MATERIALS AND METHODS

This was a phantom study to compare the verification accuracy and organ doses delivered by the MVCT and CBCT. The head-and-neck and pelvic regions were selected as they accounted for about 70% of the patients treated by IMRT in local departments. In this study, the head-and-neck phantom was used to represent the case of nasopharyngeal carcinoma (NPC), whereas the pelvic phantom was for prostate cancer.

The anthropomorphic phantoms of head-and-neck and pelvic regions (Figure 1) were used because they had bony and soft tissues of a human body which provided sufficient definitions in CT image matching. The two phantoms were immobilised by the alpha cradles and scanned in superior-first supine position mimicking the normal treatment position by the CT-simulator following routine protocols with the scan reference marked prior to the scanning. After the CT scanning, a reference standard setup position for each phantom was determined and marked using the laser system. The markings included the anterior mid-line, the reference transverse plane and the two lateral reference levels. The same reference marks and setup centre of the two phantoms were used for the MVCT and kV CBCT verifications. For the verification by CBCT, the DICOM CT images of the two phantoms were sent directly to the linear accelerator, whereas for the MVCT verification, the DICOM CT was first fed to the helical tomotherapy treatment planning system (TomoPlanning, version 3.2.2.7; Hi-Art TomoTherapy Inc, Madison, WI) from which a dummy plan was computed for each phantom. The dummy plans including the CT images were then sent to the helical tomotherapy treatment unit for conducting the verification procedure.

Figure 1. The anthropomorphic head-and-neck phantom showing the cross-section at the mid-neck level. The circular spots are the possible locations for the placement of thermoluminescent dosimeter (some of the holes have been un-plagued for illustration purpose).

To evaluate the verification accuracy of the two systems, markings of known deviations in the lateral translation (x direction: ± 3 mm and ± 7 mm), longitudinal translation (y direction: ± 5 mm and ± 9 mm) and yaw rotation (± 2° and ± 4°) were made on the phantoms with the help of the field light and laser system of a conventional simulator (Figures 2 and 3). The vertical translation (z direction) was not included because it would involve the vertical movement of the treatment table. The possible difference in table adjustment accuracy would affect the result of detection error measurement.

Figure 2. Schematic diagram showing the assigned deviation lines in the x and y directions marked on the phantom (not to scale).

Figure 3. Schematic diagram showing the assigned deviation lines in yaw rotation marked on the phantom (not to scale).

The verification by kV CBCT was conducted in the linear accelerator equipped with a kV CBCT device (VolumeView™ cone-beam CT, Elekta Synergy). The anthropomorphic phantoms were positioned in the reference setup position on the treatment couch to acquire the CBCT images. The phantoms were then shifted in steps manually by moving the phantom according to the deviations as stated above by aligning the corresponding lines of deviation using the laser system. The images were reconstructed with 1 mm resolution in the transverse, coronal and sagittal planes. For each position of the phantom, after taking the kV CBCT images, they were automatically registered with the planning CT images using the registration software. In the automatic image registration, the bone algorithm was used for the head region whereas the soft tissue algorithm was used for the pelvic region. The x, y and z and rotational deviations from the reference position would be calculated and displayed on the monitor. These values were recorded and compared with the known phantom displacements at the standard position. The discrepancy between the measured data and the known data (δD) would indicate the detection error of the verification system. The same procedure was repeated twice for each deviated position to increase reliability and the average values were taken for comparison.

The verification by MVCT was conducted in the tomotherapy treatment unit (Hi-Art System, Tomotherapy). The phantoms were positioned on the treatment couch in the reference positions by aligning the reference marks that represented the setup centre. After positioning, the phantom was moved to the setup centre and a set of reference images was generated with MVCT. The phantoms were then shifted in steps manually by moving the phantom with the help of the laser system according to the assigned lines of deviations. A set of MVCT images was generated for each phantom position. A scan length of 268 mm with slice thickness of 2 mm (total 135 slices) were produced for the head-and-neck region, whereas a scan length of 296 mm with slice thickness of 4 mm (with 75 slices) were generated for the pelvic phantom. The scan lengths of the two phantoms were similar to that of the kV CBCT. The MVCT images were registered with the planning CT images using automatic “bone and tissue” algorithm. The positional deviations detected by the MVCT system were recorded and compared with the known deviations. The same procedure was repeated twice and the average values were used for comparison. Two-tailed pair t-test was used to evaluate the significance of the differences of the mean detection errors between kV CBCT and MVCT. To eliminate the subjectivity of the operators in the image matching process, only the automatic registration provided by the system was employed in this study.

