A Contactless and Biocompatible Approach for 3D Active Microrobotic Targeted Drug Delivery

Abstract

As robotic tools are becoming a fundamental part of present day surgical interventions, microrobotic surgery is steadily approaching clinically-relevant scenarios. In particular, minimally invasive microrobotic targeted drug deliveries are reaching the grasp of the current state-of-the-art technology. However, clinically-relevant issues, such as lack of biocompatibility and dexterity, complicate the clinical application of the results obtained in controlled environments. Consequently, in this work we present a proof-of-concept fully contactless and biocompatible approach for active targeted delivery of a drug-model. In order to achieve full biocompatiblity and contacless actuation, magnetic fields are used for motion control, ultrasound is used for imaging, and induction heating is used for active drug-model release. The presented system is validated in a three-dimensional phantom of human vessels, performing ten trials that mimic targeted drug delivery using a drug-coated microrobot. The system is capable of closed-loop motion control with average velocity and positioning error of 0.3 mm/s and 0.4 mm, respectively. Overall, our findings suggest that the presented approach could augment the current capabilities of microrobotic tools, helping the development of clinically-relevant approaches for active in-vivo targeted drug delivery.

Keywords: magnetic actuation; microrobotics; minimally-invasive surgery; surgical robotics; targeted drug delivery; ultrasound tracking.

Figure 1

(Left) A close-up image of the used setup depicting the ultrasound probe, as well as the coils used to generate high frequency $\left(\begin{array}{|c|}\hline \mathrm{H}\\ \hline\end{array}\right)$ and quasi-static magnetic fields $\left(\begin{array}{|c|}\hline \mathrm{M}\\ \hline\end{array}\right)$. (Central inset) One of the used microrobots after 15 s of induction heating (18 mT at 126 kHz), captured using a cooled-detector thermal camera with 12 mm extension ring (FLIR Systems, Wilsonville, OR, USA). Scalebar in degree Celsius. (Right) Illustration of (sectioned) proposed microrobots navigating a blood vessel to perform targeted drug delivery.

Figure 2

The electromagnetic setup. The nine metal-core electromagnets $\left(\begin{array}{|c|}\hline \mathrm{M}\\ \hline\end{array}\right)$ are capable of generating quasi-static fields of up to 10 mT while allowing access to objects smaller than a φ160 mm sphere. A linear stage is used to move the ultrasound probe $\left(\begin{array}{|c|}\hline \mathrm{U}\\ \hline\end{array}\right)$ along the gravitational axis (z). Finally, the liquid-cooled high frequency coil $\left(\begin{array}{|c|}\hline \mathrm{H}\\ \hline\end{array}\right)$ can be seen in the right end of the image. (Top-right inset) A representative B-mode ultrasound image. The position of the microrobot is marked by a green square. Due to diffraction, artifacts, and noise the footprint of microrobot in the ultrasound image is significantly larger than its real size. (Bottom-right inset) A coated microrobot navigating the phantom of vascular vessels (outlined by the colored dashed lines).

Figure 3

Schematic depicting the approach to determine the out-of-plane component (z) in the tracking procedure. The ultrasound transducer continuously sweeps the workspace at a frequency of 1 Hz, hence providing two positions per second. The in-plane (x and y) components of the position of the microrobot are tracked using the procedure illustrated in Figure 4. The graph shows the dependence of the tracked blob size (blue line) on the position of the transducer. This dependence is used to triangulate the position of the microrobot using the point of maximum size (marked by the dashed red line).

Figure 4

Schematic of the control loop. The user provides the position reference. This is preprocessed by a fourth order filter that removes frequency components above the bandwidth of the controller and ensures continuous derivatives. This prevents the filtered reference from increasing with a dynamic that is faster than the maximum one of the controller The filtered reference is then provided to a Proportional, Integral and Derivative (PID) controller. This controller designed to minimize disturbances with frequencies higher than a decade below that of the tracker [17]. A feedforward component is added to improve the control performance. As the low-level controllers feed the currents determined by the force-to-current map, the microrobot moves. This motion is detected by the ultrasound tracking algorithm using the procedure shown in the image (combined with the approach of Figure 3). Such tracking procedure begins using a Gaussian mixture-based segmentation algorithm for background subtraction [18]. A Region Of Interest (ROI) around the estimated position of the microrobot is then selected and binarized using a variable-threshold. A dilation morphological filter is then applied to the image. Finally, the center of the largest blob is selected as the position of the microrobot. The computed position is then provided to a sampled-data observer, which provides the controller with intersample state estimations [19].

Figure 5

Representative timelapse of the experiments. In the first image, yellow, green, and red dashed lines mark the outline of the channels in the phantom. A yellow dashed line is also used to highlight the position of the microrobot in the frames. The blue arrows approximate the future trajectory of the microrobot. The microrobot starts from the end of the yellow-outlined channel, and navigates to the end of the green one, where it releases the drug-model. Finally, the microrobot returns to the starting point. Moreover, the changes in position of the ultrasound transducer, continuously sweeping through the workspace, can be noticed in the background. Please, refer to the accompanying video for the visualization of the experiment in the Supplementary Materials.

Figure 6

(Left) Representative example of the trajectory followed by a microrobot through the experiments. The blue curve shows the ultrasound tracking results. The red curve represent the same trajectory as computed offline by an optical tracker [10]. (Right) Histogram of the positioning and tracking errors over the performed ten trials. The positioning error (blue) is defined as the divergence between the filtered reference and the state as tracked by the ultrasound tracking system; i.e., the error as computed in the control loop. Conversely, the tracking error (yellow) is defined as the difference in tracked position between the ultrasound and optical (computed offline) tracker. Consequently, we can only compute the tracking error for x and z components. However, for reasons of symmetry, we expect the y component, which cannot be analyzed optically, to have similar tracking error to that of the x component.