Adaptive control of phase leading compensator parameters applied to respiratory motion compensation system

Abstract

PURPOSE:

This study evaluates the feasibility of our previously developed Respiratory Motion Compensation System (RMCS) combined with the Phase Lead Compensator (PLC) to eliminate system delays during the compensation of respiration-induced tumor motion. The study objective is to improve the compensation effect of RMCS and the efficay of radiation therapy to reduce its side effects to the patients.

MATERIAL AND METHODS:

In this study, LabVIEW was used to develop the proposed software for calculating real-time adaptive control parameters, combined with PLC and RMCS for the compensation of total system delay time. Experiments of respiratory motion compensation were performed using 6 pre-recorded human respiration patterns and 7 sets of different sine waves. During the experiments, a respiratory simulation device, Respiratory Motion Simulation System (RMSS), was placed on the RMCS, and the detected target motion signals by the Ultrasound Image Tracking Algorithm (UITA) were transmitted to the RMCS, and the compensation of respiration induced motion was started. Finally, the tracking error of the system is obtained by comparing the encoder signals bwtween RMSS and RMCS. The compensation efficacy is verified by the root mean squared error (RMSE) and the system compensation rate (CR).

RESULTS:

The experimental results show that the calcuated CR with the simulated respiration patterns is between 42.85% ∼3.53% and 33.76% ∼2.62% in the Right-Left (RL) and Superior-Inferior (SI), respectively, after the RMCS compensation of using the adaptive control parameters in PLC. For the compensation results of human respiration patterns, the CR is between 58.95% ∼8.56% and 62.87% ∼9.05% in RL and SI, respectively.

CONCLUSIONS:

During the respiratory motion compensation, the influence of the delay time of the entire system (RMCS+RMSS+UITA) on the compensation effect was improved by adding an adaptive control PLC, which reduces compensation error and helps improve efficacy of radiation therapy.

Keywords

Real-time adaptive control parameters respiratory motion compensation ultrasound image tracking organ motion correction

1 Introduction

When a cancer patient undergoes radiation therapy, certain locations of the tumor, such as chest and abdomen, are often affected by the patient’s movement, respiration, and heart beats, resulting in inability to accurately deliver the radiation dose to the lesion, and thus, reducing the efficacy of radiation therapy [1, 2]. Some studies indicate that tumors located in the thorax and abdomen, such as lung, liver, spleen, and pancreas, cause the tumor displacements due to respiration-induced motion. The maximum displacements in superior-inferior (SI), right-left (RL) and anterior-posterior (AP) are 52 mm, 23 mm, and 16 mm, respectively [3]. In order to solve the tumor movement caused by respiration and improve the therapeutic effect, many techniques for tumor tracking and compensation have been proposed [4 –6]. Currently real-time tracking and compensation technology is one of the best methods to track tumors and improve the treatment efficacy. Moreover, the monitoring equipment such as CT, fluoroscopy and ultrasound were used to track tumors in real time and combined with some compensation equipment to compensate the tumor motion, thereby improving the accuracy and efficacy of radiation therapy.

Real-time respiratory compensation technology has been developed for many years. Although it is possible to track and compensate for the tumor displacement, and to reduce the Planning Target Volume (PTV) during the radiation therapy, this technique is easily limited by the total signal delay time of the compensation system. Since the tracking compensation system often has a problem, system delay time, preventing the tumor motion to be compensated to the correct position in time, and thus, the compensation effect is reduced [7, 8]. To obtain a better treatment quality, many research teams have proposed different methods to solve the system delay time issue. In 2004, Sharp et al. [9] used some methods, such as linear prediction, artificial neural network (ANN), and Kalman filter to predict the tumor location. In 2014, Prall et al. [10] used an ANN to predict the next input waveforms. The results showed that the radiation dose was delivered more concentrated than when the ANN was not used. In our previous study [11], we applied a Phase Lead Compensator (PLC) for compensating the system delay caused by the ultrasonic imaging and Respiratory Motion Compensation System (RMCS), and has a good effect on the input waveform of standard sine wave; however when the PLC is applied to the input waveform of human respiratory signal, the compensation effect was not well performed due to the significant changes in human respiration patterns [12].

