Image-Guided Radiotherapy Using MV for Prostate Cancer: A Correlation Analysis between Electronic Portal Imaging with Fiducial Markers and Cone Beam CT

Citation: Salas Buzón Mdel C, Gutiérrez Bayard L, Lagares-Franco C, de Ingunza Barón L (2016) ImageGuided Radiotherapy Using MV for Prostate Cancer: A Correlation Analysis between Electronic Portal Imaging with Fiducial Markers and Cone Beam CT. J Adv Radiol Med Image 1(1): 101. doi: 10.15744/24565504.1.101 Volume 1 | Issue 1 Journal of Advances in Radiology and Medical Imaging Background Background: To study the reliability and agreement between electronic portal imaging (EPI) and cone beam computed tomography (CBCT) with megavoltage (MV) and fiducial markers (FM), in image-guided radiotherapy (IGRT) for prostate cancer. IGRT combined with three-dimensional conformal radiation therapy (3D CRT) and intensity-modulated radiation therapy (IMRT) is gaining quick acceptance in radiotherapy treatment of prostate cancer [1-4]. The advantage of this technique is that it localizes the target volume accurately, thus increasing precision in treatment delivery. Also, IGRT allows to reduce planning target volume (PTV) margins, which spares rectal and bladder toxicity and makes modern, daily-use technological tools highly reliable [5-8] .

A condition on the use of FM is a clear identification of markers on the images without introducing a significant amount of artifacts. The reduction of artifacts on the planning CT is especially desirable, since extensive artifacts could interfere with structure delineation, dose calculation and affect the clarity of the anatomical region of interest [18].
The gold markers, are widely used in IGRT, having studied the visibility and artifacts resulting in various modalities of image [19]. After analysis of gold FM, carbon and polymers as fiducial markers on the CT images, gold markers produced a significant artifact, while no artifacts were observed with polymer markers. Also the contrast increased with KV for the gold markers, whit decreasing for the polymer and carbon marker [20,21]. If MV will be the sole imaging modality for positioning verification, it may be necessary to use gold markers despite the artifacts they create on the CT simulation images [20].
Although we cannot ignore the fact that insert FM is an invasive procedure, and that involve an economic cost and human resources [22,23], fiducial marker implantation for prostate IGRT is associated with a small percentage of patients experiencing moderate to severe complications requiring further medical interventions [24].
The MV LINAC allow 2D orthogonal images systems such as electronic portal image (EPI) [36] and volumetric or 3D imaging systems, such as cone beam computed tomography (CBCT) [37,38]. Both imaging methods have different characteristics and acquisition protocols and are performed differently. However, such differences do not mean that any of these systems is superior to the other. There is no clear evidence on the best imaging method in daily clinical practice [39,40], the most accurate in target localization, or the preferable method -planar or volumetric images-when an IGRT protocol is initiated for treating a prostate cancer [30,31,41,42].
Both, 2D and 3D imaging systems have advantages and disadvantages, and there is no general agreement on the best method to be used in centers fitted with MV IGRT equipment [30]. It depends on the technical characteristics of the LINAC, the energy available for the acquisition of images, the use or not of FM, and the organizational reliability of each institution [29,34,37].
There are some studies conducted to assess agreement between EPI and CBCT but have different methods of assessment to ours [29][30][31]34,43], and performed mainly with KV energy, nevertheless there are few data quantifying the agreement between EPI and CBCT with MV energy.

