Core-based portal image registration for automatic radiotherapy treatment verification

Innovations in Three-Dimensional Treatment Planning and Quality Assurance

Radiation therapy treatment planning and treatment delivery are in the process of changing dramatically over the next several years. This change has been driven in large part by continued advances in computer hardware and software and in medical imaging. Three-dimensional radiation treatment planning systems are rapidly being implemented in clinics around the world. These developments in turn have prompted manufacturers to employ advanced microcircuitry and computer technology to produce treatment delivery systems capable of precise shaping of dose distributions via computer-controlled multileaf collimators which cause the beam intensity to be varied across the beam. Image-based 3D planning and beam intensity modulated delivery systems show significant potential for improving the quality of radiotherapy and improving the efficiency with which radiation therapy can be planned and delivered. However, significant research and development work on these systems and their clinical use remains to be performed. The techniques used for the treatment planning and the methods used for quality assurance procedures and testing must all be revised and/or redesigned to allow efficient clinical use of these technological advances. Although much of the current 3D radiation therapy process requires interactive tasks (and some still very laborious) the path is clear toward solving the technological obstacles so that a nearly automated planning, delivery, and verification system will become a reality over the next decade. Such systems will allow radiation oncologists to significantly increase dose to many tumor sites while concomitantly lowering doses to critical organs-at-risk. Most of the tasks will be automated, thus lowering the overall costs currently needed to provide high-quality external beam radiation therapy.

A frequency-based approach to locate common structure for 2D-3D intensity-based registration of setup images in prostate radiotherapy

In many radiotherapy clinics, geometric uncertainties in the delivery of 3D conformal radiation therapy and intensity modulated radiation therapy of the prostate are reduced by aligning the patient's bony anatomy in the planning 3D CT to corresponding bony anatomy in 2D portal images acquired before every treatment fraction. In this paper, we seek to determine if there is a frequency band within the portal images and the digitally reconstructed radiographs (DRRs) of the planning CT in which bony anatomy predominates over non-bony anatomy such that portal images and DRRs can be suitably filtered to achieve high registration accuracy in an automated 2D-3D single portal intensity-based registration framework. Two similarity measures, mutual information and the Pearson correlation coefficient were tested on carefully collected gold-standard data consisting of a kilovoltage cone-beam CT (CBCT) and megavoltage portal images in the anterior-posterior (AP) view of an anthropomorphic phantom acquired under clinical conditions at known poses, and on patient data. It was found that filtering the portal images and DRRs during the registration considerably improved registration performance. Without filtering, the registration did not always converge while with filtering it always converged to an accurate solution. For the pose-determination experiments conducted on the anthropomorphic phantom with the correlation coefficient, the mean (and standard deviation) of the absolute errors in recovering each of the six transformation parameters were Theta(x):0.18(0.19) degrees, Theta(y):0.04(0.04) degrees, Theta(z):0.04(0.02) degrees, t(x):0.14(0.15) mm, t(y):0.09(0.05) mm, and t(z):0.49(0.40) mm. The mutual information-based registration with filtered images also resulted in similarly small errors. For the patient data, visual inspection of the superimposed registered images showed that they were correctly aligned in all instances. The results presented in this paper suggest that robust and accurate registration can be achieved with intensity-based methods by focusing on rigid bony structures in the images while diminishing the influence of artifacts with similar frequencies as soft tissue.

Evaluation of whole-body MR to CT deformable image registration

Multimodality image registration plays a crucial role in various clinical and research applications. The aim of this study is to present an optimized MR to CT whole‐body deformable image registration algorithm and its validation using clinical studies. A 3D intermodality registration technique based on B‐spline transformation was performed using optimized parameters of the elastix package based on the Insight Toolkit (ITK) framework. Twenty‐eight (17 male and 11 female) clinical studies were used in this work. The registration was evaluated using anatomical landmarks and segmented organs. In addition to 16 anatomical landmarks, three key organs (brain, lungs, and kidneys) and the entire body volume were segmented for evaluation. Several parameters — such as the Euclidean distance between anatomical landmarks, target overlap, Dice and Jaccard coefficients, false positives and false negatives, volume similarity, distance error, and Hausdorff distance — were calculated to quantify the quality of the registration algorithm. Dice coefficients for the majority of patients (>75%) were in the 0.8–1 range for the whole body, brain, and lungs, which satisfies the criteria to achieve excellent alignment. On the other hand, for kidneys, Dice coefficients for volumes of 25% of the patients meet excellent volume agreement requirement, while the majority of patients satisfy good agreement criteria (>0.6). For all patients, the distance error was in 0–10 mm range for all segmented organs. In summary, we optimized and evaluated the accuracy of an MR to CT deformable registration algorithm. The registered images constitute a useful 3D whole‐body MR‐CT atlas suitable for the development and evaluation of novel MR‐guided attenuation correction procedures on hybrid PET‐MR systems.