Thermoluminescent dosimeters (TLD-100), lithium fluoride doped with magnesium and titanium were used for dose measurements of kV CBCT and MVCT. Sensitivity variation test was performed to exclude the over or under responded TLDs (> ±3%). Before the dose measurement, the selected TLDs were separately calibrated under the kV and MV x-ray machines following the departmental protocols to determine the kV and MV calibration curves respectively, from which the counts to dose conversion factors were obtained. The localisation of the TLD measurement points in the phantom, which represented the various organs of interest were conducted under a conventional simulator using lead markers placed in the corresponding holes inside the phantoms. The organs of interest in the head-and-neck region included the skin (anterior and two lateral surfaces), lenses (covered by 3 mm wax to mimic the eyelid), parotid glands (covered by 5 mm wax to mimic the superficial skin), spinal cord, brain stem and pituitary gland; whereas for the pelvic phantom, the skin, bladder, rectum, testes and femoral heads were included. During actual dose measurement, which was done simultaneously with the accuracy detection measurement, three TLDs were packed in each specified organ location in the phantom. The average value of the three TLDs was calculated as the counts for that specific organ of interest. After the irradiation of TLDs, their counts were read by the TLD reader system (standard TLD system, version 2.08; Rialto, NE Technology) and the individual base counts and the average background counts were subtracted from the readout counts. The net counts were then converted to the absorbed dose using the corresponding calibration curves and the doses delivered to the organs of interest per scan were calculated for the data analysis. Two-tailed pair t-test was used to evaluate the significance of the differences between kV CBCT and MVCT.

RESULTS

For the assessment of verification accuracy by the kV CBCT and MVCT in the x and y directions, the mean detection errors (δD) were ranged from 0.23 mm to 1.13 mm, whereas the range was 0.24°–0.84° for the yaw rotation. For the δD of the x direction, kV CBCT demonstrated lower values than the MVCT in both the head-and-neck and the pelvis phantoms (Table 1), in which only the difference in the 7 mm deviation line of the head-and-neck phantom did not show significance (p = 0.061). For the y direction measurements, there was no significant difference between the kV CBCT and the MVCT in both head-and-neck and pelvic phantoms (Table 2). In general, the values of δD were greater in the pelvic phantom than the head-and-neck phantom in both x and y directions. For the yaw rotation measurement, δD was significantly larger in the MVCT compared to that of the kV CBCT for the head-and-neck phantom. However the differences were insignificant in the pelvic phantom (Table 3).

Table 1. Comparison of δD (discrepancy between the measured data and the known data) in the x direction between kV CBCT and MVCT

Table 2. Comparison of δD (discrepancy between the measured data and the known data) in the y direction between kV CBCT and MVCT

Table 3. Comparison of δD (discrepancy between the measured data and the known data) in the yaw rotation between kV CBCT and MVCT

For the head-and-neck phantom, the mean doses per scan for the kV CBCT were significantly lower than that of the MVCT in all organs of interest except skin (Table 4). However the difference of the skin dose was not statistically significant (p = 0.145). For the pelvic phantom, kV CBCT delivered higher mean dose per scan to the skin, bladder and testis, in which only the bladder dose did not reach significance (p = 0.331). On the other hand, MVCT delivered significantly higher mean dose per scan to the femoral heads than the kV CBCT (Table 5).

Table 4. Comparison of organ doses per scan in the head-and-neck region between kV CBCT and MVCT

Table 5. Comparison of organ doses per scan in the pelvis region between kV CBCT and MVCT

DISCUSSION

Both kV CBCT and MVCT are advanced image-guided systems for radiotherapy and they provide 3D–3D matching of the anatomical structures in the verification process. In this study, both modalities were proved to be effective verification systems and produced sub-millimetre and sub-degree mean translational and rotational detection errors (δD), with the only exception in the x direction of the pelvic phantom.

Comparatively, the kV CBCT performed better in detecting the deviations in the x direction and yaw rotation, in which all detection errors (δD) were smaller than that of the MVCT, only the differences in 7 mm deviation of the head-and-neck phantom and yaw rotations of the pelvic phantom did not reach significance. The main reason for this result was due to the kV x-ray interaction of the CBCT produced relatively better image contrast, which allowed better image recognition during the matching process. However, the situation was not the same in the y direction in which the differences of δD were not significant. This was largely due to the reconstruction algorithm used by the kV CBCT was Feldkamp (FDK) algorithm. Based on the information obtained from the projections acquired in a single circular orbit, the FDK algorithm had been shown to introduce artefacts when constructing planes parallel to the mid-plane (central transverse plane of the cone-beam projection). As the distance from the mid-plane increased along the longitudinal direction, the intensity of the images dropped and caused slight degrading of the image quality.Reference Zhu, Starman and Fahrig^11,Reference Renström¹² Consequently, the slight weakness in error detection ability in the longitudinal direction of the kV CBCT reduced the error detection power and made it comparable to that of the MVCT.

