Variation in target volume and centroid position due to breath holding during four‐dimensional computed tomography scanning: A phantom study

Abstract This study investigated the effects of respiratory motion, including unwanted breath holding, on the target volume and centroid position on four‐dimensional computed tomography (4DCT) imaging. Cine 4DCT images were reconstructed based on a time‐based sorting algorithm, and helical 4DCT images were reconstructed based on both the time‐based sorting algorithm and an amplitude‐based sorting algorithm. A spherical object 20 mm in diameter was moved according to several simulated respiratory motions, with a motion period of 4.0 s and maximum amplitude of 5 mm. The object was extracted automatically, and the target volume and centroid position in the craniocaudal direction were measured using a treatment planning system. When the respiratory motion included unwanted breath‐holding times shorter than the breathing cycle, the root mean square errors (RSME) between the reference and imaged target volumes were 18.8%, 14.0%, and 5.5% in time‐based images in cine mode, time‐based images in helical mode, and amplitude‐based images in helical mode, respectively. In helical mode, the RSME between the reference and imaged centroid position was reduced from 1.42 to 0.50 mm by changing the reconstruction method from time‐ to amplitude‐based sorting. When the respiratory motion included unwanted breath‐holding times equal to the breathing cycle, the RSME between the reference and imaged target volumes were 19.1%, 24.3%, and 15.6% in time‐based images in cine mode, time‐based images in helical mode, and amplitude‐based images in helical mode, respectively. In helical mode, the RSME between the reference and imaged centroid position was reduced from 1.61 to 0.83 mm by changing the reconstruction method from time‐ to amplitude‐based sorting. With respiratory motion including breath holding of shorter duration than the breathing cycle, the accuracies of the target volume and centroid position were improved by amplitude‐based sorting, particularly in helical 4DCT.


| INTRODUCTION
Advances in technology have led to the development of high-precision radiotherapy that can capture moving targets. [1][2][3] To achieve this, the technology must recognize the motion of a target during the planning of radiation treatment. Four-dimensional computed tomography (4DCT) has been used to obtain temporal and spatial information for a given target. Some studies have reported that 4DCT imaging can help determine the optimal irradiation field, including the planning target volume margin. [4][5][6][7] One such study showed that 4DCT is highly beneficial and should be used for radiation treatment planning when the tumor shows respiratory movement of more than 8 mm. 4 In addition, Langner et al. reported that errors in target extraction and dose calculation may occur if motion artifacts in 4DCT images are not addressed. 8 The two main 4DCT scanning modes are cine CT scanning mode (cine mode) and low-pitch helical CT scanning mode (helical mode). 9 In cine mode, image data are acquired with repeated couch movements. The images are multiple-phase images, reconstructed at the same couch position into different phases based on the respiratory data. In this mode, respiratory data are used in the sorting process following image reconstruction. In helical mode, the image data are acquired while moving the couch with low helical pitch. The respiratory signal is added directly to the projection data during the image reconstruction process.
In clinical practice, several respiratory signals are used to reconstruct 4DCT images, such as tidal volume acquired with a spirometer 10 and surface movement of the abdomen or chest wall acquired by real-time position management (RPM), 11 C-RAD, 12 and GateCT,13 or the pressure changes in a belt wrapped around the abdomen acquired with ANZAI 14 and Bellows. 15 These respiratory signals contain the amplitude and phase, and the 4DCT images are reconstructed using these data.
The amplitude-based sorting algorithm recognizes end-inspiration and end-exhalation, and then determines their amplitudes according to the degree of respiration. Phase-based sorting can be accomplished by time-and phase angle-based sorting algorithms. The time-based sorting algorithm uses uniformly spaced bins between two consecutive end-inspiration phases. Meanwhile, the phase angle-based sorting algorithm uses uniformly spaced bins between three respiratory phases (end-inspiration, end-exhalation, and the next end-inspiration).
The 4DCT images are useful for defining moving targets. However, 4DCT cannot remove motion artifacts completely because image reconstruction algorithms assume that objects are stationary during scanning. Therefore, severe motion artifacts occur when 4DCT images are reconstructed from temporally inconsistent raw data. Some investigators have reported the effects of motion artifacts on the target volumes of 4DCT images in phantom studies. [16][17][18] However, these studies did not account for the effects of irregular breathing patterns, such as unwanted breath-holding. In clinical practice, patients sometimes unconsciously hold their breath during 4DCT scans, leading to artifacts due to missing raw data. Therefore, the moving target is inaccurately depicted in 4DCT images under breath-holding conditions. In addition, for patients with irregular breathing patterns, including unwanted breath-holding, alignment errors in the imaged target volume occur between planning CT and conebeam CT, which may result in under-dosage to the target volume. 19 This study investigated the effects of respiratory motion, including breath-holding, on the target volume and centroid position of 4DCT images, according to different CT scanning modes and respiratory-correlated sorting algorithms. The respiratory motion was as follows:

