Radiotherapy portal verification: an observer study

Assessment of set-up discrepancies using daily portal imaging during radiotherapy treatment for patients with spine and bone metastases

It is well established that patients with bone metastases get good pain relief from radiotherapy. The aim of treatment is to achieve maximum pain relief with minimum morbidity. Accuracy and reproducibility of the patient’s position are fundamental to the successful delivery of radiation therapy. It has been recognised for many years, that the accuracy of patient positioning will improve the success of radiation treatment. A previous study carried out in the department showed that the use of only a single tattoo for the set-up of palliative patients resulted in poor accuracy. The aim of this study was to assess if the addition of extra skin marks improved the set-up accuracy of palliative patients being treated for spine and bone metastases. A protocol was implemented detailing the extra skin marks to be used. Daily portal images were acquired and analysed retrospectively using anatomy matching. The results obtained were then compared with those of the previous study. The use of extra skin marks resulted in a total of 45% of images within 5 mm tolerance compared with 36% of images in patients treated with a single centre tattoo. Also, the number of images with deviations greater than 15 mm was reduced by more than 50% with the addition of extra skin marks. This study has shown that extra skin marks do increase the set-up accuracy in palliative patients treated for spine and bone metastases. Therefore, the practice of using extra skin marks has become standard protocol for all palliative patients within the department.

A phantom study on target localization accuracy using cone-beam computed tomography

The purpose of this study is to evaluate the 3-dimensional target localization accuracy of cone-beam computed tomography (CBCT) using an on-board imager (OBI). An anthropomorphic pelvis phantom was used to simulate a range of offsets in the three translational directions and rotations around each of the three axes. After a translational or rotational offset was applied, a CBCT scan of the phantom was followed by image registration to detect the offsets in six degrees. The detected offsets were compared to the offset actually applied to give the detection error of the phantom position. Afterwards, the phantom was positioned by automatically moving the couch based on the detected offsets. A second CBCT scan followed by image registration was performed to give the residual error of the phantom positioning. On the average the detection errors and their standard deviations along the lateral, longitudinal and vertical axis are 0.3 ± 0.1, 0.3 ± 0.1 and 0.4 ± 0.1 mm respectively with respect to translational shifts ranging from 0 to 10 mm. The corresponding residual errors after positioning are 0.3 ± 0.1, 0.5 ± 0.1 and 0.3 ± 0.1 mm. For simulated rotational shifts ranging from 0 to 5 degrees, the average detection error and their standard deviation around lateral, longitudinal, and vertical axes are 0.1 ± 0.0, 0.2 ± 0.0, and 0.2 ± 0.0 degrees respectively. The residual errors after positioning are 0.4 ± 0.1, 0.6 ± 0.1, and 0.3 ± 0.1 mm along the lateral, longitudinal and vertical directions. These results indicate that target localization based on CBCT is capable of achieving sub-millimeter accuracy.

Radiation therapy treatment verification imaging in Australia and New Zealand

An original questionnaire was used to investigate the available types of reference and treatment image verification equipment and specific practices related to image analysis. A section on treatment site-specific imaging was included. The questionnaire was distributed to all radiation oncology facilities in Australia and New Zealand. A response rate of 87% (40/46) was achieved. Most facilities (90%) in Australia and New Zealand reported the availability of electronic portal imaging devices. Use of computer software to assist with image interpretation was indicated by 92% of centres. Frequency of image acquisition and tolerance levels used for radical treatment sites were variable, but palliative treatment site protocols were more consistent between treatment facilities. In conclusion, departments should strive to use evidence-based protocols and guidelines to ensure acceptable accuracy in treatment delivery.