Methods
Therefore, the aim of this study was quantify the reliability or reproducibility of the different imaging methods provided by a exclusive MV image device, by calculating agreement between EPI, CBFM and CBST.
The study sample was composed of 40 patients with early T1-T3a prostate adenocarcinoma included in a prospective IGRT protocol between January 2012 and June 2013. The study was approved by the Ethics Committee of the Puerta del Mar University Hospital, Cadiz, Spain.
Four (3mm x 1mm) gold FM (Lorca Marín®) were implanted in the prostate under trans-rectal ultrasound guidance. Treatments were delivered using the LINAC ONCOR, SIEMENS, of 6 MV. The Planning Target Volume (PTV) was defined as the prostate ± the seminal vesicles with a 7 mm. margin in each direction, except for a 5 mm. posterior margin. Ninety-five percent of the dose was delivered to 95% of the target volume. Patients and treatment characteristics are presented in Table 1. Journal of Advances in Radiology and Medical Imaging clearly and easily identified and localized on both simulation and verification images. An important requirement is that serve to reduce uncertainties related to the process of positioning and administration of radiotherapy [2,17] . Abbreviations.3D CRT: 3D conformational radiotherapy; IMRT: intensitymodulated radiation therapy All patients underwent bowel, colon and vesicle preparation according to an adapted protocol of Smitsmans et al. [44]. With the patients in supine position with a support under the knees and feet, CT scans (Phillips Brillance CT 64-slices, 120 Kvp, 60 mA) of the pelvis were acquire with a slice thickness of 3 mm and a transversal pixel size of 0,65 x 0,65 mm 2 . Reference points were marked on the skin. Treatment volumes, FM, bladder, rectum and femoral heads were contoured using the Virtual Simulator PCRT, version 6.0. Patients were positioned daily using the immobilization system described above and aligned to the treatment room isocentre using skin markers. Daily pre-treatment 6 MV-EPI were acquired by antero-posterior (AP) and lateral (LAT) projection, using 2 Units Monitor (UM) in AP projection, and 3 UM in LAT projection. By means of specific ONCOR software (Coherence Therapist system Work Space 2.2.09), we performed the matching of the four FM on MV EPI AP and LAT images, which were compared with digitally reconstructed radiographs (DRR) generated from the planning CT. Shifts between image sets were calculated automatically to identify the new position of the isocenter in cm. in the three axes. Isocenter deviations from initial tattoos (skin markers) were calculated and recorded, obtaining 1520 EPI (38 sessions per patient). For the purposes of concordance analysis, the EPI on days 1,2,3,4,5,10,15,20,25,30 and 35 were selected: 440 EPI (projections AP and LAT).

Treatments, EPI and FM-based registration
A total of 440 CBCT (29%) were acquired in 11 sessions per patient (days 1,2,3,4,5,10,15,20,25,30 and 35) at standard dose, using 190-degree arc therapy. Thickness of reconstruction: 10 mm, 10 x 10 cm field. Off-line reconstruction and MV-CBCT adjustment with FM (CBFM) and soft-tissue (CBST). Shifts were firstly estimated by alignment of FM on CBFM and later of soft tissues on CBST. Deviations of the isocenter were measured in the AP, SI and LR dimensions. Shifts estimated by CBCT were registered for analysis, but they were never applied to patients. The matching of the MV-EPI and MV CBFM was performed by five physicians while off-line matching of MV CBST was performed by a single physician.
With the patient immobilized, the isocenter of the treatment was immediately aligned using the shifts identified on EPI images with FM. The radiotherapist did not enter into the room. The process was performed as quickly as possible in 25 to 30 minutes.
An interobserver study was performed to evaluate the influence of interobserver variation on FM matching with the MV CBFM image set. Because these markers are clearly visible on CBCT, they are not clearly visible on KV CT reference images due to the distortion caused by the high atomic number of gold, which causes a slight blurring of the markers and needs careful matching [18][19][20]. Image sets were randomly chosen from five patients selected from the 40 study patients. The interobserver study was performed by five physicians with experience using planning system alignment tools and involved in the image-guidance program. The observers performed volumetric registration using translational shifts only, to align the FM as visualized on the CBCT dataset with the FM visualized on CT reference images. The resulting isocenter shift was computed and recorded for each observer and for each dataset. To give an estimate of intraobserver variability, one observer repeated the entire FM alignment procedure 1 week later.

