An autotuning respiration compensation system based on ultrasound image tracking

Abstract

The purpose of this study was to develop an ultrasound image tracking algorithm (UITA) for extracting the exact displacement of internal organs caused by respiratory motion. The program can track organ displacements in real time, and analyze the displacement signals associated with organ displacements via a respiration compensating system (RCS). The ultrasound imaging system is noninvasive and has a high spatial resolution and a high frame rate (around 32 frames/s), which reduces the radiation doses that patients receive during computed tomography and X-ray observations. This allows for the continuous noninvasive observation and compensation of organ displacements simultaneously during a radiation therapy session.

This study designed a UITA for tracking the motion of a specific target, such as the human diaphragm. Simulated diaphragm motion driven by a respiration simulation system was observed with an ultrasound imaging system, and then the induced diaphragm displacements were calculated by our proposed UITA. These signals were used to adjust the gain of the RCS so that the amplitudes of the compensation signals were close to the target movements. The inclination angle of the ultrasound probe with respect to the surface of the abdomen affects the results of ultrasound image displacement tracking. Therefore, the displacement of the phantom was verified by a LINAC with different inclination-angle settings of the ultrasound probe. The experimental results indicate that the best inclination angle of the ultrasound probe is 40 degrees, since this results in the target displacement of the ultrasound images being close to the actual target motion. The displacement signals of the tracking phantom and the opposing displacement signals created by the RCS were compared to assess the positioning accuracy of our proposed ultrasound image tracking technique combined with the RCS.

When the ultrasound probe was inclined by 40 degrees in simulated respiration experiments using sine waves, the correlation between the target displacement on the ultrasound images and the actual target displacement was around 97%, and all of the compensation rates exceeded 94% after activating the RCS. Furthermore, the diaphragm movements on the ultrasound images of three patients could be captured by our image tracking technique. The test results show that our algorithm could achieve precise point locking and tracking functions on the diaphragm. This study has demonstrated the feasibility of the proposed ultrasound image tracking technique combined with the RCS for compensating for organ displacements caused by respiratory motion.

This study has shown that the proposed ultrasound image tracking technique combined with the RCS can provide real-time compensation of respiratory motion during radiation therapy, without increasing the overall treatment time. In addition, the system has modest space requirements and is easy to operate.

Keywords

Ultrasound image real-time image tracking respiration compensating system

1 Introduction

Therapeutic instruments and methods have improved significantly due to rapid developments in modern science and technology. The improvements in radiation therapy have focused on concentrating the radiation dosage on the target tumor volume and improving the treatment effects. However, respiration-induced organ motion increases the safety margins needed around a target tumor, which represents a major problem that needs to be solved in the field of oncotherapy. Typical methods for reducing organ displacement caused by respiratory motion include increasing the planning target volume, breath-holding [1, 2], gating [3, 4], real-time tracking [5, 6], compensating for organ motion [7, 8],and image-guided radiation therapy [9, 10]. Many of these methods are problematic; for example, the breath-holding requires patients to perform breathing exercises in advance in order to improve their control of suspended breathing during the treatment. Moreover, long-term gating treatment can result in baseline drift [11], thereby requiring increased radiation doses. However, real-time tracking can be effective at compensating for the organ displacement caused by respiration [12].

Real-time tracking approaches focus on reducing the tumor displacement caused by respiratory motion during radiation therapy. Real-time compensation of the tumor displacement ensures that the tumor receives almost all of the radiation dosage during radiation therapy, thereby sparing surrounding noncancerous tissue and improving the overall treatment effect [13 –16]. Huang et al. [17] proposed a synchronous respiration tracking system to perform real-time tracking by implanting gold fiducial markers that are used to determine the instantaneous tumor positions. Shah et al. [18] developed an integrated intrafraction tumor tracking system that tracks the patient’s respiratory motion by monitoring the position of optical sensors placed on the patient. The detected tumor motion is then compensated using a dynamic couch-based motion correction system. Hiraoka et al. [19] used the Vero 4DRT system combined with a gimbaled X-ray head for dynamic tumor tracking during irradiation, which can monitor the three-dimensional (3D) position of a tumor in real time. Mao et al. [20] demonstrated a real-time 3D markerless tumor tracking system to provide direct tumor motion information for compensating the tumor motion throughout a session of radiation therapy. Depuydt et al. [21] reported on the first clinical application of the Vero SBRT gimbaled linear accelerator (LINAC) system in combination with a single fiducial marker implanted in the tumor for addressing respiration-induced tumor motion. Poels et al. [22] compared two tumor tracking systems, CyberKnife and Vero systems, with a focus on geometrical positioning accuracy for internal tumor motion. The algorithms proposed in previous studies [13 –16] are not applied to the purpose of images tracking and literatures [17 –21] are describing the use of different methods and equipment to measure the displacement of the tumor. Thus, the existing algorithms used in previous studies cannot be used to compare with the proposed algorithm (UITA) in the present study. In addition, most of cited literatures are using the invasive methods for tracking tumor. This study designed a UITA for tracking the motion of a specific target. The developed UITA is mainly using the calculated value based on the gray-scale value from the ultrasound image and its location. It is relatively easier to identify the motion of specific target and with a friendly operation interface, which allows the operator to directly lock and track the displacement of the moving target. Moreover, this study used the ultrasound images to observe the motion of the diaphragm without implanting any markers, no radiation dose, and has the characteristics of real-time imaging and tracking.

