Evaluation of the combined use of two different respiratory monitoring systems for 4D CT simulation and gated treatment

Abstract Purpose Two different respiratory monitoring systems (Varian's Real‐Time Position Management (RPM) System and Siemens’ ANZAI belt Respiratory Gating System) are compared in the context of respiratory signals and 4D CT images that are accordingly reconstructed. This study aims to evaluate the feasibility of combined use of RPM and ANZAI systems for 4DCT simulation and gated radiotherapy treatment, respectively. Methods The RPM infrared reflecting marker and the ANZAI belt pressure sensor were both placed on the patient's abdomen during 4DCT scans. The respiratory signal collected by the two systems was synchronized. Fifteen patients were enrolled for respiratory signal collection and analysis. The discrepancies between the RPM and ANZAI traces can be characterized by phase shift and shape distortion. To reveal the impact of the changes in respiratory signals on 4D images, two sets of 4D images based on the same patient's raw data were reconstructed using the RPM and ANZAI data for phase sorting, respectively. The volume of whole lung and the position of diaphragm apex were measured and compared for each respiratory phase. Results The mean phase shift was measured as 0.2 ± 0.1 s averaged over 15 patients. The shape of the breathing trace was found to be in disagreement. For all the patients, the ANZAI trace had a steeper falloff in exhalation than RPM. The inhalation curve, however, was matched for nine patients, steeper in ANZAI for five patients and steeper in RPM for one patient. For 4D image comparison, the difference in whole‐lung volume was about −4% to +4% and the difference in diaphragm position was about −5 mm to +4 mm, compared in each individual phase and averaged over seven patients. Conclusions Combined use of one system for 4D CT simulation and the other for gated treatment should be avoided as the resultant gating window would not fully match with each other due to the remarkable discrepancy in breathing traces acquired by the two different surrogate systems.


| INTRODUCTION
Management of respiratory motion is an important component in the workflow of thoracic and abdominal radiotherapy. 1-3 4D CT incorporates the patient's respiratory information into a stack of 3D images such that the sequential snapshot images at different respiratory phases can be reconstructed. 4,5 With the 4D images, the tumor excursion range with respiration can be obtained and a patient-specific internal margin can be decided for contouring the internal tumor volume. 1,6,7 Respiratory gating in radiation delivery provides effective motion control by disabling radiation beam when respiration magnitude exceeds certain threshold, which is beneficial for reducing the planned target size and sparing more healthy tissues. [8][9][10] 4D CT acquisition requires oversampling CT data for the same slices such that images at different respiratory phases can be reconstructed over a full breathing cycle, which is known as the data sufficiency condition. 11 The 3D volumetric images at individual respiratory phases are obtained by sorting the oversampled images or sinograms based on the associated respiratory phase-angle or amplitude from the breathing trace that is acquired simultaneously during scanning. 12 In the treatment planning process, if a gated treatment plan is decided, the 4D images are combined with the respiratory motion data to select an appropriate gating window. 10 Several respiratory monitoring systems using external surrogates are available in the market for the purposes of measuring respiratory motion for 4D CT imaging and gated treatment. 5 These systems measure the respiratory motion as the surface movement of abdomen or chest wall (eg, Real-Time Position Management [RPM], 4 C-RAD, 13 GateCT 14 ), or the pressure change in a belt wrapped around the abdomen (eg, ANZAI, 15 Bellows 16 ), etc. There is no consensus of which is the best external surrogate with respect to the correlation with the internal tumor motion.
In our institution, both RPM (Varian Medical Systems, Palo Alto, CA, USA) and ANZAI belt systems (Anzai Medical Co. Ltd, Tokyo, Japan) are used for 4D CT image acquisition while only RPM is available for respiratory gated treatment. In the scenario of mixed use of the ANZAI system for 4D CT and RPM for treatment gating, the differences between the two systems may be detrimental to the desired clinic outcome. Previous studies [14][15][16]  AS (Open 20 RT) (helical scan mode, source rotation time 0.5 s or 1 s). The CT scanner can receive respiratory signal from either the RPM system or the ANZAI system for sorting sinogram to reconstruct 4D images. When the ANZAI system is connected to the scanner as the online respiratory monitoring system, the ANZAI data are integrated into the CT raw data, making it difficult to retrieve the ANZAI respiratory signal for offline analysis. Instead, when the RPM system is used, the RPM data file is saved separately and can be easily imported into the CT workstation through an open interface. In this study, both the RPM and ANZAI systems were placed on the same patient during 4D CT scan, with the RPM connected to the scanner as the online respiratory monitoring system while the ANZAI signal was collected independently by a laptop using the AZ-733V software (ANZAI Medical). The RPM infrared reflecting marker box and the ANZAI pressure sensor were placed between the umbilicus and xiphoid process and adjacent to each other to avoid the potential impact of surrogate positioning difference on the respiratory signal. Data collection was started by clicking the "Record" icon on the RPM interface and the "Start" icon on the AZ-733V interface. To synchronize the respiratory signal acquired by the two systems, the left button of the two mice that were each connected to each computer were soldered together such that a single click on either mouse can start data collection on both systems simultaneously. Figure 1 shows the interfaces of the RPM workstation and the AZ-733V software, as well as the connection of the paired mice.

2.B | Analysis of respiratory data
After completion of 4D CT scans, the respiratory data files were EOE to EOI. The linear correlation coefficient was calculated between the ANZAI and RPM data over the entire traces.
The phase shifts between the two traces were calculated as the time latency to reach the EOI. To compare the trace shape between ANZAI and RPM, each individual breathing cycle between two adjacent EOE were extracted and plotted together. The amplitude was renormalized from 0 to 100 for each cycle and all cycle curves were aligned at the EOI. The average trace shape within one breathing cycle was obtained by averaging the amplitude of the aligned curves as a function of time.