Interobserver and intraobserver study
An intraobserver study was performed to evaluate the influence of intraobserver variation for soft tissue matching on MV-CBST following the same procedure. Both studies were conducted over a period of seven days.
Statistical analysis was performed using IBM SPSS Statistics 21 and the freeware EPIDAT 4.0. Quantitative variables are described as frequencies and tables. Agreement or reliability of quantitative variables are frequently assessed using Pearson's correlation coefficient (r). However, in this study the Pearson's correlation coefficient is not useful, given that two measuring tools can systematically provide different values other and obtain a perfect correlation (r=1) but null reliability [45]. Therefore, Intraclass Correlation Coefficient (ICC) [46], was used to assess the agreement or reliability of quantitative data for the two techniques i.e. displacement of the isocenter in the three spatial axes between MV EPI-FM and MV CBCT (FM or ST). ICC is defined as the proportion of variance among patients and provides an objective coefficient that is especially useful when comparing two methods, as it is the case of this study.
The difference between measured couch shifts was then plotted against the average couch shift as recommended by Bland and Altman [48] when new methods of measurement emerge. The 95% confidence interval (CI) of the error distribution is also reported. The percentage of shifts within a ± 3 mm and ±5 mm tolerance was calculated for the three techniques, in order to make a comparison with previous concordance studies. The frequency of differences between measurements in each plane was calculated and represented graphically (Figure 2).

Statistical Analysis
As a rough guide, we followed the ratings suggested by Landis and Koch et al. An agreement level of 0-0.2 was considered poor, 0.2-0 .4 fair, 0.4-0.6 moderate, 0.6-0.8 substantial, and 0.8-<1.0 almost perfect [47]. Interobserver agreement for MV EPI, MV CBFM and planning CT was calculated by the ICC in each axis, while intraobserver agreement was calculated for MV CBST (Figure 1).

Results
To evaluate the correlation between the three imaging method, each one was compared to the other two. As summarized in Table  2, ICC values are shown along with their corresponding 95% CI and significance values for all three axis. We found an "almost perfect" strength of agreement between MV EPI-FM and MV CBFM in LR, SI and AP, as well as between MV EPI and MV CBST and between MV CBFM and MV CBST.     Inter and intraobserver Intraclass Correlation Coeficient Figure 2 shows histograms of the frequency of differences between shifts in each technique in each axis.

Discussion
To the extent of our knowledge, there are no previous studies comparing agreement between MV energy and FM imaging methods used before treatment and sequentially in the same patient. Such lack of literature is surprising, considering the large number of institutions that use MV IGRT exclusively.
We evaluated agreement between planar systems and MV volumetric systems using EPI images with FM, since according to numerous studies, it is a valid reference method in IGRT [11,15,16,41,[49][50][51]. As alternative methods, we used CBCT with fiducials and CBCT with soft-tissue. Table 4 summarizes the studies assessing agreement between IGRT methods with FM and transponders.  2009 Ogunleye P. et al. [39] Cone-beam computed tomography and KV fiducial imaging are similar; however, more than one fourth of CBCT and KV shifts differed enough to affect target coverage.   The highest coefficient of concordance (ICC) was found between MV-FM EPI and MV CBFM: 0.911, 0.849, 0.924 in the LR, SI and AP axes respectively, which is an "almost perfect" agreement, similar to that reported in previous studies [31,34]. Moseley et al. reported a Pearson's correlation coefficient of 0.95, 0.81, and 0.84 in the LD, SI, and AP dimensions between MV and KV using FM, and 0.90, 0.49 and 0.51 respectively using soft tissue [34]. Foster el al. reported high coefficients of concordance between kV imaging systems, CBCT and the Calypso-4D system. Although the mean of mismatches among the three methods was lower than 0.8 mm, Foster observed more uncertainty in prostate localization by CBCT as compared to 2D orthogonal images, although such differences were not clinically significant [31].