2.A | Phantom and simulated respiratory motion
where y(t) is the target position at time t, A is the maximum amplitude of 5 mm, T is the motion period of 4.0 s, and C is the constant used to determine the starting phase of the respiratory motion. The following three respiratory motion patterns were used (as shown in Fig. 1): • Type A, as described by formula (1). 20 • Type B, as described by formulas (2) to (4) (n ≥ 1): • Type C, as described by formulas (5) to (6) (n ≥ 1): In Types B and C, unwanted breath-holding appeared once every three respiratory cycles.

2.B | 4DCT data acquisition and image reconstruction process
The 4DCT images were acquired using two different CT scanners with different scanning modes: a cine mode CT scanner (LightSpeed The 4DCT scan parameters in cine mode were as follows (Table 1) ADVANTAGE 4D software was used to assign a phase to each CT slice according to the temporal correlation between the RPM data and the CT image, and 10 respiratory phase images acquired at regular intervals over a respiratory cycle were exported.
The 4DCT scan parameters in helical mode were as follows (Table 1)

2.C | Target volume and centroid position analyses
The 4DCT images were imported into a commercial three-dimensional radiation

3.A | Target volume analyses
The ratios of target volume to V 0 of the three reconstruction methods are shown in Fig. 2 for the various respiratory motion patterns.
With Type A respiratory motion, the root mean square errors (RMSEs) between the reference and imaged target volume were  in helical mode, 4DCT images do not contain unnecessary temporal phase data. In general, the effective temporal resolution of a scan using a short scan reconstruction is equal to half the gantry rotation time in helical mode. 21 We did not discuss the temporal interval for the Siemens scanner because the operators cannot change the temporal interval. The RMSE between the reference and target volume with Type A respiratory motion were smaller in time-based images in helical mode than in time-based images in cine mode. We consider that helical mode has superior in temporal resolution to cine mode, which would cause less motion artifacts. In addition, whereas the time-based sorting algorithm recognizes only inspiration phases, and then divides the phase accordingly, the amplitude-based sorting algorithm recognizes both inspiration and exhalation phases. Therefore, amplitude-based sorting yields a more accurate centroid position.

3.B | Target centroid position analyses
Previous studies have reported various artifacts in 4DCT images, including blurring, as well as duplicated, overlapping, and incomplete structures. 16,22,23 Among these, overlapping and incomplete structure artifacts (Fig. 4)  is not likely to be improved by changing scan mode or reconstruction method. In particular, zonal truncation artifacts, which are specific to helical 4DCT, occurred with Type C respiratory motion due to undersampling of the respiratory cycle (Fig. 4). The target volume was affected by the extent of the zonal truncation artifacts and tended to be overestimated. In addition, the imaged centroid position was different from the reference position due to the artifacts. increase the radiation dose delivered to the patient. Therefore, it is important to encourage radiological technologists engaged in 4DCT scanning to acquire reproducible and regular respiration data during CT data acquisition.
Our study had several limitations. First, the maximum amplitude of the simulated respiratory motion was low, at only 5 mm. However, if the maximum amplitude of the respiratory motion is 10 mm or more, the accuracy of the target volume and centroid position decreases because motion artifacts become significant. 17 Second, simplified respiratory patters were used in this study, while actual respiratory patterns are more complicated (Fig. 1). In addition, such motions would involve centroid movement coupled with couch motion in helical scanning mode due to interplay between couch speed and phantom motion. Therefore, it should be noted that larger errors would occur in clinical practice than observed in our study.
Third, the two CT scanners used could not reconstruct images of the same slice thickness (2.5 mm in cine mode and 2.0 mm in helical mode, appropriate for clinical radiotherapy planning). The impact of the partial volume effect may differ by slice thickness. However, the size of the spherical object used in this study was large enough compared to the difference in spatial resolution, so we believe that the difference would not have affected our results.

| CONCLUSION
We investigated the effects of respiratory motion, including breathholding, on target volume and centroid position in 4DCT images acquired with different CT scanning modes and respiratory-correlated sorting algorithms.
Our results suggest that 4DCT images acquired in helical mode depict the target volume and centroid position more accurately than cine 4DCT. In addition, with respiratory motion including breathholding of shorter duration than the breathing cycle, the accuracy of the target volume and centroid position are improved by amplitudebased sorting, particularly in helical 4DCT.

CONF LICT OF I NTEREST
No conflict of interest.