In this study, the diaphragm was tracked by using ultrasound images because it is highly correlated with tumor movement near the lungs and liver [13 –15]. Furthermore, the tracking technique was also used in combination with our previously developed Ultrasound Image Tracking Algorithm (UITA) [16] and RMCS [17] for compensation experiments. In order to reduce the impact of system delay time during the radiation therapy, a PLC was added to the RMCS in this study and an adaptive control method of the PLC was proposed to make it suitable for various sudden changes in respiratory signals and other conditions, and thus, further improves the radiation therapy effect.

2 Materials and methods

2.1 Experimental apparatus

The purpose of this study was to improve the influence of delay time on the compensation effect caused by the imaging and compensation system during respiratory motion compensation. The experimental equipment includes ultrasonic imaging system (UF-4000 Fukuda Denshi, BSUS20-32 C Broadsound Corporation), with no radiation exposure and a non-invasive method, compare to the X-ray tracking systems, RMCS, Respiratory Motion Simulation System (RMSS), motion control I/O card (PCI-7344, National Instruments), video capturing card (CE310B, AverMedia) and diaphragm phantom. Both the RMCS and the RMSS are composed of an Acrylic sheet, a linear slide and a ball-screw and DC and servo motors were used as well. The RMCS is a couch tracking device and it comprised a couch, motor (BA1SGM7A-04AFA61, Yaskawa), ball screw (KK60, HIWIN), linear ball slide (EGH15CA, HIWIN), and acrylic plate. The couch is driven by the motor to offset the diaphragm motion in real-time. A rubber belt of PVC material was used to simulate the diaphragm, which is attached to the edge of a basin, fill with agarose to simulate human tissue. Finally, the ultrasound coupling gel and a small amount of water were applied to the surface of the agarose to enhance the conductivity of the ultrasonic wave. The experimental setup is shown in Fig. 1.

Fig.1

Experimental setup of respiratory motion compensation, A: ultrasonic imaging system, B: RMSS, C: RMCS, D: setup of diaphragm phantom (agarose, universal chuck, ultrasonic probe).

Software used in the experiment included our previously developed algorithm UITA and the LabVIEW graphical programming that controls RMSS and RMCS. The main function of UITA (written in Visual C) is to track the diaphragm phantom motion in real-time, and it could capture the respiration signals from ultrasound images to analyze the displacement of chosen pixel. Then, the threshold image processing was performed to eliminate the image other than the target, diaphragm phantom. In UITA, a small white square box was designed to select the target to be tracked [16]. Finally, the displacement of the small white square box in the SI and RL directions is output to the RMCS as the target motion signals, and the ultrasound image captured by UITA is shown in Fig. 2. The LabVIEW controlling program is mainly used to drive the motors of the RMSS and RMCS and the encoders of each motor to monitor the position of the compensation couch, and display the real-time data of the compensation system on a designed human-machine interface for convenient operation and monitoring. In this study, the adaptive control parameters in PLC were also completed by LabVIEW software and combined with the RMCS control program to eliminate system delay time.

Fig.2

UITA captured images, (a) diaphragm phantom image captured by UITA, (b) UITA output signals (motion signals of the diaphragm phantom).

2.2 Experimental respiratory signal patterns

In order to verify the feasibility of compensation experiment, in this study 7 sets of simulations and 6 groups of human respiratory signals were used. The simulated breathing signals were referenced from Dynamic Phantom model 008PL manufactured by CIRS (Computerized Imaging Reference Systems Inc.) with their 3 breathing modes, 1-2cos4(t), 1-2cos6(t) and shark-fin respiratory patterns [18] as shown in Fig. 3. Moreover, the 4 sine wave based simulated respiratory signals are shown in Fig. 4. The human respiratory signals were obtained by using UITA with LabVIEW to capture the diaphragmatic respiratory motion signals from 6 volunteers, as shown in Fig. 5. The ultrasonic probe was placed on the abdomen of the human body to observe the diaphragm movement, and then the captured image was instantly transmitted to the computer through the image capturing card. The UITA was used to track the position of diaphragm in real time, and the LabVIEW was used to record the real-time motion of the diaphragm.

Fig.3

Simulated breathing signals of the CIRS phantom (10 seconds), (a) simulated breathing signal with the 1-2cos4(t) function, (b) simulated breathing signal with the 1-2cos6(t) function, and (c) simulated breathing signal with the shark-fin pattern.

Fig.4

4 types of simulated breathing signals based on sine waves, (a) simulated breathing signals with baseline shift characteristics, (b) simulated breathing signals with amplitude variation characteristics, (c) simulated breathing signals with both baseline shift and amplitude variation characteristics, (d) simulated breathing signals with both amplitude and frequency variation characteristics.

