Ultrasound image based visual servoing for moving target ablation by high intensity focused ultrasound

Abstract Background Although high intensity focused ultrasound (HIFU) is a promising technology for tumor treatment, a moving abdominal target is still a challenge in current HIFU systems. In particular, respiratory‐induced organ motion can reduce the treatment efficiency and negatively influence the treatment result. In this research, we present: (1) a methodology for integration of ultrasound (US) image based visual servoing in a HIFU system; and (2) the experimental results obtained using the developed system. Materials and methods In the visual servoing system, target motion is monitored by biplane US imaging and tracked in real time (40 Hz) by registration with a preoperative 3D model. The distance between the target and the current HIFU focal position is calculated in every US frame and a three‐axis robot physically compensates for differences. Because simultaneous HIFU irradiation disturbs US target imaging, a sophisticated interlacing strategy was constructed. Results In the experiments, respiratory‐induced organ motion was simulated in a water tank with a linear actuator and kidney‐shaped phantom model. Motion compensation with HIFU irradiation was applied to the moving phantom model. Based on the experimental results, visual servoing exhibited a motion compensation accuracy of 1.7 mm (RMS) on average. Moreover, the integrated system could make a spherical HIFU‐ablated lesion in the desired position of the respiratory‐moving phantom model. Conclusions We have demonstrated the feasibility of our US image based visual servoing technique in a HIFU system for moving target treatment.


| INTRODUCTION
High intensity focused ultrasound (HIFU) is well known as a promising non-invasive tumor treatment modality, and its applications are increasing. However, treatment for moving organs is still a challenge in currently available commercial HIFU systems due to treatment efficiency and safety. Motions of abdominal organs are caused by heartbeat, respiration, or even bowel peristalsis. 1 Any of these motions can cause a mismatch of the target position and put neighboring organs at risk during lengthy HIFU interventions. Among these organ motions, respiratory-induced motion exhibits the largest displacement, and so substantially influences the performance of HIFU treatment.
Several tracking strategies for respiratory-induced target motion have been introduced. In 2 and, 3 a 2D motion model acquired from the initial learning phase was used to correct the target position during HIFU irradiation. In MR-guided HIFU, pencil-beam navigator echoes were used for motion compensation of both the MR thermometry and the target position. 4,5 Auboiroux et al. 6 combined US imaging and MR imaging for image guidance of HIFU treatment in moving tissue.
Respiratory gating for HIFU ablation was also studied by, 7,8 and. 9 In a This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. comparison of model-based motion compensation and respiratory gating, Rijkhorst et al. 8 showed that the model-based method showed lower heating rates than gating. Auboiroux et al. 7 exploited an MRcompatible digital camera to obtain the triggers of respiratory gating and applied the method in the liver and kidney of sheep in vivo. However, in respiratory gating in HIFU ablation, it is difficult to precisely deposit the heat due to the inherent heat diffusion in tissue and the reduced overall duty cycle of the gated HIFU energy delivery. Therefore, for precise ablation of a moving target, the organ motion should be monitored in real time and continuously compensated during HIFU irradiation. Diagnostic ultrasound (US) imaging is an adequate modality for real-time monitoring of a target, is cheaper than MRI, and is easily integrated as a hardware component in HIFU systems. In, 10 US image based visual servoing was applied to track the respiratory motion of renal stones for lithotripsy. As a visual servoing technique for soft tissue motion, in our previous research, we made an artificial marker (visible in US imaging) in the protein phantom by initial HIFU irradiation and tracked the marker for motion compensation. 11 Chanel et al. presented a US-guided robotic HIFU system in, 12 which applied a speckle tracking method based on normalized 1D cross-correlation to reduce computation time. The method is still difficult to apply for real organ motion due to the out-of-plane motion.
In this paper, we present a US image based visual servoing technology integrated into a HIFU system for moving target ablation.
Briefly, the organ motion is monitored by biplane US imaging and the target position is calculated in real time by registration with a preoperative 3D model. The distance between the target and the current HIFU focal position becomes an error that must be compensated in the visual servoing system. In experiments, organ motion was simu-  The PID gains are fixed in both cases. Inc., Tokyo, Japan). Since the scattered HIFU signals cause serious noise artifacts in US imaging, we need to construct a sophisticated timeline schedule in order to interlace two US sources. In particular, we need to consider the target tracking speed to secure the maximum efficiency of HIFU irradiation. A 50% duty ratio of HIFU irradiation was selected as the best setting. Therefore, we skip biplane US imaging alternately and irradiate HIFU during that time.