Novel image registration quality evaluator (RQE) with an implementation for automated patient positioning in cranial radiation therapy

In external beam radiation therapy, digitally reconstructed radiographs (DRRs) and portal images are used to verify patient setup based either on a visual comparison or, less frequently, with automated registration algorithms. A registration algorithm can be trapped in local optima due to irregularity of patient anatomy, image noise and artifacts, and/or out-of-plane shifts, resulting in an incorrect solution. Thus, human observation, which is subjective, is still required to check the registration result. We propose to use a novel image registration quality evaluator (RQE) to automatically identify misregistrations as part of an algorithm-based decision-making process for verification of patient positioning. A RQE, based on an adaptive pattern classifier, is generated from a pair of reference and target images to determine the acceptability of a registration solution given an optimization process. Here we applied our RQE to patient positioning for cranial radiation therapy. We constructed two RQEs-one for the evaluation of intramodal registrations (i.e., portal-portal); the other for intermodal registrations (i.e., portal-DRR). Mutual information, because of its high discriminatory ability compared with other measures (i.e., correlation coefficient and partitioned intensity uniformity), was chosen as the test function for both RQEs. We adopted 1 mm translation and 1 degree rotation as the maximal acceptable registration errors, reflecting desirable clinical setup tolerances for cranial radiation therapy. Receiver operating characteristic analysis was used to evaluate the performance of the RQE, including computations of sensitivity and specificity. The RQEs showed very good performance for both intramodal and intermodal registrations using simulated and phantom data. The sensitivity and the specificity were 0.973 and 0.936, respectively, for the intramodal RQE using phantom data. Whereas the sensitivity and the specificity were 0.961 and 0.758, respectively, for the intermodal RQE using phantom data. Phantom experiments also indicated our RQEs detected out-of-plane deviations exceeding 2.5 mm and 2.50. A preliminary retrospective clinical study of the RQE on cranial portal imaging also yielded good sensitivity > or = 0.857) and specificity (> or = 0.987). Clinical implementation of a RQE could potentially reduce the involvement of the human observer for routine patient positioning verification, while increasing setup accuracy and reducing setup verification time.

Dynamic targeting image-guided radiotherapy

Volumetric imaging and planning for 3-dimensional (3D) conformal radiotherapy and intensity-modulated radiotherapy (IMRT) have highlighted the need to the oncology community to better understand the geometric uncertainties inherent in the radiotherapy delivery process, including setup error (interfraction) as well as organ motion during treatment (intrafraction). This has ushered in the development of emerging technologies and clinical processes, collectively referred to as image-guided radiotherapy (IGRT). The goal of IGRT is to provide the tools needed to manage both inter- and intrafraction motion to improve the accuracy of treatment delivery. Like IMRT, IGRT is a process involving all steps in the radiotherapy treatment process, including patient immobilization, computed tomography (CT) simulation, treatment planning, plan verification, patient setup verification and correction, delivery, and quality assurance. The technology and capability of the Dynamic Targeting IGRT system developed by Varian Medical Systems is presented. The core of this system is a Clinac or Trilogy accelerator equipped with a gantry-mounted imaging system known as the On-Board Imager (OBI). This includes a kilovoltage (kV) x-ray source, an amorphous silicon kV digital image detector, and 2 robotic arms that independently position the kV source and imager orthogonal to the treatment beam. A similar robotic arm positions the PortalVision megavoltage (MV) portal digital image detector, allowing both to be used in concert. The system is designed to support a variety of imaging modalities. The following applications and how they fit in the overall clinical process are described: kV and MV planar radiographic imaging for patient repositioning, kV volumetric cone beam CT imaging for patient repositioning, and kV planar fluoroscopic imaging for gating verification. Achieving image-guided motion management throughout the radiation oncology process requires not just a single product, but a suite of integrated products to manipulate all patient data, including images, efficiently and effectively.

Consistency in electronic portal imaging registration in prostate cancer radiation treatment verification

Background

A protocol of electronic portal imaging (EPI) registration for the verification of radiation treatment fields has been implemented at our institution. A template is generated using the reference images, which is then registered with the EPI for treatment verification. This study examines interobserver consistency among trained radiation therapists in the registration and verification of external beam radiotherapy (EBRT) for patients with prostate cancer.

Materials and methods

20 consecutive patients with prostate cancer undergoing EBRT were analyzed. The EPIs from the initial 10 fractions were registered independently by 6 trained radiation therapist observers. For each fraction, an anterior-posterior (AP or PA) and left lateral (Lat) EPIs were generated and registered with the reference images. Two measures of displacement for the AP EPI in the superior-inferior (SI) and right left (RL) directions and two measures of displacement for the Lat EPI in the AP and SI directions were prospectively recorded. A total of 2400 images and 4800 measures were analyzed. Means and standard deviations, as well as systematic and random errors were calculated for each observer. Differences between observers were compared using the chi-square test. Variance components analysis was used to evaluate how much variance is attributed to the observers. Time trends were estimated using repeated measures analysis.

