures in a blood pump. These are potential hazards which regularly cause problems in intensive care routine. We propose an indicator that considers the nonlinear shear thinning flow properties of blood. Based on a threefold of physiological motivated measures, we calculate an indicator which is not only able to detect ongoing events like gas in the blood phase but also to predict upcoming events like the suction of the withdrawing cannula to the surrounding vessels wall. We present an algorithm that is embedded in a distributed 32 bit microcontroller network and holding hard real-time constraints. We were able to evaluate out algorithms in-vivo. For this algorithm we analyzed online data of more than 140 hours of animal experiments.

Decentralized safety concept for closed-loop controlled intensive care : Supervision of a blood pump during extracorporeal circulation

Abstract: This paper presents a decentralized safety concept for networked intensive care setups, for which a decentralized network of sensors and actuators is realized by embedded microcontroller nodes. It is evaluated for up to eleven medical devices in a setup for automated acute respiratory distress syndrome (ARDS) therapy. In this contribution we highlight a blood pump supervision as exemplary safety measure, which allows a reliable bubble detection in an extracorporeal blood circulation. The approach is validated with data of animal experiments including 35 bubbles with a size between 0.05 and 0.3 ml. All 18 bubbles with a size down to 0.15 ml are successfully detected. By using hidden Markov models (HMMs) as statistical method the number of necessary sensors can be reduced by two pressure sensors.

Physiological closed-loop control of mechanical ventilation and extracorporeal membrane oxygenation.

Abstract A new concept is presented for cooperative automation of mechanical ventilation and extracorporeal membrane oxygenation (ECMO) therapy for treatment of acute respiratory distress syndrome (ARDS). While mechanical ventilation is continuously optimized to promote lung protection, extracorporeal gas transfer rates are simultaneously adjusted to control oxygen supply and carbon dioxide removal using a robust patient-in-the-loop control system. In addition, the cooperative therapy management uses higher-level algorithms to adjust both therapeutic approaches. The controller synthesis is derived based on the introduced objectives, the experimental setup and the uncertain models. Finally, the autonomous ARDS therapy system capabilities are demonstrated and discussed based on in vivo data from animal experiments.

Fault Identification in a Blood Pump Using Neural Networks

This paper compares two fault identification implementations based on a neural network and a model based approach. Our worked example is the detection of gas bubbles in the pump head of a centrifugal blood pump. We focus on algorithms applicable on minimal sensor data with a reasonable implementation effort. The approaches were restricted to the desired blood flow and the measured rotational speed of the pump. We evaluated both implementations with data from an ECMO system.

Mock loop for bubble generation in a centrifugal blood pump for fault simulation

Abstract In extracorporeal blood circulation intensive care treatments, the occurrence of gas within the circulation is one known major hazard. This gas volume can cause severe harm to the patient like infarctions. Consequently, within risk assessment for these treatments gas bubbles are usually addressed by either constructive or signal based approaches. All signal-based approaches do have in common that they need a sufficient amount of data to be parameterized. These data can only be acquired in animal trials or laboratory experiments, as they could result in harm to patients. Hence, we designed a mock loop, which is automatically able to create annotated data of gas bubbles injected into an extracorporeal circulation. We were able to run this setup with a periodicity of 15 seconds, which results in 240 annotated measurements per hour. For the evaluation, we created 1095 bubbles of varying sizes (0.3 to 0.5 ml). The elaborated setup enables us to produce a great amount of annotated data, which is shown to be comparable to manually generated data in a convenient and fully automated manner.

Case studies on automated verification with slope boundaries for block diagrams

Abstract Model-based development is increasingly used in embedded systems, which are often deployed in a safety-critical environment. Verification techniques, supporting the development process can not only increase safety, but also help to speed up the process. In many cases models are designed with block diagrams, assisting rapid prototyping. However, automated verification is, thereafter, often applied to the resulting code. Instead of focusing on code, we apply verification techniques to models consisting of block diagrams and MATLAB code. We propose to combine a value and slope range analysis, with symbolic methods. In this way, our concept can not only prove properties in models, but also check rate requirements automatically, which arise from physical constraints of the environment. We evaluate our work in case studies from ongoing research projects. These case studies cover the domains of clinical intensive care, autonomous drone control and driver assistance. All systems are also evaluated with a commercial verification tool, highlighting benefits of the tool and our implementation.