Agreement between planar and volumetric imaging techniques with FM and ST registration
In our study, interobserver variability in MV CBFM did not have any impact on agreement among images. According to the ICC values obtained for the five observers, agreement was excellent. And it was so despite the potential difficulty involved in seed matching due to sligth distortion and artifacts caused on gold seeds on reference CT [18][19][20], which is more dramatic in the LR and SI dimensions. The high agreement observed demonstrates that the MV CBCT system has millimeter accuracy and precision when localizing unambiguous objects such as FM [25].
Analyzing the possible impact of the size of the FM used, we do not believe to be significant. DeLangen et al, and Chan et al, investigated a large number of commercial FM including solid gold, gold coil, and polymer types, in various imaging IGRT modalities. They found that solid markers were preferred in EPI, that only the large diameter markers are visible in MV, that the marker measuring 0,9 mm x 5 mm appears to be clinically optimal in pelvic radiotherapy patients (80% visualization success in lateral EPID) [18,52] and visibility of gold coils of 0,75 mm diameter or larger is comparable across all imaging modalities studied.
When the planar or volumetric system was compared with CBST, agreement diminished in the three axes. Agreement was lower in the LR and SI axes, and higher in the AP axis, where agreement was "almost perfect". When comparing the two volumetric imaging methods, the results obtained are very similar, with higher agreement in the AP axis. There is an explanation to such results, since although the quality of MV images is lower than that of KV images [53], the interface and gray gradients on rectum and prostate tissue MV images (often with gas inside) and prostate and seminal vesicle tissue images is higher as compared to the interface and gray gradients on prostate and peripheral prostate tissue images in the LR axis [37,43], and on prostate and penile bulb tissue images, in the SI axis. The quality of the image is also deleterious to the reproducibility of IGRT based on soft-tissue, which could increase interobserver variability [53]. Langen et al. stated that 3D data delivered by CBCT images may hinder the localization of the prostate, as it adds more uncertainty than two 2D orthogonal images. In addition, reported that FM are more advantageous and accurate, which would reduce inter-user variability [54].
We reviewed the studies available comparing EPI with CBCT based on FM/electromagnetic transponders and observed that most studies were performed with KV energy and yielded higher percentages of concordance when using FM/transponders than when using volumetric images without markers [29,34,54].
If the rotation of the prostate is not considered, only Logadottir et al. reported similar position uncertainties when comparing 2D/3D techniques based on FM and KV soft-tissue for patient positioning, soft-tissue requiring only a 1mm higher positioning margin [30].
The highest percentage of shifts within a ± 3 mm. differences was between EPI-FM and CBFM: 94.5%, 89.5% and 90.7% in LR, SI and AP directions, as compared to CBST, which is in agreement with the results obtained by Moseley et al. [34] and Foster et al [31] and slightly lower in the SI axis. The CT slice thickness also adds to the uncertainty inherent to localization by CBCT, as compared to 2D localization based on FM. In our study, CT slice thickness was 3 mm. Slices of lower thickness such as 2.5 mm. [34], or 1 mm. [31], would improve precision, since a better resolution in the SI direction would increase the precision of CBMF scans in this axis.
A limitation to this study is that FM was visible during off-line reconstruction of CBST, which was performed by a single physician. As FM could not be suppressed using any software, the strategy used was to initiate matching by bone references and then by prostate soft tissues, ignoring FM. We are aware that this could lead to bias, therefore, concordance results in matching based on CBST would be considered cautiously. This bias was unavoidable, since we wanted to assess agreement among MV imaging techniques prior to delivering the treatment to each patient.
This study contributes useful comparative data on the effectiveness of MV in IGRT that could be used in the selection of the most appropriate imaging method for positioning patients with prostate cancer prior to the treatment.

Conclusions
The precision, agreement and reliability of IGRT based on MV imaging for the localization of the prostate is similar among EPI and CBCT based on FM. The lower precision or reliability observed in the absence of FM could be due to the lower quality on MV CBCT images localizing soft tissues, which have a higher impact on the shifts observed. It is true that el use of soft tissue resulted in greater discrepancy, but the difference is clinically insignificant. The difference between MV CBFM and MV CBST was less than 2 mm in any direction. Considering the PTV margin, it is safe to use soft tissue for guidance to ensure adequate coverage.