Results

Inter-observer variation expressed as the standard deviation of the six observers' measurements within each image were 0.7, 1.0, 1.7 and 1.4 mm for APLR, APSI, LatAP and LatSI respectively. Variance components analysis showed that the variation attributed to the observers was small compared to variation due to the images. On repeated measure analysis, time trends were apparent only for the APLR and LatSI measurements. Their magnitude however was small.

Conclusion

No clinically important systematic observer effect or time trends were identified in the registration of EPI by the radiation therapist observers in this study. These findings are useful in the documentation of consistency and reliability in the quality assurance of treatment verification of EBRT for prostate cancer.

Observer variability when evaluating patient movement from electronic portal images of pelvic radiotherapy fields

BACKGROUND AND PURPOSE

A study has been performed to evaluate inter-observer variability when assessing pelvic patient movement using an electronic portal imaging device (EPID).

MATERIALS AND METHODS

Four patient image sets were used with 3-6 portal images per set. The observer group consisted of nine radiographers with 3-18 months clinical EPID experience. The observers outlined bony landmarks on a digital simulator image and used matching software to evaluate field placement errors (FPEs) on each portal image relative to the reference simulator image. Data were evaluated statistically, using a two-component analysis of variance technique, to quantify both the inter-observer variability in evaluating FPEs and inter-fraction variability in patient position relative to the residuals of the analysis. Intra-observer variability was also estimated using four of the observers carrying out three sets of repeat readings.

RESULTS

Eight sets of variance data were analysed, based on FPEs in two orthogonal directions for each of the four patient image sets studied. Initial analysis showed that both inter-observer variation and inter-fraction-patient position variation were statistically significant (P<0.05) in seven of the eight cases evaluated. The averaged root-mean-square (RMS) deviation of the observers from the group mean was 1.1 mm, with a maximum deviation of 5.0 mm recorded for an individual observer. After additional training and re-testing of two of the observers who recorded the largest deviations from the group mean, a subsequent analysis showed the inter-observer variability for the group to be significant in only three of the eight cases, with averaged RMS deviation reduced to 0.5 mm, with a maximum deviation of 2.7 mm. The intra-observer variability was 0.5 mm, averaged over the four observers tested.

CONCLUSIONS

We have developed a quantitative approach to evaluate inter-observer variability in terms of its statistical significance compared to inter-fraction patient movement. This will assist us in training and assessing observers required to perform this task on a routine basis.

[Quality control during radiation therapy]

The aim of quality control procedures during radiation therapy is to check the consistency between actual and prescribed treatments. Given the technical complexity of modern radiotherapy, stricter policies are necessary to meet increasing requirements for quality and safety. Among the various tools available, electronic imaging systems play an increasing role in patient-beam position checking and in vivo dose measurements. Written procedures will have to be established in order to describe the control modalities and frequency, as well as the rules for error corrections according to the treatment intent. Non medical staff will be devoted to new tasks, under the radiation oncologist's responsibility. A special attention should be directed at electronic archives, since the present technology is unlikely to meet the legal requirement to keep medical records accessible for at least 30 years.

Electronic and film portal images: a comparison of landmark visibility and review accuracy

PURPOSE

To quantitatively compare a scanning liquid ion chamber electronic portal imaging device (SLIC-EPID) and an amorphous silicon flat panel (aSi) EPID with portal film in clinical applications using measures of landmark visibility and treatment review accuracy.

METHODS AND MATERIALS

Six radiation oncologists viewed 39 electronic portal images (EPIs) from the SLIC-EPID, 36 EPIs from the aSi-EPID, and portal films of each of these treatment fields. The physicians rated the clarity of anatomic landmarks in the portal images, and the scores were compared between EPID and film. Nine hundred portal image reviews were performed. EPID and film portal images were acquired with known setup errors in either phantom or cadaver treatments. Physicians identified the errors visually in portal films and with computerized analysis for EPID.