Fig.5

Captured diaphragmatic motion signals of 6 volunteers.

2.3 Design of phase lead compensator (PLC)

During the respiratory motion compensation experiment, the ultrasonic imaging system combined with RMCS generates a significant delay time, and the overall compensation rate (CR) was greatly reduced. Therefore, this study uses a PLC to eliminate the delay time generated by the system, and expects to greatly increase the CR. The formula of the PLC used in the experiment is shown in Equation (1). $(\frac{s + 1}{s + a}) k$ (1) where “a’’ is the term of time variable calculated from the system delay time and the frequency of the input signal. After adding this timing term “a’’ into the control system, the phase can be advanced and the delay time caused by the transmission, calculation and analysis of the system can be eliminated. However, this will change the magnification of the input signal, causing the amplitudes of the input and output signals to be unequal, so the gain factor “k’’ must be added to compensate for the difference in amplitude. Equation (1) was derived from the phase lead compensator theory, which introduces a pole-zero pair into the open loop transfer function. The transfer function can be written in the Laplace domain as: $(\frac{s + 1}{s + a}) k = (\frac{s}{s + a} + \frac{1}{s + a}) k$ (2) where ‘s+1’ is the input to the compensator, ‘s+a’ is the output term, ‘s’ is the complex Laplace transform variable, ‘1’ is the zero frequency, ‘a’ is the pole frequency and ‘k’ is the gain factor.

2.4 Automatic detection of real-time respiration frequency

The important parameters “a’’ and “k’’ in PLC change with the frequency of the input breathing signal. Therefore, the ability to perform the real-time detection of the respiration in compensation experiments is a major factor in improving the efficacy of radiation therapy. In order to perform automatic detection of the instantaneous respiratory frequency, this study developed a monitoring module for analyzing respiratory signals. The breathing signal acquisition is performed through the captured ultrasound image and analyzed by the UITA. After applying the image binarization method to remove the imaging noise, the displacement of the brightest point in the ultrasound imaging of diaphragm phantom motion can be tracked. In this study, the ultrasound image is processed with the image binarization method and after the binarization threshold was setup, it filters according to the grayscale value of each pixel on the image, so that the image shows a clear black-and-white effect, and thus, highlighting the outline of the target. The analysis process of the program first filters the input breathing signal to remove the noise, and then calculates the average amplitude value of the input signal in the certain period of time in the past, and thus, it can establish the reference line of the breathing signal in order to calculate and predict the respiration pattern signals for the next cycle. When it is actually applied in clinical trial, in order to increase the correctness of the proposed algorithm, it is necessary to carry out the pre-operation for a certain period of time before starting the analysis.

2.5 System delay time

In this study, the respiratory motion compensation experiment requires certain procedures such as ultrasonic imaging, UITA analysis, control parameters calculation, motor drive and data transmission, which all contribute to the delay time. The total delay time of the compensation system is 500 ms, where the main delay is about 300 ms and 150 ms in the ultrasonic image processing and control program respectively, and the rest of the delay time is caused by data transmission.

2.6 Experiments of respiratory motion compensation

When performing a respiratory motion compensation experiment, the RMSS was placed on the RMCS and the RMSS is driven with a pre-prepared input breathing signal to move the diaphragm phantom. Ultrasonic images were used to track the instantaneous motion of the diaphragm phantom, and UITA was used to analyze the captured ultrasound images to calculate the displacement values of the specific position of the diaphragm phantom in the SI and RL directions. The displacement values were transmitted to the RMCS for real-time respiratory motion compensation, and the schematic diagram of the respiratory motion compensation experiment is shown in Fig. 6. In the RMCS, the data input by UITA will be analyzed after the filtering of the respiratory signals. After calculating the appropriate parameters for control system, the input respiratory signals were subjected to the calculation of the adaptive control compensation, and then output to the motor-driven RMCS for respiratory motion compensation experiments. Finally, the encoder value of each motor will be feedback to the control system to increase the compensation accuracy.

Fig.6

Schematic diagram of respiratory motion compensation experiment, A: RMCS B: RMSS C: diaphragm phantom D: ultrasonic probe.