The respiration-induced tumor motion is generally maximal in the superior-inferior (SI) direction, which is close to the lower lobe of the adjacent lung [23]; it is three- to fourfold smaller in the anterior-posterior (AP) direction, and three- to tenfold smaller in the left-right (LR) direction. Keall et al. [24] also found that the tumor displacement was much larger in the SI direction than in the AP and LR directions. Several studies [25 –28] have investigated the compensation effect on the tumor displacement only in the SI direction, and developed a one-dimensional couch-based compensation system with a tumor motion detection device for compensating the tumor motion. Buzurovic et al. [29] described an active tracking and dynamic dose delivery technique that employed a robotic treatment table for real-time tumor motion compensation. Lang et al. [30] developed a prototype system for use with the Protura treatment couch for one-dimensional real-time tracking of respiratory motion. Chuang et al. [31] used an ultrasound imaging system to monitor the motion of internal organs during radiation therapy, and combined a strain gauge with the treatment couch [32, 33] to acquire the real-time breathing signal and compensate for organ displacement.

Current cancer treatments mainly involve the direct application of high-energy radiation to the tumor, and thus the exact location of an internal tumor needs to be observed using an external imaging method, such as four-dimensional computed tomography (4D CT) [34 –37], nuclear magnetic resonance imaging (MRI) [38 –40], fluoroscopy [41, 42], positron-emission tomography [43, 44], and ultrasound imaging [45]. The present study used ultrasound imaging to monitor the motion of internal organs and the displacement of the diaphragm, due to the four advantageous properties of this imaging modality: noninvasiveness, high spatial resolution, high frame rate, and the capability of real-time imaging. In addition, ultrasound imaging can be used to acquire breathing signals. Hwang et al. [46] proposed a novel respiratory motion monitoring technique using ultrasound imaging, in which the diaphragm was located using a fitting technique. Zhu et al. [47] used an additional electromagnetic body sensor to provide real-time information on human respiration, and then used a specific model to correct the measured displacements. Kim et al. [48] proposed a method for tracking moving organs based on a 3D organ model constructed using MRI or CT images. Ultrasound imaging of the target allowed the tumor position to be tracked in real time. Blanc et al. [49] proposed encasing an ultrasound probe within an extracorporeal high-intensity focused ultrasound device that aligns the imaging plane with the acoustic axis of the device. They demonstrated that 4D CT imaging can be used to track the location of a tumor in real time during treatment.

In this study, human breathing signals were captured in grayscale ultrasound images, in which objects were identified based on the intensity and contrast of the pixels. The “peak and valley method” was used in the present study [50, 51] for target tracking. Mahmood et al. [52] proposed a watershed algorithm to enhance liver ultrasound images, while Murala et al. [53] developed a novel pattern-based feature called the local mesh peak–valley edge pattern for the indexing and retrieval of biomedical images. Singh et al. [54] suggested an algorithm for segmenting the tumor in liver ultrasound images, and validated their proposed algorithm using a database of human images.

The present study developed an ultrasound image tracking algorithm (UITA) for tracking the displacements of the diaphragm in vivo. This algorithm can also be combined with our previously reported respiration compensating system (RCS), which is set up on the treatment couch, to compensate for diaphragm motion and thereby reduce the planning target volume and hence also the overall radiation dose. The proposed algorithm first captures the breathing signals from ultrasound images and then analyzes the target motion (e.g., of the diaphragm) in ultrasound images to obtain the magnitude of the target displacement. The real-time target displacement values are sent to the controller to drive the RCS so that it can compensate the target motion caused by respiration. The proposed autotuning RCS with an ultrasound image tracking technique is implemented using a small and simple device that is easy to move, which makes it suitable for setting up as part of different radiation therapy and CT systems.

The developed algorithm has been verified in different experiments involving simulated displacements driven by a respiration simulation system (RSS) with phantoms and actual human respiration signals in cooperation with the Department of Radiation Oncology, Taipei Medical University Hospital. Based on the validation results, the use of the developed ultrasound image tracking technique in combination with the RCS for compensating respiration-induced diaphragm displacements is discussed herein.

2 Materials and methods

2.1 Respiration compensating system

A basin of water was set up on the RSS containing a phantom of the diaphragm, and an ultrasound probe was used to detect the real-time displacement of this phantom. Imaging motion control software (VisSim, Visual Solutions, Westford, MA) was used to control the linear actuators, and to drive the RSS for simulating breathing signals as shown in Fig. 1. The tracking displacement data of the phantom were analyzed by the UITA developed in this study by using a compensation signal to drive the RCS and thereby compensate for the displacement driven by the RSS. The compensation speed was then calculated to assess the accuracy of the proposed UITA combined with the RCS.

The center frequency of the ultrasound traducer used in this study is 3 MHz. The higher the frequency, the better the image we can obtain for displacement estimation. However, the irradiation depth of the ultrasound becomes shallower, therefore it is not easy to clearly observe the diaphragm. Thus, the displacement estimation accuracy depends on the ultrasound image resolution, which is defined by the center frequency of the ultrasound transducer.