PACS number: 07.05.Pj

Automated 2D-3D registration of portal images and CT data using line-segment enhancement

In prostate radiotherapy, setup errors with respect to the patient's bony anatomy can be reduced by aligning 2D megavoltage (MV) portal images acquired during treatment to a reference 3D kilovoltage (kV) CT acquired for treatment planning purposes. The purpose of this study was to evaluate a fully automated 2D-3D registration algorithm to quantify setup errors in 3D through the alignment of line-enhanced portal images and digitally reconstructed radiographs computed from the CT. The line-enhanced images were obtained by correlating the images with a filter bank of short line segments, or "sticks" at different orientations. The proposed methods were validated on (1) accurately collected gold-standard data consisting of a 3D kV cone-beam CT scan of an anthropomorphic phantom of the pelvis and 2D MV portal images in the anterior-posterior (AP) view acquired at 15 different poses and (2) a conventional 3D kV CT scan and weekly 2D MV AP portal images of a patient over 8 weeks. The mean (and standard deviation) of the absolute registration error for rotations around the right-lateral (RL), inferior-superior (IS), and posterior-anterior (PA) axes were 0.212 degree (0.214 degree), 0.055 degree (0.033 degree) and 0.041 degree (0.039 degree), respectively. The corresponding registration errors for translations along the RL, IS, and PA axes were 0.161 (0.131) mm, 0.096 (0.033) mm, and 0.612 (0.485) mm. The mean (and standard deviation) of the total registration error was 0.778 (0.543) mm. Registration on the patient images was successful in all eight cases as determined visually. The results indicate that it is feasible to automatically enhance features in MV portal images of the pelvis for use within a completely automated 2D-3D registration framework for the accurate determination of patient setup errors. They also indicate that it is feasible to estimate all six transformation parameters from a 3D CT of the pelvis and a single portal image in the AP view.

ESTERR-PRO: a setup verification software system using electronic portal imaging

The purpose of the paper is to present and evaluate the performance of a new software-based registration system for patient setup verification, during radiotherapy, using electronic portal images. The estimation of setup errors, using the proposed system, can be accomplished by means of two alternate registration methods. (a) The portal image of the current fraction of the treatment is registered directly with the reference image (digitally reconstructed radiograph (DRR) or simulator image) using a modified manual technique. (b) The portal image of the current fraction of the treatment is registered with the portal image of the first fraction of the treatment (reference portal image) by applying a nearly automated technique based on self-organizing maps, whereas the reference portal has already been registered with a DRR or a simulator image. The proposed system was tested on phantom data and on data from six patients. The root mean square error (RMSE) of the setup estimates was 0.8 ± 0.3 (mean value ± standard deviation) for the phantom data and 0.3 ± 0.3 for the patient data, respectively, by applying the two methodologies. Furthermore, statistical analysis by means of the Wilcoxon nonparametric signed test showed that the results that were obtained by the two methods did not differ significantly (P value > 0.05).

Automated 2D-3D registration of a radiograph and a cone beam CT using line-segment enhancement