Both kV CBCT and MVCT generally showed relatively large detection errors (δD) in the pelvic phantom. It was mainly due to the inherent limitation of the pelvic phantom, which did not contain the soft tissues structures such as bladder and prostate apart from the bones for the image matching process. As a result, it might reduce the accuracy in organ matching and error detection. The head-and-neck region was less affected because the image registration was mainly based on the bony structures, which was readily available in the phantom. Therefore, it is expected that the verification accuracy of the pelvic region will be improved if the scan were applied on real patients, where the soft tissue landmarks are present.

MVCT delivered significantly higher doses per scan to most organs than kV CBCT, except for the skin and testis. These differences in organ doses were largely due to the differences in the methods of image acquisition and x-ray energy of the two verification systems. MVCT image was generated with the gantry continuously rotating at 360° round the phantoms, which resulted in higher integral dose. Besides, a relatively higher dose was required for MVCT to achieve the same contrast resolution as kV CBCT due to the inverse relationship of the attenuation coefficient and radiation energy.Reference Shah, Langen, Ruchala, Cox, Kupelian and Meeks¹³ In addition, due to the more penetrating MV radiation, the doses delivered to the deep seated organs were higher in the MVCT than that of the kV CBCT. The difference in x-ray penetration also explained why kV CBCT delivered higher doses to the more superficial organs such as the skin and testis, as the maximum intensity of kV x-ray was at the surface. For the lens and parotid glands, where the dose measurement points were situated about 3–5 mm below skin, they received slightly higher doses in the kV CBCT scan relative to other more deep seated structures. It was worth noting that in the head-and-neck phantom, the highest dose in the MVCT scan was delivered to the cervical spinal cord. It was because it was the narrowest part of the body and experienced the least dose attenuation from the MV radiation. It was also noted that the doses per scan for the pelvic organs of the kV CBCT were higher than that of the head-and-neck phantom. This was because higher exposure factors and larger gantry rotation angle (full 360° compared to 235° in head-and-neck) were required for the generation of images in the pelvic phantom due to its larger size.

To simulate an actual radiotherapy treatment, an estimation of the total dose to the organs of interest based on the doses per scan obtained for an actual treatment course would be useful. NPC and prostate cancer were chosen as the diseases for the head-and-neck and pelvic regions respectively. It was because both of them were common malignant diseases encountered locally and required a cancerocidal dose of about 70 Gy.

Based on a 35 fractions radiotherapy course to NPC, the total extra organ doses calculated for the kV CBCT were not alarming as most of them did not exceed 6 cGy (Table 6). Even for the skin, the extra 70 cGy was not expected to have any clinical significance when added to the dose received from the main treatment course, which was 3,000–4,000 cGy. However, for the MVCT, special attention was needed for the doses to the lens, brain stem and spinal cord because they were the dose-limiting organs at risk for NPC treatment, the addition of 70–75 cGy to these organs might bring their total doses close to their tolerance.

Table 6. Comparison of estimated total organ doses received during a full course of radiotherapy (35 fractions) between kV CBCT and MVCT

For the prostate case, the extra total doses to the organs of interest between the kV CBCT and MVCT were similar. All differences were <15 cGy and not expected to cause any clinical difference. The doses to the bladder and rectum were ranged from 56.70 to 66.85 cGy. These might require some attention particularly when the doses to these organs were already high in the main treatment. However, the actual situation would depend on whether the measurement points were at the high dose region of the organs or not. Because the testis was usually outside the radiation field in the whole pelvic irradiation, the additional 70 cGy from the CT scan would not be a serious issue. Furthermore, the estimated total doses for the kV CBCT might be exaggerated in this study because kV CBCT was usually not performed daily in many centres, unlike the mandatory daily MVCT.

CONCLUSION

Both kV CBCT and MVCT achieved effective verification capability for the head-and-neck and pelvic phantoms with most mean detection errors fell below 1.0 mm or 1.0°. Only the detection errors in the lateral translation and yaw rotation by kV CBCT were relatively smaller than MVCT. With regard to the additional organ doses, it was unlikely that the extra dose delivered by the kV CBCT will increase the complication risk. However, the doses to the deep seated organs were higher in the MVCT. Special attention was needed for the lens, brain stem and spinal cord in the MVCT verification in the nasopharyngeal cancer case because an extra dose of about 70 cGy in a treatment course might bring their total doses close to their tolerance. Further studies on other treatment regions such as the chest and abdomen would provide a more comprehensive picture about the strengths and weaknesses of the two verification systems.