RESULTS

There were no statistically significant (p < 0.05) differences between film and SLIC-EPID in ratings of landmark clarity. Eleven of 12 landmarks were more visible in aSi-EPID than in film. Translational setup errors were identified with an average accuracy of 2.5 mm in film, compared to 1.5 mm with SLIC-EPID and 1.3 mm with aSi-EPID.

CONCLUSIONS

Both EPIDs are clinically viable replacements for film, but aSi-EPID represents a significant advancement in image quality over film.

Set-up verification using portal imaging; review of current clinical practice

In this review of current clinical practice of set-up error verification by means of portal imaging, we firstly define the various types of set-up errors using a consistent nomenclature. The different causes of set-up errors are then summarized. Next, the results of a large number of studies regarding patient set-up verification are presented for treatments of patients with head and neck, prostate, pelvis, lung and breast cancer, as well as for mantle field/total body treatments. This review focuses on the more recent studies in order to assess the criteria for good clinical practice in patient positioning. The reported set-up accuracy varies widely, depending on the treatment site, method of immobilization and institution. The standard deviation (1 SD, mm) of the systematic and random errors for currently applied treatment techniques, separately measured along the three principle axes, ranges from 1.6-4.6 and 1.1-2.5 (head and neck), 1.0-3.8 and 1.2-3.5 (prostate), 1.1-4.7 and 1.1-4.9 (pelvis), 1.8-5.1 and 2.2-5.4 (lung), and 1.0-4.7 and 1.7-14.4 (breast), respectively. Recommendations for procedures to quantify, report and reduce patient set-up errors are given based on the studies described in this review. Using these recommendations, the systematic and random set-up errors that can be achieved in routine clinical practice can be less than 2.0 mm (1 SD) for head and neck, 2.5 mm (1 SD) for prostate, 3.0 mm (1 SD) for general pelvic and 3.5 mm (1 SD) for lung cancer treatment techniques.

A double exposed portal image comparison between electronic portal imaging hard copies and port films in radiation therapy treatment setup confirmation to determine its clinical application in a radiotherapy center

PURPOSE

At the William Buckland Radiotherapy Center (WBRC), field-only electronic portal image (EPI) hard copies are used for radiation treatment field verification for whole brain, breast, chest, spine, and large pelvic fields, as determined by a previous study. A subsequent research project, addressing the quality of double exposed EPI hard copies for sites where field only EPI was not considered adequate to determine field placement, has been undertaken. The double exposed EPI hard copies were compared to conventional double exposed port films for small pelvic, partial brain, and head and neck fields and for a miscellaneous group.

METHODS AND MATERIALS

All double exposed EPIs were captured during routine clinical procedures using liquid ion chamber cassettes. EPI hard copies were generated using a Visiplex multi-format camera. In sites where port film remained the preferred verification format, the port films were generated as per department protocol. In addition EPIs were collected specifically for this project. Four radiation oncologists performed the evaluation of EPI and port film images independently with a questionnaire completed at each stage of the evaluation process to assess the following: Adequacy of information in the image to assess field placement. Adequacy of information for determining field placement correction. Clinician's preferred choice of imaging for field placement assessment

RESULTS

The results indicate that double exposed EPI hard copies generally do containsufficient information to permit evaluation of field placement and can replace conventionaldouble exposed port films in a significant number of sites. These include the following:pelvis fields < 12 X 12 cm, partial brain fields, and a miscellaneous group. However forradical head and neck fields, the preferred verification image format remained port film dueto the image hard copy size and improved contrast for this media. Thus in this departmenthard copy EPI is the preferred modality of field verification for all sites except radical headand neck treatments. This should result in an increase in efficiency of workloadmanagement and patient care.

Subjectivity in interpretation of portal films

PURPOSE

We have measured the variability in identifying geometric errors and the variability in clinical decision making when using portal films.

METHODS AND MATERIALS

Eight observers (four radiation oncologists and four radiation therapists) viewed 40 film pairs from 40 different patients. All films, which were acquired using conventional simulator and portal film cassettes, were selected retrospectively from a large clinical database. The observers compared the simulator and portal films under standard conditions, identified the field placement errors, and decided whether adjustments in treatment were required. In addition, all films were digitized and the field placement errors were measured objectively using image registration software.