The objective of this study was to develop a fully automated two-dimensional (2D)-three-dimensional (3D) registration framework to quantify setup deviations in prostate radiation therapy from cone beam CT (CBCT) data and a single AP radiograph. A kilovoltage CBCT image and kilovoltage AP radiograph of an anthropomorphic phantom of the pelvis were acquired at 14 accurately known positions. The shifts in the phantom position were subsequently estimated by registering digitally reconstructed radiographs (DRRs) from the 3D CBCT scan to the AP radiographs through the correlation of enhanced linear image features mainly representing bony ridges. Linear features were enhanced by filtering the images with "sticks," short line segments which are varied in orientation to achieve the maximum projection value at every pixel in the image. The mean (and standard deviations) of the absolute errors in estimating translations along the three orthogonal axes in millimeters were 0.134 (0.096) AP(out-of-plane), 0.021 (0.023) ML and 0.020 (0.020) SI. The corresponding errors for rotations in degrees were 0.011 (0.009) AP, 0.029 (0.016) ML (out-of-plane), and 0.030 (0.028) SI (out-of-plane). Preliminary results with megavoltage patient data have also been reported. The results suggest that it may be possible to enhance anatomic features that are common to DRRs from a CBCT image and a single AP radiography of the pelvis for use in a completely automated and accurate 2D-3D registration framework for setup verification in prostate radiotherapy. This technique is theoretically applicable to other rigid bony structures such as the cranial vault or skull base and piecewise rigid structures such as the spine.

A gradient feature weighted Minimax algorithm for registration of multiple portal images to 3DCT volumes in prostate radiotherapy

PURPOSE

To develop an accurate, fast, and robust algorithm for registering portal and computed tomographic (CT) images for radiotherapy using a combination of sparse and dense field data that complement each other.

METHODS AND MATERIALS

Gradient Feature Weighted Minimax (GFW Minimax) method was developed to register multiple portal images to three-dimensional CT images. Its performance was compared with that of three others: Minimax, Mutual Information, and Gilhuijs' method. Phantom and prostate cancer patient images were used. Effects of registration errors on tumor control probability (TCP) and normal tissue complication probability (NTCP) were investigated as a relative measure.

RESULTS

Registration of four portals to CTs resulted in 30% lower error when compared with registration with two portals. Computation time increased by nearly 50%. GFW Minimax performed the best, followed by Gilhuijs' method, the Minimax method, and Mutual Information.

CONCLUSIONS

Using four portals instead of two lowered the registration error. Reduced fields of view images with full feature sets gave similar results in shorter times as full fields of view images. In clinical situations where soft tissue targets are of importance, GFW Minimax algorithm was significantly more accurate and robust. With registration errors lower than 1 mm, margins may be scaled down to 4 mm without adversely affecting TCP and NTCP.

A novel method for automatic detection of patient out-of-plane rotation by comparing a single portal image to a reference image

A novel method for detecting out-of-plane patient rotation by comparing a single portal image to its reference image is presented. Out-of-plane rotation results in an apparent distortion of the anatomy in a portal image. This distortion can be mathematically predicted with the magnification varying at each point in the image. While scaling of points at equal depth is invariant under in-plane rotation or translation, and changes equally in both dimensions for an axial shift of the patient, a change of scaling in only one dimension can be ascribed to an out-of-plane rotation. For the two conditions that are used in this study, it is shown that out-of-plane rotation yields a different scaling of the image in two perpendicular directions and therefore it is feasible to calculate the scale factors as a function of out-of-plane rotation. Conversely the recovery of scale factors in two different directions at the same time would enable the magnitude of the out-of-plane rotation to be recovered. The properties of the Fourier transform of the image are used to align the portal image with the reference image (a simulator image or first approved portal image) prior to the recovery of the scale factors. Correlating the Fourier transform of the portal image on a log-scale with that of the reference image enables the scale factors to be automatically extracted from a single portal image. In the two approaches investigated, out-of-plane rotations of up to 41 degrees and 20 degrees (respectively) have been recovered with a maximum error of 2.4 degrees. This technique could be used to automatically detect patient roll or tilt prior to or during a treatment session.

Clinical use of electronic portal imaging

Accurate and routine target localization is necessary for successful outcome in radiation therapy treatments. Electronic portal imaging devices (EPIDs) provide an advanced tool with digital technology to improve target localization and maintain clinical efficiency. EPIDs are ubiquitous in the radiation therapy clinic, and they provide a powerful and flexible tool to collect and process data in a quantitative manner to improve treatment accuracy for virtually any treatment site. This manuscript provides an overview of the clinical implementation process for effective use of EPIDs. It continues with a review of correction strategies and finally highlights numerous examples of effective clinical application of EPID.