2.7 Compensation rate with adaptive control phase lead compensator (PLC)

In this study, the positioning signals of the RMSS and RMCS were collected by using the encoder of the motor. The RMSS encoder signals represent the motion of the diaphragm phantom, while the RMCS encoder signals represent the position of the compensation couch. The tracking error of the respiratory motion compensation experiment is defined as the difference of two motor encoders between RMSS and RMCS, i.e. the residual diaphragm motion signals. In this study, the experimental residual diaphragm motion signals were analyzed by the root mean squared error (RMSE) to verify the system tracking error during respiratory motion compensation experiments. In order to compare the compensation effects of using the adaptive control PLC, this experiment uses the RMSE (RMSE_RMSS) of the diaphragm motion signals (RMSS encoder) and the RMSE (RMSE_RMCS) of the residual diaphragm motion signals (RMCS encoder) to verify the CR, the calculation formula is shown as Equation (3). $CR (%) = (1 - \frac{{RMSE}_{RMCS}}{{RMSE}_{RMSS}}) \times 100 .$ (3)

3 Results

In this experiment, 7 sets of simulation and 6 groups of human respiratory signals were input to RMCS, and the results of the respiratory motion compensation experiments were compared with three different compensation methods: compensation without using PLC, compensation with constant parameters in PLC, and compensation with adaptive control parameters in PLC. Figures 7 and 8 show the respiratory motion signals and residual motion signals in different input respiration pattern signals. It can be seen that after using a PLC for compensation experiments, the tracking error has a tendency to decrease regardless of whether using the constant parameters or the adaptive control parameters in PLC. However, our experiment results show that the use of adaptive control parameters in PLC can further reduce the residual diaphragm motion signals. Table 1 shows the calculated RMSE of the residual diaphragm respiratory motion signals in the directions of RL and SI under three different compensation methods. After the respiratory motion compensation and the calculation of the RMSE, the CR without using the PLC can be calculated in the directions of RL and SI is between 0.81% ∼49.37% and 1.16% ∼53.93% respectively. The calculated CR in the directions of RL and SI under compensation with constant parameters in PLC is between 24.13% ∼83.58% and 20.05% ∼81.95% respectively. However, the CR in the directions of RL and SI with using the adaptive control parameters in PLC is between 42.85% ∼83.53% and 33.76% ∼82.62% respectively.

Fig.7

The input respiratory motion signal of pattern #D and its residual motion signal after compensation. Figure (a) (c) and (e) are in RL direction, and (b) (d) and (f) are in SI direction. (a) (b) represent the compensation without PLC; (c) (d) represent compensation with constant parameters in PLC, and (e) (f) represent compensation with adaptive control parameters in PLC.

Fig.8

The input respiratory motion signal of pattern #2 and its residual motion signal after compensation. Figure (a) (c) and (e) are in RL direction, and (b) (d) and (f) are in SI direction. (a) (b) represent the compensation without PLC; (c) (d) represent compensation with constant parameters in PLC, and (e) (f) represent compensation with adaptive control parameters in PLC.

Table 1

Root mean squared error (RMSE) by three different types of compensation methods under different input respiration pattern signals

Pattern	Uncompensated RMSE (mm)		Compensated without PLC RMSE (mm)		Compensated with constant parameters PLC RMSE (mm)		Compensated with adaptive control PLC RMSE (mm)
	RL	SI	RL	SI	RL	SI	RL	SI
A	3.82	7.64	3.79	7.52	2.26	4.00	1.74	3.98
B	4.00	8.00	3.96	7.91	3.04	6.40	2.29	5.30
C	6.80	13.60	5.63	11.28	1.99	4.06	2.12	4.63
D	7.60	15.20	4.45	8.59	1.86	3.72	1.25	2.64
E	5.89	11.78	4.02	7.71	0.97	2.13	1.06	2.06
F	6.53	13.05	4.10	7.94	1.80	3.28	1.66	2.82
G	6.07	12.13	4.84	9.76	2.54	5.20	1.72	3.58
#1	3.77	10.37	2.19	5.43	1.22	3.42	0.97	2.45
#2	4.26	11.70	2.17	5.39	1.54	4.14	0.91	2.45
#3	4.52	12.43	2.29	5.73	1.31	3.94	1.34	2.91
#4	2.02	5.54	1.90	4.83	0.95	2.37	0.83	2.06
#5	3.93	10.79	2.08	5.56	1.29	3.38	1.11	2.80
#6	3.36	9.24	2.28	5.76	0.89	2.65	1.03	2.29

Abbreviations: SI: superior-inferior, RL: right-left, RMSE: root mean squared error, CR: compensation rate, Pattern A∼G are simulated respiratory signals, Pattern #1∼6 are human respiratory motion signals.