RESULTS

There was much variability in identifying field placement errors and even more variability in the number of recommended adjustments. The field placement errors identified by the different observers differed by up to 50 mm for the same film pair. The number of adjustments of treatment or block position recommended by the observers also varied between 8 and 25 for the same set of films. The average field placement error, before correction, for AP lung, AP pelvis, and lateral pelvis films was 5.7 mm, 6.3 mm, and 8.9 mm, while the average error after correction (i.e., correcting all errors identified by the observers) was 5.5 mm, 4.9 mm, and 5.7 mm, respectively. Thus, for lateral pelvis films, where the initial errors were larger, the observers were able to make an improvement in patient setup.

CONCLUSIONS

The results suggest that human observers have difficulty identifying field placement errors accurately when the errors are around 5 mm or smaller. Although there is some evidence that experience influenced the performance of the observers, the effect of experience is not large. In routine clinical environments, the use of visual inspection will detect large field placement errors. However, tools other than visual inspection will be required if field placement errors 5 mm or smaller are to be identified accurately.

Computer-assisted decision making in portal verification--optimization of the neural network approach

PURPOSE

Conventional portal verification requires that a qualified radiation oncologist make decisions as to the set-up acceptability. This scheme is no longer sustainable with the large numbers of images available on-line and stringent time constraints. Therefore the objective of this study was to develop, optimize, and evaluate on clinical data an artificial intelligence decision-making tool for portal verification. The tool, based on the artificial neural network (ANN) approach, should approximate, as closely as possible, portal verification assessments made by a radiation oncologist expert.

METHODS AND MATERIALS

A total of 328 electronic portal images of tangential breast irradiations were included in the study. A radiation oncologist expert evaluated these images and rated the treatment set-up acceptability on a scale from 0 to 10. Translational and rotational errors in the placement of the radiation field boundaries formed seven-dimensional feature vectors that represented each of the 328 portal images/treatments. The feature vectors were used as inputs to a three-layer, feedforward ANN. The neural network was trained on the oncologist's ratings.

RESULTS

The rms discrepancy between the ANN and the expert's ratings was 1.05 rating points. Using the decision threshold equal to 5 for both sets of ratings, the ANN classifier was capable of detecting 100% of the portals classified as "unacceptable" by the expert. Only 6.5% of portals acceptable to the oncologist were misclassified as "unacceptable" by the ANN.

CONCLUSION

The results of this study indicate the feasibility of using the ANN portal image classifier as an automated assistant to the radiation oncologist. Its role would be to recommend an appropriate decision as to the acceptability or otherwise of a given treatment set-up depicted in a portal image.

Application of a fuzzy pattern classifier to decision making in portal verification of radiotherapy

With the large volume of electronic portal images acquired and stringent time constraints, it is no longer feasible to follow the convention whereby the radiation oncologist reviews and approves or rejects all portals. For that purpose we have developed a portal image classifier based on the fuzzy k-nearest neighbour (k-NN) algorithm. Each portal image is represented by a feature vector that consists of translational and rotational errors in the placement of radiation field borders that were measured in the portal image. Memberships in the acceptable portal class for the reference portal images within a training dataset were defined by a radiation oncologist expert. The fuzzy k-NN portal image classifier was trained and tested on a dataset of 328 portal images acquired during tangential irradiations of the breast. The memberships in the acceptable portal class produced by the fuzzy k-NN algorithm agreed very well with those defined by the expert. The linear correlation coefficient was equal to 0.89. Performance of the fuzzy k-NN classifier was also evaluated from the portal decision-making point of view using the measures of accuracy, sensitivity and specificity. The fuzzy k-NN portal classifier was capable of identifying almost all the truly unacceptable portals with an acceptably low false alarm rate.