2.2 Respiratory motion tracking based on ultrasound image processing

In this study, the respiratory signals were captured from the ultrasound images of the diaphragmatic motion caused by respiration. In order to track the diaphragm motion, the present study performed the analysis of the brightness weight on each gray-scale value of the ultrasound image as shown in Fig. 2. The signal of the diaphragm movement can be obtained by tracking the brightest point on the diaphragm image such that: $℘ (x_{i}, y_{i}) = M ax (W (a, b))$ (1)

where, ℘ (x₀, y₀) is defined as the starting track position and ℘ (x_i, y_i) is defined as the desired brightest point and coordinates to track after the weight analysis, W is the weight analysis, and a, b represent processing parameters of the center position in the small frame region (m×m); m is the size of the small frame region. Equation (1) indicates the highest brightness point and position of the image in the m×m matrix, and this point is defined as the brightest point for tracking. Meanwhile, an m×m matrix is created by taking this point as the matrix center. The tracking of the brightest point can be continuously conducted by repeating the above calculation. Parameters, “a” and “b”, are set as follows: $a = x_{i - 1} - (\frac{m - 1}{2}) \sim x_{i - 1} + (\frac{m - 1}{2})$ (2) $b = y_{i - 1} - (\frac{m - 1}{2}) \sim y_{i - 1} + (\frac{m - 1}{2})$ (3)

where the range “a” is from $- (\frac{m - 1}{2})$ to $(\frac{m - 1}{2})$ , a total of m and the range “b” is from $- (\frac{m - 1}{2})$ to $(\frac{m - 1}{2})$ , a total of m.

However, the most important step is to perform the analysis of the brightness weight by the following equation (4) within a region of a solid black frame (n×n). $W (a, b) = ρ \times G (a, b)$ (4) $ρ = [\begin{matrix} ρ_{1, 1} & \dots & ρ_{1, n} \\ ⋮ & ⋱ & ⋮ \\ ρ_{n, 1} & \dots & ρ_{n, n} \end{matrix}]$ (5) $G = [\begin{matrix} (a - (\frac{n - 1}{2}), b - (\frac{n - 1}{2})) & \dots & (a - (\frac{n - 1}{2}), b + (\frac{n - 1}{2})) \\ ⋮ & ⋱ & ⋮ \\ (a + (\frac{n - 1}{2}), b - (\frac{n - 1}{2})) & \dots & (a + (\frac{n - 1}{2}), b + (\frac{n - 1}{2})) \end{matrix}]$ (6)

where, n is the size of the solid black frame. In equation (4), ρ is the weight matrix, and G is the gray-scale matrix with a center coordinates (a, b), where the weight matrix represents the weight matrix of the applied gray-scale values and the gray-scale matrix is extracted from the ultrasound B-mode images. The numbers in the matrix indicate the magnitude of gray-scale values of the images. Moreover, the gray-scale matrix represents the gray-scale values from the ultrasound B-mode images and the brightness matrix is defined as the result of multiplying the gray-scale matrix and the weight matrix. In equation (5), the “ρ” is defined as the weight matrix. During the gray-scale pixel imaging synthesis process, it may create several independent brightnesses due to the influence of the signal noise. If we directly track the highest brightness value, it is likely to be the brightness of the false position. Thus in this study, the weight matrix is set to a matrix of n×n (each element value is 1/n²) and multiplied by the gray-scale matrix, aimed at smoothing the image to eliminate the noise generated from the independent brightnesses.

In this study, an m×m matrix is created at the center of the start-tracking position, and a new n×n matrix is generated at the center of each element in the m×m matrix. The brightness and position of the brightest point can be obtained by comparing the brightness weight in each n×n matrix, where ρ is the weight matrix (n×n) and G is the gray-scale matrix (n×n) with center coordinates (a, b). Multiplying ρ and G (a, b) we can obtain the brightness and position of a specific brightness point, $W (a, b)$ . Thus, a total of m×m brightness points within the $W (a, b)$ are compared to find the brightest point. The size of the matrix “n” is depending on the moving velocity of tracking target within a certain period of time. The higher the moving velocity, the larger the matrix “n” is. The matrix “m” is defined as the masking matrix and its value is generally smaller than “n”. The size of the matrixes “n” and “m” are depending on the empirical values, and typical values for “n” and “m” are 9 and 3, respectively.

Ultrasound images were directly displayed on a PC screen, and a mouse could be used to select any point of interest with an ultrasound image tracking program (developed in Visual C), such as the small white block shown in Fig. 3. The proposed algorithm then automatically analyzes the brightness differences among all the pixels within the selected area. The analysis results serve as a reference for the subsequent dynamic analysis of the brightest point, allowing the real-time displacement of the brightest point to be tracked. The dynamic tracking and analysis algorithm works as follows:

The initial tracking position (small white point) is selected by using the mouse and it is selected as the center of n×n pixels (i.e., solid black frame). In this solid black frame, the brightness of each pixel is analyzed based on the calculation of the gray-scale values within the small frame (m×m pixels). The small frame moves around and searches for the highest brightness value within the solid black frame at different times, and therefore in Fig. 2 (first step), it is shown that the small frames with different colors (yellow, green, blue) are at the different search positions at different times.