The measured CR value of 20.05% is one of the two data after compensation less than 60%, out of total 13 input respiration signals. The characteristics of these two respiratory signals are that the respiration pattern is relatively irregular with the frequency changes rapidly, and the amplitude transition has a large change. Therefore, the developed adaptive control algorithm in this paper cannot follow up with these rapid changes. The remaining 11 CR values are all higher than 60%. In addition, the physical meaning of CR value can be defined as: when CR is equal to 100% means the respiratory motion is fully compensated, whereas CR equals to 0%, indicating no compensation at all.

By performing a statistical analysis of the residual motion signal of the diaphragm, a whisker plot can be obtained as shown in Fig. 9. The distribution of the compensation error exhibited by the graph was used to evaluate the accuracy of respiratory compensation and concentration of errors. Figure 9 reveals that the compensation without using PLC shows that the median of the error is very high as well as the range of Q1∼Q3, which means that the compensation error is not only large but also not concentrated. In contrast, the error after using PLC is obviously reduced, and under the human respiratory signals, the use of adaptive control parameters in PLC has better compensation effect than the constant parameters in PLC.

Fig.9

The error concentration range (whisker plot) of the respiratory signal compensated by three different types of compensation methods under different input respiration pattern signals. Figure (a) (c) are in RL direction and (b) (d) are in SI direction. Pattern #A∼G are simulated respiratory signals and Pattern #1∼6 are human respiratory motion signals.

Table 2 shows that the compensation effect of pattern #4 is the best among all the human breathing signals. The compensation error without using PLC in the directions of RL and SI is 1.85 (1.22 ∼ 2.28) and 4.73 (3.2 ∼ 5.79), respectively. The compensation error using constant parameters in PLC in the directions of RL and SI is 0.73 (0.35 ∼ 1.18) and 1.87 (0.88 ∼ 2.91), respectively, and the compensation error using adaptive control parameters in PLC is 0.58 (0.27 ∼ 1) and 1.43 (0.69 ∼ 2.49) in RL and SI, respectively.

Table 2

The error concentration range of the respiratory signal compensated by three different types of compensation methods under different input respiration pattern signals

Pattern	Compensated without PLC RMSE (mm)		Compensated with constant parameters PLC RMSE (mm)		Compensated with adaptive control PLC RMSE (mm)
	RL	SI	RL	SI	RL	SI
A	2.9 (0.95 ∼ 5.12)	6.07 (1.86 ∼ 10.25)	1.54 (0.7 ∼ 2.71)	2.86 (1.68 ∼ 4.67)	1.33 (0.7 ∼ 2.03)	3.36 (1.7 ∼ 4.7)
B	2.5 (0.64 ∼ 5.4)	5.17 (1.02 ∼ 10.93)	1.76 (0.98 ∼ 3.46)	4.2 (2.49 ∼ 6.96)	1.87 (1.09 ∼ 2.4)	4.51 (2.33 ∼ 5.49)
C	5.08 (3.32 ∼ 7.03)	10.09 (6.4 ∼ 14.03)	1.46 (0.74 ∼ 2.28)	2.99 (1.57 ∼ 4.78)	1.4 (0.7 ∼ 2.35)	3.1 (1.44 ∼5.49)
D	4.37 (2.31 ∼ 5.69)	8.43 (4.59 ∼ 10.96)	1.31 (0.59 ∼ 2.27)	2.47 (1.17 ∼ 4.38)	0.87 (0.42 ∼ 1.46)	1.89 (0.97 ∼ 3.12)
E	3.9 (2.07 ∼ 5.07)	7.59 (4.28 ∼ 9.67)	0.71 (0.37 ∼ 1.17)	1.65 (0.91 ∼ 2.49)	0.8 (0.38 ∼ 1.28)	1.7 (0.83 ∼ 2.49)
F	4.03 (2.03 ∼ 5.33)	7.78 (4.09 ∼ 10.12)	1.37 (0.71 ∼ 2.17)	2.41 (1.2 ∼ 3.86)	1.3 (0.59 ∼ 2.03)	1.99 (0.94 ∼ 3.41)
G	4.64 (2.32 ∼ 6.02)	9.22 (4.83 ∼ 12.14)	1.53 (0.76 ∼ 2.75)	3.07 (1.44 ∼ 5.65)	1.09 (0.5 ∼ 1.85)	2.31 (1.09 ∼ 4.04)
#1	2.01 (1.37 ∼ 2.62)	5.06 (3.5 ∼ 6.51)	0.71 (0.33 ∼ 1.37)	1.97 (0.94 ∼ 3.89)	0.65 (0.32 ∼ 1.1)	1.51 (0.72 ∼ 2.77)
#2	1.92 (1.03 ∼ 2.72)	4.99 (2.82 ∼ 6.79)	1.06 (0.47 ∼ 1.84)	2.72 (1.24 ∼ 4.95)	0.62 (0.28 ∼ 1.08)	1.76 (0.89 ∼ 2.94)
#3	1.98 (1.12 ∼ 2.86)	5.06 (3.07 ∼ 7.16)	0.99 (0.53 ∼ 1.6)	2.95 (1.29 ∼ 4.99)	0.97 (0.51 ∼ 1.6)	1.97 (0.96 ∼ 3.43)
#4	1.85 (1.22 ∼ 2.28)	4.73 (3.2 ∼ 5.79)	0.73 (0.35 ∼ 1.18)	1.87 (0.88 ∼ 2.91)	0.58 (0.27 ∼ 1)	1.43 (0.69 ∼ 2.49)
#5	1.91 (0.98 ∼ 2.49)	5.12 (2.8 ∼ 6.78)	0.93 (0.41 ∼ 1.66)	2.41 (1.18 ∼ 4.48)	0.51 (0.23 ∼ 1.01)	1.38 (0.64 ∼ 2.69)
#6	2.19 (1.4 ∼ 2.8)	5.74 (3.83 ∼ 7.19)	0.67 (0.35 ∼ 1.05)	1.92 (0.88 ∼ 3.24)	0.74 (0.35 ∼ 1.27)	1.56 (0.72 ∼ 2.69)