| Phantom model for experiment
For the experiments of US image based visual servoing with HIFU ablation we need a phantom model that enables US imaging as well as lesion formation by HIFU thermal ablation. In this research, we developed a new phantom model combining a rectangular tissue-mimicking phantom and a kidney-shaped protein gel made of 7% bovine serum albumin (BSA). 14 The protein gel was formed as the preoperative 3D kidney model and sandwiched between the tissue phantoms, as shown in Figure 4. The surface between the protein gel and the tissue phantom is shown as echogenic boundaries in the US image (see Figure 1B and Figure 3B. Therefore, the boundaries are utilized for registration with the preoperative 3D kidney model so that the target position in the US image can be calculated. Moreover, because BSA gel is optically transparent, the coagulated lesion is visually identified after HIFU ablation.

| Experimental conditions
The experimental system is composed of lower and upper water tanks as shown in Figure 5. The water in both tanks is degassed and circulated during the experiments.

| Visual servoing and error analysis
To evaluate the performance of visual servoing, we recorded the    Table 1. As shown by the results, the visual servoing system can compensate for the respiratory motion of the target with an error of 1.73 mm (E Y-fr , RMS) with HIFU irradiation.

| HIFU ablation for the moving target
After the visual servoing, we checked whether  To improve E Y-fr , therefore, we need to reduce E Y-f d and E f d -fr . E Y-f d , which depends on the speed of the target tracking, is reduced by adjusting the field of view of the US imaging or by reducing the calculation time. E f d -fr depends on the hardware performance of the robot system. Increasing the control gain will decrease this error. However, the higher values of PID gain will cause a mechanical oscillation, which can increase both E Y-f d and E f d -fr . Without mechanical servoing, the electronic HIFU beam steering will reduce this error. However, the tracking range that the system can cover will be smaller. Because these errors in visual servoing for the HIFU system are mutually dependent, we need to carefully consider their relationships when changing any feature of a system.

| HIFU duty ratio in the US-guided visual servoing system
In the experimental system, US imaging and HIFU irradiation were interlaced, because the scattered HIFU signals can be a significant source of noise in US target imaging. For the interlacing, the maximum duty ratio of HIFU was set to 50% to secure the maximum US imaging speed. A higher HIFU duty ratio reduces the US imaging speed and decreases tracking accuracy, which lowers the efficiency of HIFU treatment. Therefore, 50% of the HIFU duty ratio is the optimal for current US-guided HIFU systems. However, 15 and 16 proposed a method to achieve a 100% HIFU duty ratio with US imaging. The basic idea of 15 is that the US imaging signal is encoded as a 13-bit Barker code and is simultaneously burst with the HIFU signal. The reflected HIFU signal by the imaging array can be removed by notch filtering and pulse compression. Based on that method, 16 introduced an adaptive method to determine the parameters of notch filtering, and the reported results seem to demonstrate an improvement over the previous method.

| Visual servoing induced HIFU lesion control
No current visual servoing technique can precisely compensate for the motion of a target. Therefore, in visual servoing in the HIFU system, the error always accumulates near the target location; i.e. the heat is deposited across a region, not at a sharp point. This can be visualized FIGURE 7 2D error histograms of motion compensation with HIFU irradiation (experimental time: 5 min): results of Experiment 3 in Table 1  by the 2D error histogram shown in Figure 7. Here, we found that the shape of the ablated lesion after motion compensation was usually spherical, as shown in the Figure 9, whereas normally the focal region is a long cigar shape 17 along the axis of HIFU irradiation. The larger error or longer HIFU irradiation enlarges the lesion along the direction of the target motion rather than the direction of irradiation. However, appropriate conditions could be applied intentionally to create a spherical lesion. This irradiation shape might be very useful for spherical and small targets that conventional HIFU cannot treat.

| Future improvements to US-guided robotic HIFU systems
In the developed system, the end-effector is submerged in a water tank and transmits US energy to the patient via a silicon membrane covering a bottom hole in the water tank. In clinical applications, therefore the water tank will be fixed on the patient's body and the endeffector will track the target motion in the water tank. However this design limits the range of imaging and treatment, so we are planning to redesign the end-effector with a small water bag in the next version. With the developed visual servoing system, we aimed to produce a single lesion in the moving target. Therefore servoing for three translational axes was acceptable even if the target spun during the motion. However, additional degrees of freedom should be considered when multiple thermal ablations are required in the volumetric target.

| CONCLUSION
In this paper, a US image based visual servoing method for a HIFU system was presented. We assumed the condition of respiratory motion, and the target organ was a kidney. In experiments, we built a moving organ model with a linear actuator and kidney-shaped phantom model to simulate the respiratory-induced cranio-caudal motion of the kidney. The motion was based on real human respiration data. The kidney phantom model was monitored by biplane US imaging and the target position was calculated in real time (40 Hz) by registration of a preoperative 3D model. The distance between the target and the current HIFU focal position becomes an error that needs to be compensated in the visual servoing system. Because the motion compensation is conducted during HIFU irradiation, the system must avoid the intervention of two US sources. Therefore, a sophisticated timeline strategy