Anatomic feature-based registration for patient set-up in head and neck cancer radiotherapy

Modern radiotherapy equipment is capable of delivering high precision conformal dose distributions relative to isocentre. One of the barriers to precise treatments is accurate patient re-positioning before each fraction of treatment. At Massachusetts General Hospital, we perform daily patient alignment using radiographs, which are captured by flat panel imaging devices and sent to an analysis program. A trained therapist manually selects anatomically significant features in the skeleton, and couch movement is computed based on the image coordinates of the features. The current procedure takes about 5 to 10 min and significantly affects the efficiency requirement in a busy clinic. This work presents our effort to develop an improved, semi-automatic procedure that uses the manually selected features from the first treatment fraction to automatically locate the same features on the second and subsequent fractions. An implementation of this semi-automatic procedure is currently in clinical use for head and neck tumour sites. Radiographs collected from 510 patient set-ups were used to test this algorithm. A mean difference of 1.5 mm between manual and automatic localization of individual features and a mean difference of 0.8 mm for overall set-up were seen.

Registration of electronic portal images for patient set-up verification

Images acquired from an electronic portal imaging device are aligned with digitally reconstructed radiographs (DRRs) or other portal images to verify patient positioning during radiation therapy. Most of the currently available computer aided registration methods are based on the manual placement of corresponding landmarks. The purpose of the paper is twofold: (a) the establishment of a methodology for patient set-up verification during radiotherapy based on the registration of electronic portal images, and (b) the evaluation of the proposed methodology in a clinical environment. The estimation of set-up errors, using the proposed methodology, can be accomplished by matching the portal image of the current fraction of the treatment with the portal image of the baseline treatment (reference portal image) using a nearly automated technique. The proposed registration method is tested on a number of phantom data as well as on data from four patients. The phantom data included portal images that corresponded to various positions of the phantom on the treatment couch. For each patient, a set of 30 portal images was used. For the phantom data (for both transverse and lateral portal images), the maximum absolute deviations of the translational shifts were within 1.5 mm, whereas the in-plane rotation angle error was less than 0.5 degrees. The two-way Anova revealed no statistical significant variability both within observer and between-observer measurements (P > 0.05). For the patient data, the mean values obtained with manual and the proposed registration methods were within 0.5 mm. In conclusion, the proposed registration method has been incorporated within a system, called ESTERR-PRO. Its image registration capability achieves high accuracy and both intra- and inter-user reproducibility. The system is fully operational within the Radiotherapy Department of 'HYGEIA' Hospital in Athens and it could be easily installed in any other clinical environment since it requires standardized hardware specifications and minimal human intervention.

A new algorithm for the registration of portal images to planning images in the verification of radiotherapy, as validated in prostate treatments

The manual verification of a radiotherapy treatment, where a portal image is matched onto a planning image, is very time consuming and subject to inter- and intraobserver variability. Therefore, a fully automatic matching procedure (image registration) is required. Existing automatic matching algorithms are confounded, however, by irrelevant information in the portal images (i.e., air in the intestines). Therefore, we have developed a new method, which is an extension of chamfer matching and uses, apart from the distance to the nearest edge, additional information on the correspondence of the gradient angle and magnitude of the edges, making the method less sensitive to confounding information in the images. To validate the automatic matching procedure in clinical practice, we applied the new method on 157 images of 29 randomly selected patients treated for carcinoma of the prostate. Three experts manually matched these images in consensus. Subsequently, the same observers assessed the results of the automatic registration. When regular chamfer matching is used for the fully automatic matching procedure, only 5% of the image pairs could be matched correctly, whereas the new method successfully registered 80% by using additional information on the angle of the edges. From the results of the validation study it can be concluded that a significant reduction in workload for the physicians and technicians can be achieved with this method.

Entropy-based dual-portal-to-3-DCT registration incorporating pixel correlation

For patient setup verification in external beam radiotherapy (EBRT) of prostate cancer, we developed an information theoretic registration framework, called the minimax entropy registration framework, to simultaneously and iteratively segment portal images and register them to three-dimensional (3-D) computed tomography (CT) image data. The registration framework has two steps, the max step and the min step, and evaluates appropriate entropies to estimate segmentations of the portal images and to find the transformation parameters. In the initial version of the algorithm (Bansal et al. 1999), we assumed image pixels to be independently distributed, an assumption not true in general. Thus, to better segment the portal images and to improve the accuracy of the estimated registration parameters, in this initial formulation of the problem, the correlation among pixel intensities is modeled using a one-dimensional Markov random process. Line processes are incorporated into the model to improve the estimation of segmentation of the portal images. In the max step, the principle of maximum entropy is invoked to estimate the probability distribution on the segmentations. The estimated distribution is then incorporated into the min step to estimate the registration parameters. Performance of the proposed framework is evaluated and compared to that of a mutual information-based registration algorithm using both simulated and real patient data. In the proposed registration framework, registration of the 3-D CT image and the portal images is guided by an estimated segmentation of the pelvic bone. However, as the prostate can move with respect to the pelvic structure, further localization of the prostate using ultrasound image data is required, an issue to be further explored in future.