4 Discussion

In this study, after using the adaptive control PLC for respiratory motion compensation, a smaller compensation error was obtained, compared with our previous compensation method. However, some simulated breathing pattern signals can’t achieve good compensating results in the use of the adaptive control PLC, because the amplitude of the simulated breathing signals used in this experiment is generally larger than that of the human respiratory signals. Moreover, the slope of the simulated signals changes more obviously, causing the compensating system to oscillate, and thus, the compensation error is worse than that of the constant parameter PLC. However, the human respiratory pattern signals have a good compensation result despite the unstable waveform, mainly because the respiratory frequency of normal person is low. It can be found from Fig. 9 that most of the respiratory pattern signals with the adaptive control PLC compensation, the error is more concentrated (Q1 ∼ Q3 spacing is smaller). That means that when this technique was applied to clinical radiotherapy, the radiation will be more concentrated to the target, and if it is without using the PLC compensation, the error is more dispersed (larger spacing between Q1 and Q3), the radiation will be more dispersed to the target. Therefore, after the respiratory motion compensation with the adaptive control PLC, it improves the accuracy of radiation delivered to the lesion, and meanwhile reducing the dose damage of the tissue near the tumor.

In this study, the RMSE of the simulated respiratory motion signals in both RL and SI directions were below 5.3 mm by applying adaptive control PLC with RMCS for respiratory motion compensation experiments. The RMSE of human respiratory motion signals in both RL and SI directions were all below 2.91 mm. Compared with other research teams, Lee et al. [19] used the 3D camera of the AlignRT^® system to track the target and compensated the respiratory motion with the CBTS three-dimensional respiratory motion compensation couch. The experimental results under load conditions show that the maximum of RMSE was 7.54 mm. D’Souza et al. [20] used the HexapodTM treatment couch to compensate the respiration-induced motion, and their results showed that the RMSE was below 3 mm using the step signals. Although their RMSE results are not much different from ours, their input signals are step signals, which is quite different from the human respiratory signals. Haas et al. [21] developed a patient support system (PSS) to enable internal tumors to be fixed under radiation during radiotherapy and to track the target with a digital camera. The experimental results show the maximum RMSE in the RL and SI directions after compensation is 5.21 mm and 4.46 mm, respectively. In this study, the tracking method is different from other studies. We use ultrasound images to directly observe the movement of organs inside the body, which can show the real-time movement of organs in vivo compared with the in vitro tracking method.