Through the brightness analysis within each small frame, the highest brightness pixel can be identified in the solid black frame; for example the small blue point in Fig. 2 (second step) is considered for the new center position. Thus, a new range of n×n pixels is created with the new center (small blue point) for the next moving area (dashed frame) on the ultrasound image.

All the brightness of each pixel within the small frame in the dashed frame is evaluated again for the analysis of the brightness weight, and thus, the next center position of the image is determined according to the analysis results as shown in the Fig. 2 (third step).

The above steps (1) (2) (3) are repeated continuously for searching the new center position (small white point) on the new ultrasound images as shown in Fig. 2 (fourth step), and the motion path of this new center position is the trajectory of our tracking target.

2.2.1 Ultrasound image processing I: Point lock

One of the most important features of the proposed UITA is that a mouse can be used to select any point of interest in the ultrasound images on the computer screen, such as any point in the region containing the diaphragm. Once a specific point is selected, a tracking box (the small white block in Fig. 3) is then locked at the mouse-click position, as also shown in Fig. 3.

2.2.2 Ultrasound image processing II: Fixed-mode point tracking

The position of the diaphragm in the obtained ultrasound images changes with time due to respiratory motion, and thus the selected tracking box also moves. With the fixed-mode function in our developed algorithm, the tracking box may be fixed on the X-axis and only move on the Y-axis along with the diaphragm (i.e., the small white block only moves vertically), as shown in Fig. 4.

2.2.3 Ultrasound image processing III: Line-mode point tracking

In order to also be able to cope with irregular motion of internal organs, the present study further developed a line-mode tracking function. In this mode, the user first uses the mouse to select any two distinct points on the target (the diaphragm, in example given), and the program then automatically draws a straight white line between these points, as shown in the left image of Fig. 5. In this mode the movements of the brightest points are tracked while the tracking box is constrained to move along the specified line only. This line can be adjusted as necessary to different positions and orientations in accordance with the motions of different human organs, as shown in the right image of Fig. 5. Normally, it does require the operator to receive adequate training to make the line orthogonal to the moving diaphragm, in order to increase the displacement estimation accuracy.

2.3 Experimental methods

2.3.1 Verification of the diaphragm displacement on ultrasound imaging

The main function of the UITA is to read in the ultrasound images and track a specific point. The displacement data are then sent to the RCS as the input signal to compensate for the target motion caused by respiration. However, the ultrasound probe must be set up at a certain inclination angle with respect to the target, and the reflected acoustic wave differs depending on the density, volume, and hardness of the various objects present. The correlation between the analyzed displacement data obtained by the proposed UITA and the actual organ displacement was verified in a collaboration with Taipei Medical University Hospital. A diaphragm phantom setup in the RSS with a metal marker was irradiated by a linear accelerator (LINAC), while an ultrasound probe was set up to irradiate the same phantom, as shown in Fig. 6. This made it possible to compare the displacement of the marker detected by the ultrasound probe with the actual marker displacement caused by the LINAC. In this experiment, a sine wave with a frequency of 0.25 Hz was used as the input signal to drive the RSS so as to simulate the respiration waveform. In addition, the ultrasound probe was positioned at different inclination angles (10 degrees, 20 degrees, 30 degrees, and 40 degrees), and different RSS displacements were used (5 mm, 10 mm, 15 mm, 20 mm, 25 mm, and 30 mm).

Two sets of displacement data detected using ultrasound imaging and obtained from the LINAC were compared, which yielded the individual ultrasound image signal gains obtained for different inclination angles of the ultrasound probe. This allowed the best inclination angle of the ultrasound probe to be determined. Moreover, the obtained gain values can be used as a correction factor of the actual displacement for different inclination angles of the ultrasound probe, which means that the displacement detected in the ultrasound image will be close to the actual displacement of the organ if the pixel values are multiplied by the appropriate gain value.

2.3.2 Verification of the respiration compensating system

One major purpose of this study was to verify the compensating effect of the UITA in combination with the RCS. The RCS moves in the direction opposite to the RSS in order to compensate for the diaphragm displacement caused by the simulated breathing movement. The setup of the RSS and RCS is shown in Fig. 1, and the method used to verify the displacement was the same as that used previously [55], in order to compare the compensation rates of the proposed UITA combined with the RCS.

In this study, sine waves with an amplitude of 10 mm (corresponding to a total displacement of 20 mm) and different frequencies (0.167 Hz, 0.2 Hz, 0.25 Hz, 0.333 Hz, and 0.5 Hz) for simulating the respiration period were sent to the RCS as input signals. The movement of the phantom and diaphragm within the basin could be detected from the ultrasound probe after activating the RSS, and the computer received the ultrasound images instantaneously. The proposed UITA tracks the diaphragm displacements in real time, and these data are sent to image motion control software (VisSim, Visual Solutions). VisSim converts the displacement value into a voltage signal and then outputs this to the LINAC of the RCS. The voltage signal gain and shift on the RCS are adjusted so as to optimize the compensation rate, as shown in Fig. 7. Finally, the displacement error signals of both the RSS and RCS were compared to calculate the corresponding compensation rates and the tracking errors.