Thresholds for human detection of patient setup errors in digitally reconstructed portal images of prostate fields

PURPOSE

Computer-assisted methods to analyze electronic portal images for the presence of treatment setup errors should be studied in controlled experiments before use in the clinical setting. Validation experiments using images that contain known errors usually report the smallest errors that can be detected by the image analysis algorithm. This paper offers human error-detection thresholds as one benchmark for evaluating the smallest errors detected by algorithms. Unfortunately, reliable data are lacking describing human performance. The most rigorous benchmarks for human performance are obtained under conditions that favor error detection. To establish such benchmarks, controlled observer studies were carried out to determine the thresholds of detectability for in-plane and out-of-plane translation and rotation setup errors introduced into digitally reconstructed portal radiographs (DRPRs) of prostate fields.

METHODS AND MATERIALS

Seventeen observers comprising radiation oncologists, radiation oncology residents, physicists, and therapy students participated in a two-alternative forced choice experiment involving 378 DRPRs computed using the National Library of Medicine Visible Human data sets. An observer viewed three images at a time displayed on adjacent computer monitors. Each image triplet included a reference digitally reconstructed radiograph displayed on the central monitor and two DRPRs displayed on the flanking monitors. One DRPR was error free. The other DRPR contained a known in-plane or out-of-plane error in the placement of the treatment field over a target region in the pelvis. The range for each type of error was determined from pilot observer studies based on a Probit model for error detection. The smallest errors approached the limit of human visual capability. The observer was told what kind of error was introduced, and was asked to choose the DRPR that contained the error. Observer decisions were recorded and analyzed using repeated-measures analysis of variance.

RESULTS

The thresholds of detectability averaged over all observers were approximately 2.5 mm for in-plane translations, 1.6 degrees for in-plane rotations, 1 degrees for out-of-plane rotations, and 8% change in magnification for out-of-plane translations along the central axis. When one inexperienced observer is excluded, the average threshold for change in magnification is 5%. Experienced observers tended to perform better, but differences between groups were not statistically significant. Thresholds were computed as averages over all observers. Because of the broad range of observer capabilities, some detection tasks were too difficult for some observers, leading to missing threshold values in our data analysis. The missing values were excluded from computation of the average thresholds reported above. The effect of the missing values is to bias the average values toward the best human performance.

CONCLUSIONS

Under favorable conditions, humans can detect small errors in setup geometry. The thresholds for error detection reported in this study are believed to represent rigorous but reasonable benchmarks that can be incorporated into studies evaluating algorithms for computer-assisted detection of setup errors in electronic portal images.

Portal imaging

Portal imaging is the acquisition of images with a radiotherapy beam. Imaging theory suggests that the quality of portal images could be much higher if the efficiency of the imaging media in detecting radiation could be improved. Introduction of new media (films and electronic portal imaging devices) has confirmed this by markedly increasing the quality of portal images. Images from these devices can then be used to verify a patient's treatment. Geometric verification requires the portal image to be registered with a reference image. Dosimetric verification requires the portal imager to be calibrated for dose. This review gives a brief overview of the current areas of interest in portal imaging: imaging theory; imaging media, film and electronic portal imaging devices; image registration; and dosimetry using these devices.

Clinical use of electronic portal imaging: report of AAPM Radiation Therapy Committee Task Group 58

AAPM Task Group 58 was created to provide materials to help the medical physicist and colleagues succeed in the clinical implementation of electronic portal imaging devices (EPIDs) in radiation oncology. This complex technology has matured over the past decade and is capable of being integrated into routine practice. However, the difficulties encountered during the specification, installation, and implementation process can be overwhelming. TG58 was charged with providing sufficient information to allow the users to overcome these difficulties and put EPIDs into routine clinical practice. In answering the charge, this report provides; comprehensive information about the physics and technology of currently available EPID systems; a detailed discussion of the steps required for successful clinical implementation, based on accumulated experience; a review of software tools available and clinical use protocols to enhance EPID utilization; and specific quality assurance requirements for initial and continuing clinical use of the systems. Specific recommendations are summarized to assist the reader with successful implementation and continuing use of an EPID.