An image registration scheme applied to verification of radiation therapy

The introduction of modern conformal radiation therapy techniques requires high geometric precision in treatment delivery which must be verified. For that purpose we have developed an automated system based on registration of portal and simulation (or planning) image pairs. The image registration is performed on anatomical features which are automatically extracted from the portal image. The portal image is then registered with a planning or simulation radiographic image which represents the geometric prescription for the treatment, using an optimized version of the chamfer matching algorithm. Subsequently, the magnitude of the radiation field displacement during treatment is measured by registering the prescribed and treated field boundaries. Algorithms based on chamfer matching and polygon matching have been used for the field boundary registration. Performance of the entire scheme was evaluated on a series of 15 portal images of a pelvic phantom representing various known degrees of the radiation field displacement. The measurements of the radiation field displacements performed by the automated system proved very reliable and after correction for systematic bias agreed to within 1.5 mm or 1 degree with the displacements applied. Second test series involved comparisons between the automated registrations and those performed manually/visually by an experienced human observer, on 31 portal images acquired during treatments of 18 pelvic patients. These tests showed close agreement (in 80% of cases discrepancies were smaller than 1.5 mm or 1.5 degrees) between the automated scheme and the human observer. It is concluded that the developed scheme would be suitable for online geometric verification of radiation therapy treatments.

Quantitative vs. subjective portal verification using digital portal images

PURPOSE

Off-line, computer-aided prescription (simulator) and treatment (portal) image registration using chamfer matching has been implemented on PC based viewing station. The purposes of this study were (a) to evaluate the performance of interactive anatomy and field edge extraction and subsequent registration, and (b) to compare observer's perceptions of field accuracy with measured discrepancies following anatomical registration.

METHODS AND MATERIALS

Prescription-treatment image pairs for 48 different patients were examined in this study. Digital prescription images were produced with the aid of a television camera and a digital frame grabber, while the treatment images were obtained directly from an on-line portal imaging system. To facilitate perception of low contrast anatomical detail, on-line portal images were enhanced with selective adaptive histogram equalization prior to extraction of anatomical edges. Following interactive extraction of anatomical and field border information by an experienced observer, the identified anatomy was registered using chamfer matching. The degree of conformity between the prescription and treatment fields was quantified using several parameters, which included relative prescription field coverage and overcoverage, as well as the translational and rotational displacements as measured by chamfer matching applied to the boundaries of the two fields. These quantitative measures were compared with subjective evaluations made by four radiation oncologists.

RESULTS

All the images in this series that included a range of the most commonly seen treatment sites were registered and the conformity parameters were found. The mean treatment/prescription field coverage and overcoverage were approximately 95 and 7%, respectively before registration. The mean translational displacement in the transverse and cranio-caudal directions were 2.9 and 3.4 mm, respectively. The mean rotational displacement was approximately 2 degrees. For all four oncologists, the portals classified as unacceptable, in terms of the field placement, exhibited significantly higher (p < 0.03) translational errors in the transverse direction. The field coverages were significantly lower (p < 0.05) and the translational errors in the cranio-caudal direction were significantly higher (p < 0.05) for the portals rated as unacceptable by two of the oncologists.

CONCLUSIONS

From the parameters that were used to quantify the degree of conformity between the prescription and treatment fields, the translational error in the transverse direction correlated best with the oncologists' assessments on the field placement. Field coverage and translational error in the cranio-caudal direction correlated well with assessments of only two out of the four participating oncologists. This can be explained by the fact that for the majority of treatment sites included in the study the positioning of field borders was more critical for the transverse direction. A conclusion for the design of future quantitative and automated on-line portal verification systems is that they will have to model different perceived significances of different types of localization errors intrinsic to oncologist evaluation of portal images.

A diagnostic-quality electronic portal imaging system

Initial clinical experience is presented, on the use of a prototype portal imaging system which is designed to provide diagnostic-quality images on-line at the accelerator. The system comprises a compact diagnostic X-ray unit mounted on the accelerator head, with its source in the isocentric plane exactly 37 degrees around from the therapy source. The image detector is an image intensifier with digital image storage/processing facilities. Images were taken at the accelerator of treatment fields in seven patients (pelvic, head and neck, chest fields) and these images were comparable in anatomical contrast to simulator radiographs. Two techniques for marking the therapy field onto the portal images were successfully demonstrated. One was to mark relevant corners or edges of the therapy light field on the skin using small Pb markers which become clearly visible in the 'diagnostic' portal image. The other was to record a separate 'therapy' image and through the software, extract and superimpose the field edges on to the 'diagnostic' image. The system and method proved fundamentally sound on criteria of image quality, geometric precision of rotation between therapy and imaging conditions, and its potential for development as a practical clinical tool.