2.3.3 Real-time human respiration signal tracking

The main purpose of the human respiration signal tracking experiments performed in this study was to verify the effect of tracking the displacement of an object such as the diaphragm using the developed UITA. Clinical trials were performed after obtaining consent from 3 male participants, who had a mean age of 23 years, a mean height of 170 cm, and a mean weight of 65 kg. The tracking targets were the diaphragms of these three subjects. The diaphragm motions were monitored using ultrasound imaging and analyzed by the developed UITA. Before the experiments, the three patients were informed to change between different irregular breathing conditions, such as normal breathing (smooth and stable), rapid breathing, and deep breathing, in order to verify that our proposed algorithm can respond rapidly to different breathing conditions. The experimental setup is shown in Fig. 8.

3 Results and discussion

3.1 Ultrasound imaging displacement error analysis

It was found that when the inclination angle of the ultrasound probe was set to 10 degrees, the measured displacement data needed to be multiplied by a gain of 0.57 in order to match the actual target displacement. The signal gains for different inclination angles are presented in Table 1 and Fig. 9. The gain value is varied among people. Normally before the radiation therapy treatment, the patient will be asked to perform the CT positioning. In this study, the ultrasound transducer is setup with an angle of 40 deg and at a fixed position to obtain a clear US image during this CT positioning process. In diaphragm motion measurement, the correlation (gain) between the diaphragm displacement under CT and US can be determined. Thus, the measured displacement by the US is multiplied by a gain value to be consistent with the CT.

In this study, the ultrasound probe was placed at the last rib near the lower edge of the human chest (as shown in Fig. 8). This was because positioning the probe vertically on the chest could exert a vertical squeezing force against the human chest during respiration, which would make the patient uncomfortable. In addition, the main observation target of this study was the movement of the diaphragm in the SI direction, and the movement of the diaphragm in that direction could still be accurately measured when the ultrasound probe was tilted.

Figure 10 shows that the analyzed displacement data obtained from the ultrasound images were closest to the actual displacements of the phantom diaphragm when the inclination angle of the ultrasound probe was 40 degrees with respect to the horizontal plane, as listed in Table 2. This table also indicates that a best correspondence of 99.6% was achieved between the displacement data of ultrasound image and measured displacement data for the LINAC, demonstrating the feasibility of using the proposed UITA to track the motions of internal organs.

3.2 Verification of the respiration compensating system based on ultrasound image tracking

In this study, after starting the RCS but without using a phase-lead compensator, the measured tracking error for an input frequency of 0.5 Hz was 5.40±2.73 mm (mean±SD), and the compensation rate was 73.0±13.7%. The other measured tracking errors and compensation rates when not using a phase-lead compensator for different input frequencies are listed in Table 3. However, applying a phase-lead compensator significantly improved the tracking error and compensation rate to 1.07±0.81 mm and 94.6±4.0%, respectively. The other measured tracking errors and compensation rates when using a phase-lead compensator are listed in Table 3.

Different frequencies were used as the input respiration signals to drive the RSS during the tests to verify the efficacy of compensating respiratory displacements. However, the driving speed of the RSS strongly affects the clarity of the captured ultrasound images, with a faster speed producing more noise, resulting in reading errors of the UITA. In addition, the RCS itself suffers from an inherent signal delay problem, and so this was corrected using a phase-lead compensator. The tracking error and the compensation rate with and without a phase-lead compensator are compared in Figs. 11 and 12, respectively. These figures indicate that the tracking errors and compensation rates of the displacement signals of the phantom diaphragm at different frequencies were both improved by using a phase-lead compensator, and also that the results were fairly stable. The compensation rate increased by 8.2–21.6%, and the compensation effects all exceeded 94%. These results demonstrate the feasibility of using the ultrasound image tracking technique for automatic tumor localization and compensation.

3.3 Verification of the real-time human diaphragm movement tracking based on different respiration patterns

Real-time ultrasound images and displacement signals of the diaphragms of the three patients with different respiration patterns are shown in Fig. 13. The four different conditions of normal (smooth and stable), rapid, and deep breathing, and breath-holding (with the diaphragm motionless) of each subject are evident in the charts shown on the left-hand side of Fig. 13. The respiration signals of the first subject showed very large shifts in the baseline condition, while only minor baseline shifts were evident in the second subject. The third subject had a relatively stable respiration signal. The in vivo ultrasound images of diaphragm movement of the three subjects are all different in Fig. 13, as well as the amplitudes of the organ motions caused by respiration. Moreover, the respiration signals of the three subjects all exhibit the problem of baseline shift [11]. However, the point-tracking function for these different breathing patterns could still be determined using the UITA developed in the present study, which demonstrates that it was possible to solve the problem of baseline shifts and thereby achieve high-accuracy organ displacement tracking. Therefore, the ultrasound image tracking technique developed in this study can be used for in vivo tumor positioning and displacement tracking.

4 Conclusions

This study aimed to capture human respiration signals and observe the internal organ motion by using a novel UITA combined with an automatic tumor localization system. The results suggest that when the inclination angle of the ultrasound probe is 40 degrees, the displacement of the phantom in the ultrasound images was closest to the actual displacement. Furthermore, the measured phantom displacement using a LINAC has a best dependence of 99.6%. In addition, a phase-lead compensator can be connected to the developed ultrasound image tracking device and the RCS, which produces a compensation rate of more than 94%.