Although this study has better compensation results than in the past, there are still some limitations. In order to analyze the input human respiratory signals, the patient must be allowed to rest for 30 seconds before compensation, to stabilize the breathing [22]. During this time (30 seconds) period, our proposed algorithm was used to analyze the respiratory signals, in order to determine the proper adaptive control parameters in PLC. In this study, the use of PLC affects the overall control system and might cause a slight instability of the system. Although it does not have much impact on the results, the human comfort must be considered in order to perform the clinical human trials. In addition, it was found that an excessively fast respiratory frequency (greater than 20 times per minute) would cause instability of the RMCS because the respiratory frequency would affect the parameters of the PLC. A slow respiratory frequency (less than 10 times per minute) will prevent the system from accurately capturing the respiratory frequency resulting in incorrect PLC parameters being calculated. However, considering the general clinical situation, the typical respiratory frequency of adults is about 12 to 18 times per minute, and the elderly people are about 12 to 28 times per minute [23, 24]. Although this study is not fully applicable to all patients, it can still meet the needs of most patients. In addition, the delay time in our system is not shorter than the other compensation systems used in the clinical practice. However, the main purpose of this work is to study the parameters of the developed Adaptive Control of Phase Leading Compensator. Therefore, regardless of the delay time is large or small, the proposed real-time adaptive control algorithm in this study can improve the impact of delay time on the compensation effect.

Moreover, the ultrasound image tracking system was currently used to detect the diaphragm motion, as an observational alternative to tumors. Although the respiration-induced diaphragm motion is highly correlated with lung and liver tumors, the amplitude is not exactly the same. Thus, there is still some difficulty in using it in clinical practice. In the future, it is necessary to develop a real-time motion conversion model that can calculate the displacement of the diaphragm into a tumor displacement, so that the tumor motion compensation can be substantially performed. In addition, this study only considers the compensation in RL and SI directions and does not consider the AP direction. If the linear accelerator (LINAC) is not perpendicular to the patient’s coronal plane during the radiation therapy, the displacement of the AP caused by the respiratory motion affects the efficacy of the radiotherapy. Therefore, in the future work of this study, we should consider to develop a three-axis respiratory motion compensation system.

Although applying the adaptive control PLC to the RMCS has a better compensation result and can improve the efficacy of the radiotherapy, adding the adaptive control PLC also reduces the stability of the RMCS. In the future, a PID controller can be added in the control system to enhance the system response (system response is faster), or replace the current control system with an embedded system to increase the sampling rate. If the stability of the entire system can be increased, it can also improve the patient comfort during treatment, while also improving the efficacy of radiation therapy.

5 Conclusions

This study proposes a method for adaptively controlling PLC applied to RMCS and combined with the UITA for real-time respiratory motion tracking and compensation. We compared the differences between three compensation results, without using PLC, constant parameters in PLC and adaptive control parameters in PLC. The experimental results show that it is feasible to apply the adaptive control PLC to the RMCS, which can not only obtain lower compensation error and higher CR, but also improve the efficacy of radiotherapy. Using ultrasound images for tracking targets is a non-invasive, non-radiative method to obtain the real-time internal target motion. The proposed method used in this study can calculate the real-time displacement of the internal target after combining with UITA, and compensate the respiratory motion with RMCS, which can improve the respiration-induced tumor movement in radiation therapy and reduce the dose damage of surrounding normal tissue. In the future, the system will be applied to the clinical human trials. The stability of the entire system and the method of ultrasound monitoring tumor movement must be further explored in future research.

Footnotes

Acknowledgments

This work was supported by the National Taipei University of Technology and Taipei Medical University Hospital under Contract USTP-NTUT-TMU-108-03. The authors would like to express their appreciation to the Taipei Medical University Hospital, Taiwan for providing the financial and facilities support for this study. The ethical approval is approved by the Taipei Medical University Hospital under the reference number: IRB 201501050.

References

Jiang

, Radiotherapy of mobile tumors, Semin. Radiat. Oncol. 16(4) (2006), 239–248.

, Ma

, Chen

, Tang

, Zhang

, 4DCT and CBCT based PTV margin in stereotactic body radiotherapy (SBRT) of non-small cell lung tumor adhered to chest wall or diaphragm, Radiat Oncol 11(1) (2016), 152 (9pp).

Wysocka

, Kassam

, Lockwood

, et al., Interfraction and respiratory organ motion during conformal radiotherapy in gastric cancer, Int. J. Radiat. Oncol. Biol. Phys. 77(1) (2010), 53–59.

Wilbert

, Baier

, Hermann

, et al., M. Accuracy of real-time couch tracking during 3-dimensional conformal radiation therapy, intensity modulated radiation therapy, and volumetric modulated arc therapy for prostate cancer, Int. J. Radiat. Oncol. Biol. Phys. 85(1) (2013), 237–242.