2.C | Comparisons on 4D images
As the open interface mode is used for receiving respiratory signal from RPM, the CT workstation only allows respiratory files in RPM data format to be imported. For comparisons of 4D images reconstructed using both surrogates, the ANZAI file was converted into the RPM file format with the relative pressure measurement replacing the anterior-posterior movement as in the RPM data lines.
Besides, the ANZAI file did not contain the real-time records of CT beam on/off indicators as it was not connected to the scanner during scanning. The TTLin (CT on) and TTLout (CT off) indicators in the ANZAI data lines were obtained from the corresponding RPM data lines with the timestamps aligned. The RPM file and the converted ANZAI file were sequentially imported into the CT workstation to reconstruct two separate sets of 4D images with the same CT raw data (sinogram). The local amplitude sorting algorithm was used for acquiring images at 10 respiratory phases (Fig. 2). The two sets of

3.A | Correlation between the respiratory signal acquired by ANZAI and RPM systems
The synchronized ANZAI and RPM data were plotted together with the signal amplitude renormalized from 0 to 100 in the range of the entire traces. Figure 3(a) shows the respiratory traces from patient #1, with CT beam on/off marked on both traces (scan time 190 s, trace length 230 s). A portion between 80 and 120 s is enlarged in Fig. 3(b) and shows the apparent misalignment between the two traces. Figure 4 shows

3.B | Phase shift between ANZAI and RPM traces
As shown in Fig. 3, the ANZAI trace leads the RPM trace in time, introducing a phase shift between the two traces. The phase shift was measured as the time difference to reach the EOI within each breathing cycle in RPM compared to ANZAI. The phase shift (mean value and standard deviation) for each patient is plotted in Fig. 5 with positive values representing that ANZAI reaches EOI ahead of RPM in time.
The phase shift averaged over all the 15 patients is 0.2 ± 0.1 s. The phase shift calculated as percentage of respiratory period is also plotted in Fig. 5, with the mean value 5.5% ± 2.3% averaged over 15 patients.

3.C | Distortions in respiratory trace shape
Another major discrepancy between ANZAI and RPM is the distortion in the trace shape.  aligned at the EOI. The average breathing cycle shape of ANZAI and RPM are shown in Fig. 6(b). The inhalation curves are in high agreement, whereas the exhalation curve of ANZAI is steeper than that of RPM.
For all 15 patients, the average breathing cycle shapes are shown in Fig. 7. Patients #2-9 are similar to patient #1 with a steeper falloff in exhalation measured by ANZAI than RPM. For patients #10-14, the inhalation curve of ANZAI is also steeper than that of RPM. For patient #15, in contrast, the inhalation curve of ANZAI is relatively more gradual than that of RPM. But the exhalation curve is still steeper in ANZAI than RPM, consistent with all the other patients.

3.D | Comparisons of 4D images
Image comparisons were performed for seven patients (patient #1, #2, #3, #10, #11, #12 and #15), each with two sets of 4D images reconstructed using the ANZAI and RPM data, respectively. The seven patients were selected from the three groups as described in above (Section 3.C.) with respect to the difference in inhalation curve. The local amplitude sorting algorithm was applied to obtain the images for 10 respiratory phases. The scheme of local amplitude sorting is illustrated on Fig. 2. Segmentation for whole lung on each phase was performed in the treatment planning system. Figure 8 shows   Fig. 8(b)] can be observed.
The whole-lung volume and the axial displacement of the diaphragm apex as the function of phase number are shown in Fig. 9.
The displacement of the diaphragm apex is measured as the superior-inferior position change relative to the EOE phase (phase 6, or phase 0%In). As the reconstructed axial slices have a thickness of 2 mm, the change in diaphragm position has a step length of 2 mm. The numerical single-phase differences between ANZAI and RPM measurements in whole-lung volume and diaphragm position are summarized in Table 2. The single-phase differences are given as the range of differences over 10 phases. The overall singlephase differences averaged over the seven patients are −3.7% to +4.1% in whole-lung volume and −5.1 mm to +4 mm in diaphragm position.

| DISCUSSIONS
The synchronized respiratory signal acquired by the ANZAI and RPM surrogate systems were found to be in disagreement as characterized     The maximum intensity projection (MIP) and average intensity images were also generated based on the 4D images of ANZAI and RPM, respectively. The differences in MIP and average images were found to be negligible. Figure 9 shows that the ranges of lung volume and diaphragm position over 10 phases are in agreement between ANZAI and RPM. Therefore, the 4D images of ANZAI and RPM consisting of all 10 phases can be equally used for contouring the internal target volume and treatment planning. However, the impact of the discrepancies between ANZAI and RPM occurs when treatment gating with a portion of 4D images used is implemented and the online respiratory monitoring system is different from that used for 4D CT. In such case, the gating window designed in treatment planning based on 4D CT is mismatched with that applied for gated radiation delivery, which may cause errors in dose distribution received by the patient.

| CONCLUSIONS
In this study, the correlation between two respriatory monitoring systems (ANZAI and RPM) was evaluated and discrepancies were found as characterized as phase shift and shape distortion between the respiratory traces. The results indicate that the two external surrogates have nonequivalent correlation to internal organ motion with respiration. Mixed use of the two surrogates for 4D CT and gated treatment should be avoided as the same gating window does not match on different surrogates and potential errors in dose distribution received by the patient may be thus caused.