Three patients were investigated in in vivo diaphragm displacement tracking experiments, and the results showed that the developed program is able to rapidly and accurately track different diaphragm displacements. Even when the respiration patterns had various characteristics— such as during deep, normal (smooth and stable), and rapid breathing, and breath-holding— and in the presence of baseline shift, the developed program could still track the displacement of the diaphragm of the subject in real time.

Using an ultrasound imaging system with a higher resolution and a more efficient computer would further reduce the time required to analyze the ultrasound images and the signal delay time of the RCS, thereby increasing the compensation rate. This would allow physicians to use the developed UITA combined with the RCS more safely, and allowing the application of higher irradiation doses to tumors while still sparing the surrounding noncancerous tissue of patients.

Footnotes

Acknowledgments

This work was supported by the National Taipei University of Technology and Mackay Memorial Hospital under Contract NTUT-MMH-10204. The authors would like to express their appreciation to the Mackay Memorial Hospital, Taipei Taiwan for providing the financial support and facilities for this study.

References

Demonstration videos are available at the following URLs: Fixed-mode: https://drive.google.com/open?id=0B7SnX5lqshUDaHBpaUJ5RTU0cGM Line-mode: https://drive.google.com/open?id=0B7SnX5lqshUDZkMyV09WRGNXSWs

Wang

, Zhong

, Bai

, Xu

, Zhao

, Wang

, Jiang

, Shen

, Xu

and Wei

, Lung tumor reproducibility with active breath control (ABC) in image-guided radiotherapy based on cone-beam computed tomography with two registration methods, Radiotherapy and Oncology 99 (2011), 148–154.

Glide-Hurst

, Gopan

and Hugo

, Anatomic and Pathologic Variability During Radiotherapy for a Hybrid Active Breath-Hold Gating Technique, International Journal of Radiation Oncology Biology Physics 77(3) (2010), 910–917.

Nishioka

, Nishioka

, Kawahara

, Tanaka

, Hiromura

, Tomita

and Shirato

, Exhale fluctuation in respiratory-gated radiotherapy of the lung: A pitfall of respiratory gating shown in a synchronized internal/external marker recording study, Radiotherapy and Oncology 86 (2008), 69–76.

Kim

Y.-J.

, Jang

S.-J.

, Han

J.-B.

, Dong

K.-R.

, Chung

W.-K.

and Kim

S.-C.

, Motion and volume change of tumor tissue depending on patient position in liver cancer treatment with use of tomotherapy, Annals of Nuclear Energy 65 (2014), 174–180.

Nuyt

J.J.

and Bondiau

P.-Y.

, Role of 4D stereotactic radiotherapy with the cyberknife^® system in the treatment of lung cancer, European Oncology & Haematology 8(3) (2012), 189–191.

Iwata

, Tatewaki

, Inoue

, Yokota

, Baba

, Nomura

, Shibamoto

and Sato

, Single and hypofractionated stereotactic radiotherapy with CyberKnife for craniopharyngioma, Journal of Neuro-Oncology 106(3) (2012), 571–577.

Brost

, Liao

, Strobel

and Hornegger

, Respiratory motion compensation by model-based catheter tracking during EP procedures, Medical Image Analysis 14 (2010), 695–706.

Seo

, Kim

S.K.

, Kim

Y.-S.

and Choi

, Motion compensation for ultrasound thermal imaging using motion-mapped reference model: Anmouse study, IEEE Transactions on Biomedical Engineering 61(11) (2014), 2669–2678.

10.

Goyal

and Kataria

, image guidance in radiation therapy: Techniques and applications, Division of Radiation Oncology (2014), ID705604.

11.

, Hui-Jun

and Su-Jing

, Detection and evaluation on the accuracy of fiducial tracking with spine auxiliary positioning during IGRT treatment,978-1—8, Medical Imaging Physics and Engineering (ICMIPE) (2013), 51–56.

12.

Falk

, Pommer

, Keall

, Korreman

, Persson

, Poulsen

and Munck af Rosenschöld

, Motion management during IMAT treatment of mobile lung tumors-a comparison of MLC tracking and gated delivery, Medical Physics 41(10) (2014), 101707.

13.

Paluszczyszyn

, Skworcow

, Haas

, Burnham

K.J.

and Mills

J.A.

, Model predictive control for real-time tumor motion compensation in adaptive radiotherapy, IEEE Transactions on Control Systems Technology 22(2), 635–651.

14.

Yang

, Zhong

, Rong

and Xiaohu

, Real-time GPU-aided lung tumor tracking,Fourth Pacific-Rim Symposium, Image and Video Technology (PSIVT) (2010), 495–500.

15.

McGary

J.E.

, Real-time tumor tracking for four-dimensional computed tomography using SQUID magnetometers, IEEE Transactions on Magnetics 45(9) (2009), 3351–3361.

16.

and Xing

, A kernel method for real-time respiratory tumor motion estimation using external surrogates, Machine Learning and Applications and Workshops (ICMLA) 2 (2011), 206–209.

17.

Pourhomayoun

, Fowler

and Zhanpeng

, A novel method for tumor localization and tracking in radiation therapy, Signals, Systems and Computers (ASILOMAR) (2012), Conference Record of the Forty Sixth Asilomar Conference, 1637–1640.

