Nonlinear Control for Biomedical Applications

Abstract

Access to adequate health care is essential for humans. However, many healthcare systems around the world are under pressure trying to answer to a growing need with limited financial and personal resources. Modern medical devices can help to reduce the work load of clinical staff and enable more sophisticated treatment options [1]. However, even though biophysiological signals and biomechanics of the human body express nonlinear behavior, many medical devices for clinical practice still rely on simplified linear equations. The purpose of this special issue is to highlight how non-linear control can contribute to innovate and improve healthcare.

The first topic that emerged from the papers in this special issue was the usage of non-linear control methods for system identification or more particular to derive more information about the human body and its interaction with medical devices [2, 3]. Zhu et al. demonstrate how an Hunt Crossley contact model with an iterated Kalman filter can be used to identify tissue properties from the contact forces between a robotic manipulator and the human body. This technique is a first promising step on the way towards realistic haptic real time feedback in robot assisted minimal invasive surgery [3]. The paper “Nonlinear control of a hybrid pneumo-hydraulic mock circuit of the cardiovascular system” by Alhajyounis et al. moves the usage of non-linear control to the test-bench. They employed the Lyapunov stability criterion to control the non-linear pneumatic part of a test bench that simulates the cardio-vascular system of the human body. In in silico test they demonstrate that they could simulate the behavior of normal, failing, and assisted cardiovascular function with high accuracy [2]. Test benches that correctly represent human anatomy, such as the one explored in the work of Alhajyounis et al., are urgently needed to reduce the need of animal testing and shorten the time to market for medical devices.

The second topic in which nonlinear control plays a crucial role is simulation based optimization of pharmacokinetic processes [4-6]. Real-time prediction of the reaction of the body on a medication scheme is especially crucial in anesthesiology where combinations of multiple drugs are used to exploit synergistic effects that improve efficiency and reduce toxicity [7]. Sandre-Hernandez et al. demonstrate that it is possible to model a system that controls for the depth of the hypnosis in a multi-drug regime. To model the complex system they apply multiple-input and multiple output predictive modeling. By utilizing an exponential cost function and solving the optimization problem with quadratic programming they arrive at a model that satisfies the control objective in a simulation based on data from 12 patients [5]. Pawloski et al. report how depth of hypnosis can be controlled using an event-based generalized predictive controller. They show that external predictors can be used to deal with non-linearity and inter and intra-patient variation. Using this novel control architecture the number of calculations needed for controlling the depth of hypnosis with an acceptable error [4]. The use of non-linear control in pharmacokinetics is not limited to anesthesia. In “Modelling and control of vascular dementia disease by exact dosing of medicines” Vidhyaa et al. describe how predictive controllers based on non-linear models that assumes links between the presence of certain proteins and disease progression can be used to control automatic dosing of drugs [6].

The third topic in which non-linear control methods can play a role in improving the current state of the art are assistive devices that interact closely with human users such as exoskeletons [8, 9]. A key design problem in these wearable assistive robots is the control of the interaction forces [10]. In “Fixed-time observer-based controller for the human–robot collaboration with interaction force estimation” Sharif Abadi et al. suggest to increase robustness of exoskeleton controllers and decrease chattering using sliding mode control to estimate the states of the human and the wearable robot. This control structure is analyzed in several simulations. The results suggest that it outperforms several conventional used controllers [8]. Along the same lines, Wang et al. propose a novel control structure that relies on estimated interaction torques to determine when to provide which level of assistance. In co-simulation they demonstrate that their method outperforms a position tracking error-based and a strength index-based impedance controllers in reducing the knee joint position error. Especially in patients with reduced muscle strength the novel controller outperformed benchmark algorithms. Experiments in potential users demonstrate that this control structure is able to distinguish between user intention and muscle weakness, which is a crucial property of assist as needed control [9].

The fourth and last topic that is covered by this special issue focuses on innovative non-linear control architectures and their suitability for real-time control of medical devices such as surgical robots or ventilators [11, 12]. Piccinelli et al. demonstrate how advanced control structures for surgical robots can improve patient safety by enforcing a software based remote center of motion and restricting the workspace within safe boundaries. The control structure is experimentally validated using a conventional surgical training tasks for minimal invasive surgery. The paper highlights the importance of ensuring safety and stability of the control loop while coping with delays that are inherent to complex technical systems [12]. Safety is also key in ventilator control, as the life of patients depends on timely and controlled delivery of pressurized air. In “Mathematical modeling of lung mechanics and pressure-controlled ventilation design for barotrauma minimization: A numerical simulation study” D'Orsi et al. investigate the use of model predictive control for pressure regulation. The models they suggest are optimized for maintaining optimal oxygen saturation while minimizing the risk of ventilator induced barotrauma [11].

The authors declare no conflicts of interest.