Hansen

, Ravkilde

, Worm

E.S.

, et al., Electromagnetic guided couch and multileaf collimator tracking on a TrueBeam accelerator, Med. Phys. 43(5) (2016), 2387–2398.

Lang

, Zeimetz

, Ochsner

, et al., Development and evaluation of a prototype tracking system using the treatment couch, Med. Phys. 41(2) (2014), 021720.

Fast

M.F.

, O’Shea

T.P.

, Nill

, Oelfke

, Harris

E.J.

, First evaluation of the feasibility of MLC tracking using ultrasound motion estimation, Med. Phys. 43(8) (2016), 4628–4633.

Shiinoki

, Kawamura

, Uehara

, et al., Evaluation of a combined respiratory-gating system comprising the TrueBeam linear accelerator and a new real-time tumor-tracking radiotherapy system: A preliminary study, J. Appl. Clin. Med. Phys. 17(4) (2016), 202–213.

Sharp

, Jiang

, Shimizu

, Shirato

, Prediction of respiratory tumour motion for real-time image-guided radiotherapy, Phys. Med. Biol. 49(3) (2004), 425–440.

10.

Prall

, Kaderka

, Saito

, et al., Ion beam tracking using ultrasound motion detection, Med. Phys. 41(4) (2014), 041708.

11.

Chuang

H.C.

, Hsu

H.Y.

, Nieh

S.K.

, Tien

D.C.

, The feasibility of the auto tuning respiratory compensation system with ultrasonic image tracking technique, J. Xray Sci. Technol. 23(4) (2015), 503–516.

12.

Ting

L.L.

, Chuang

H.C.

, Kuo

C.C.

, et al., Tracking and compensation of respiration pattern by an automatic compensation system, Med. Phys. 44(6) (2017), 2077–2095.

13.

Kyriakou

, McKenzie

D.R.

, Dynamic modeling of lung tumor motion during respiration, Phys. Med. Biol. 56(10) (2011), 2999–3013.

14.

Gendrin

, Furtado

, Weber

, et al., Monitoring tumor motion by real time 2D/3D registration during radiotherapy, Radiother. Oncol. 102–142(2) (2012), 274–280.

15.

Yang

, Cai

, Wang

, et al., Is Diaphragm motion a good surrogate for liver tumor motion? Int. J. Radiat. Oncol. Biol. Phys. 90(4) (2014), 952–958.

16.

Kuo

C.C.

, Chuang

H.C.

, Teng

K.T.

, et al., An autotuning respiration compensation system based on ultrasound image tracking, J. Xray Sci. Technol. 24(6) (2016), 875–892.

17.

Ting

L.L.

, Chuang

H.C.

, Liao

A.H.

, et al., Experimental verification of a two-dimensional respiratory motion compensation system with ultrasound tracking technique in radiation therapy, Phys. Med. 49 (2018), 11–18.

18.

CIRS Dynamic Platform, Model 008PL. User guide and technical information.

19.

Lee

, Chang

K.H.

, Shim

J.B.

, et al., Evaluation of mechanical accuracy for couch-based tracking system (CBTS), J. Appl. Clin. Med. Phys. 13(6) (2012), 157–169.

20.

D’Souza

, McAvoy

, An analysis of the treatment couch and control system dynamics for respiration-induced motion compensation, Med. Phys. 33(12) (2006), 4701–4709.

21.

Haas

O.C.

, Skworcow

, Paluszczyszyn

, et al., Couch-based motion compensation: Modelling, simulation and real-time experiments, Phys. Med. Biol. 57(18) (2012), 5787–5807.

22.

Kuo

C.Y.

, Chuang

H.C.

, Zhou

Y.L.

, et al., Application of ultrasound image tracking algorithm for real-time diaphragmatic excursion measurement, Journal of Medical and Biological Engineering 38(4) (2018), 678–684.

23.

Barrett

, Barman

, Boitano

, Brooks

, Ganong’s Review of Medical Physiology 24th Edition. New York: McGraw-Hill Medical Publishing Division (2012), p. 619.

24.

Rodríguez-Molinero

, Narvaiza

, Ruiz

, Gálvez-Barrón

, Normal respiratory rate and peripheral blood oxygen saturation in the elderly population, J. Am. Geriatr. Soc. 61(12) (2013), 2238–2240.