18.

Huang

, Hossain

, Chuang

, Descovich

, Gottschalk

, Larson

and Ma

, Lung tumor motion compensation during real time respiratory tracking using gold fiducial markers, International Journal of Radiation Oncology Biology Physics 72(1) (2008), 464–465.

19.

Shah

, Malinowski

, McAvoy

T.J.

and D’Souza

W.D.

, Toward Intra-fraction tumor motion tracking using on-board imaging and dynamic couch compensation, International Journal of Radiation Oncology Biology Physics 81(2) (2011), 153.

20.

Hiraoka

, Matsuo

, Sawada

, Ueki

, Miyaba

, Nakamura

, Yano

, Kaneko

, Mizowaki

and Kokubo

, Realization of dynamic tumor tracking irradiation with real-time monitoring in lung tumor patients using a gimbaled X-ray head radiation therapy equipment, International Journal of Radiation Oncology Biology Physics 84(3) (2012), 560–561.

21.

Mao

, Kearney

, Jiang

, Yordy

and Solberg

, Real time 3D markerless lung tumor tracking, International Journal of Radiation Oncology Biology Physics 87(2, Suplement 1) (2013), 685–686.

22.

Depuydt

, Poels

, Verellen

, Engels

, Collen

, Buleteanu

, Van den Begin

, Boussaer

, Duchateau

, Gevaert

, Storme

and De Ridder

, Treating patients with real-time tumor tracking using the Vero gimbaled linac system: Implementation and first review, Radiotherapy and Oncology 112(3) (2014), 343–351.

23.

Poels

, Dhont

, Verellen

, Blanck

, Ernst

, Vandemeulebroucke

, Depuydt

, Storme

and De Ridder

, A comparison of two clinical correlation models used for real-time tumor tracking of semi-periodic motion: A focus on geometrical accuracy in lung and liver cancer patients, Radiotherapy and Oncology Available online 14 May 2015 (In Press, Corrected Proof).

24.

Kyriakou

and McKenzie

D.R.

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

25.

Keall

P.J.

, Kini

V.R.

, Vedam

S.S.

and Mohan

, Motion adaptive x-ray therapy: A feasibility study, Phys Med Biol 46 (2001), 1–10.

26.

Wilbert

, Baier

, Hermann

, Flentje

and Guckenberger

, 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.

27.

Haas

O.C.

, Skworcow

, Paluszczyszyn

, Sahih

, Ruta

and Mills

J.A.

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

28.

Crijns

S.P.

, Raaymakers

B.W.

and Lagendijk

J.J.

, Proof of concept of MRI-guided tracked radiation delivery: Tracking one-dimensional motion, Phys Med Biol 57(23) (2012), 7863–7872.

29.

Wilson

P.L.

and Meyer

, A spring– dashpot system for modelling lung tumour motion in radiotherapy, Computational and Mathematical Methods in Medicine 11(1) (2010), 13–26.

30.

Buzurovic

, Yu

and Podder

T.K.

, Active tracking and dynamic dose delivery for robotic couch in radiation therapy, Conf Proc IEEE Eng Med Biol Soc (2011), 2156–2159.

31.

Lang

, Zeimetz

, Ochsner

, Daners

M.S.

, Riesterer

and Klöck

, Development and evaluation of a prototype tracking system using the treatment couch, Medical Physics 41(2), (2014).

32.

Chuang

H.C.

, Hsu

H.Y.

, Chiu

W.H.

, Tien

D.C.

, Wu

R.H.

and Hsu

C.H.

, Verification and compensation of respiratory motion using an ultrasound imaging system, Medical Physics 42(3) (2015), 1193–1199.

33.

Chuang

H.-C.

, Chiou

C.-Y.

, Tien

D.-C.

, Huang

D.-Y.

, Wu

R.-H.

and Hsu

C.-H.

, A compensating system of respiratory motion for tumor tracking: Design and verification, Journal of X-Ray Science and Technology 20 (2012), 161–174.

34.

Chuang

H.-C.

, Huang

D.-Y.

, Tien

D.-C.

, Wu

R.-H.

and Hsu

C.-H.

, A respiratory compensating system: Design and performance evaluation, Journal of Applied Clinical Medical Physics 15(3) (2014).

35.

Redmond

, Song

, Fox

, Zhou

, Rosenzweig

and Ford

, Respiratory motion changes of lung tumors over the course of radiation therapy based on respiration-correlated four-dimensional computed tomography scans, International Journal of Radiation Oncology Biology Physics 75(5) (2009), 1605–1612.

36.

Liu

H.H.

, Balter

, Tutt

, Choi

, Zhang

, Wang

, Chi

, Luo

, Pan

, Hunjan

, Starkschall

, Rosen

, Prado

, Liao

, Chang

, Komaki

, Cox

J.D.

, Mohan

and Dong

, Assessing respiration-induced tumor motion and internal target volume using four-dimensional computed tomography for radiotherapy of lung cancer, International Journal of Radiation Oncology Biology Physics 68(2) (2007), 531–540.

37.

Giraud

, De Rycke

, Dubray

, Helfre

, Voican

, Guo

, Rosenwald

J.C.

, Keraudy

, Housset

, Touboul

and Cosset

J.M.

, Conformal radiotherapy (CRT) planning for lung cancer: Analysis of intrathoracic organ motion during extreme phases of breathing, International Journal of Radiation Oncology Biology Physics 51(4) (2001), 1081–1092.

38.

Chuang

H.-C.

, Chiou

C.-Y.

, Tien

D.-C.

, Wu

R.-H.

and Hsu

C.-H.

, Motion artifacts in CT scans: A study by computer simulation and mechanical phantom, Journal of the Chinese Society of Mechanical Engineers 32(2) (2011), 147–154.

39.

Rivera

D.S.

, Cohen

M.S.

, Clark

W.G.

and Chu

A.C.

, An Implantable RF solenoid for magnetic resonance microscopy and microspectroscopy, IEEE Transactions on Biomedical Engineering 59(8) (2011), 2118–2125.

40.

Yang

, Fridjonsson

E.O.

, Johns

M.L.

, Wang

and Fane

A.G.

, A non-invasive study of flow dynamics in membrane distillation hollow fiber modules using low-field nuclear magnetic resonance imaging (MRI), Journal of Membrane Science 451 (2014), 46–54.

41.

Ichikawa

and Araki

, Fast magnetic resonance imaging of liver, European Journal of Radiology 29(3) (1999), 186–210.

42.

Christensen

J.D.

, Kirichenko

A.V.

and Gayou

, Treatment dose verification in liver SBRT using a deformable respiratory motion model and treatment beam’s-eye-view fluoroscopy with tumor tracking, International Journal of Radiation Oncology Biology Physics 84(3) (2012), 202.

43.

Shirato

, Oita

, Fujita

, Watanabe

and Miyasaka

, Feasibility of synchronization of real-time tumor-tracking radiotherapy and intensity-modulated radiotherapy from viewpoint of excessive dose from fluoroscopy, International Journal of Radiation Oncology Biology Physics 60(1) (2004), 335–341.

44.

Ojha

, Griesmer

, Hu

and Shao

, PET performance of the GEMINI TF PET– MR: The world’s first whole body PET– MRI scanner, Nuclear Science Symposium Conference Record (NSS/MIC) (2010), 2013–2015.

45.

Chang

, Clark

J.W.

and Mawlawi

O.R.

, Investigating the use of non-attenuation corrected PET images for the attenuation correction of PET data in PET/MR systems, Nuclear Science Symposium and Medical Imaging Conference (NSS/MIC) (2011), 3784–3787.

46.

Fontanarosa

, van der Meer

, Bamber

, Harris

, O’Shea

and Verhaegen

, Review of ultrasound image guidance in external beam radiotherapy: I. Treatment planning and inter-fraction motion management, Phys Med Biol 60 (2015), 77–114.

47.

Hwang

, Kim

J.-B.

, Bang

W.-C.

, Kim

J.D.K.

, Kim

C.-Y.

and Lee

, Robust real-time respiratory motion tracking using ultrasound image sequences, Ultrasonics Symposium (IUS) (2012), 1666–1669.

48.

Zhu

, Yang

, Kang

, Ding

and Wang

, A respiration correction method for real-time ultrasound image fusion with pre-acquired 3D dataset, IEEE International Conference on Medical Imaging Physics and Engineering (ICMIPE) (2013), 162–165.

49.

Kim

J.-B.

, Hwang

, Bang

W.-C.

, Kim

J.D.K.

and Kim

C.-Y.

, Real-time moving organ tracking in ultrasound video based on a 3D organ model, Ultrasonics Symposium (IUS) (2012), 2659–2662.

50.

Blanc

, Melodelima

, Rit

, Rivoire

and Sarrut

, Towards 4DCT-US image fusion for liver motion monitoring, Ultrasonics Symposium (IUS) (2013), 817–820.

51.

Hiransakolwong

, Hua

K.A.

, Vu

and Windyga

P.S.

, Segmentation of ultrasound liver images: An automatic approach, International Conference on Multimedia and Expo, 2003. ICME ’03. Proceedings 1 (2003), 573–576.

52.

Pradeep Kumar

B.P.

, Prathap

and Dharshith

C.N.

, An automatic approach for segmentation of ultrasound liver images, International Journal of Emerging Technology and Advanced Engineering 3(1) (2013).

53.

Mahmood

N.H.

, Zulkarnain

and Zulkifli

N.S.A.

, Ultrasound liver image enhancement using watershed segmentation method, International Journal of Engineering Research and Applications (IJERA) 2(3) (2012), 691–694.

54.

Subrahmanyam Murala

Q.M.

and Wu

, MRI and CT image indexing and retrieval using local mesh peak valley edge patterns, Signal Processing: Image Communication 29 (2014), 400–409.

55.

Singh

and Garg

, An approach for segmentation of ultrasound liver images and extracting the features of segmented part, International Journal of Advanced Research in Computer Science and Software Engineering 5(2) (2015).

56.

Chuang

H.-C.

, Hsu

H.-Y.

, Nieh

S.-K.

and Tien

D.-C.

, he feasibility of the auto tuning respiratory compensation system with ultrasonic image tracking technique, Journal of X-Ray Science and Technology 23(4) (2015), 503–516.