A prototype 3D CT extension for radiotherapy simulators

In this paper we present a report on our custom 3D CT extension capable of producing fully 3D tomographic studies from conventional radiotherapy simulator cone-beam fluoro output. The extension consists of a common PC system on which a proprietary software toolkit provides the appropriate environment for reconstruction and acquisition of cone-beam data provided by commercial simulator fluoro video chains. The extension may compare favorably with CT options offered by simulator manufacturers: in particular, multi-slice single-scan reconstruction seems to be achievable while the current commercial solutions are based on single-slice single-scan. Radiotherapy applications include on-line treatment planning, patient treatment verification and registration.

Guide to clinical use of electronic portal imaging

The Electronic Portal Imaging Device (EPID) provides localization quality images and computer-aided analysis, which should in principal, replace portal film imaging. Modern EPIDs deliver superior image quality and an array of analysis tools that improve clinical decision making. It has been demonstrated that the EPID can be a powerful tool in the reduction of treatment setup errors and the quality assurance and verification of complex treatments. However, in many radiation therapy clinics EPID technology is not in routine clinical use. This low utilization suggests that the capability and potential of the technology alone do not guarantee its full adoption. This paper addresses basic considerations required to facilitate clinical implementation of the EPID technology and gives specific examples of successful implementations.

Decision-making models in the analysis of portal films: a clinical pilot study

Portal films continue to play an important role in the verification of radiotherapy treatment. There is still some discussion, however, as to what action should be taken after a port film has shown a radiation field deviation from the prescribed volume. It was the aim of the present pilot study to investigate the performance of three decision-making models ('Amsterdam', 'Quebec' and 'Newcastle') and an expert panel basing their decision on intuition rather than formal rules after portal film acquisition in a clinical setting. Portal films were acquired on every day during the first week of treatment for five head and neck and five prostate cancer patients (diagnostic phase). If required, the field position was modified according to our normal practice following the recommendation of the expert panel. In order to analyse the results of the models, however, additional port films were taken in the following 3 treatment weeks with the patient moved as required by the different models (intervention phase). The portal films were taken over 4 consecutive days, positioning the patient according to each of the different models on one day each. None of the models diagnosed a field misplacement in the head and neck patients, while the 'Amsterdam' and 'Quebec' models predicted a move in one prostate patient. The 'Newcastle' model, which is based on Hotelling's T2 statistic, proved to be more sensitive and diagnosed a systematic displacement for three prostate patients. The intervention phase confirmed the diagnosis of the model, even if the three portal films taken with the patient position adjusted as required by the model proved to be insufficient to demonstrate an improvement. The 'Newcastle' model does not rely on assumptions about the random movement of patients and requires five portal films before a decision can be reached. This approach lends itself well to incorporation into electronic portal imaging 'packages', where repeated image acquisitions present no logistical difficulty.

Automatic on-line electronic portal image analysis with a wavelet-based edge detector

A fully automatic method for on-line electronic portal image analysis is proposed. The method uses multiscale edge detection with wavelets for both the field outline and the anatomical structures. An algorithm to extract and combine the information from different scales has been developed. The edges from the portal image are aligned with the edges from the reference image using chamfer matching. The reference is the first portal image of each treatment. The matching is applied first to the field and subsequently to the anatomy. The setup deviations are quantified as the displacement of the anatomical structures relative to the radiation beam boundaries. The performance of the algorithm was investigated for portal images with different contrast and noise level. The automatic analysis was used first to detect simulated displacements. Then the automatic procedure was tested on anterior-posterior and lateral portal images of a pelvic phantom. In both sets of tests the differences between the measured and the actual shifts were used to quantify the performance. Finally we applied the automatic procedure to clinical images of pelvic and lung regions. The output of the procedure was compared with the results of a manual match performed by a trained operator. The errors for the phantom tests were small: average standard deviation of 0.39 mm and 0.26 degrees and absolute mean error of 0.31 mm and 0.2 degrees were obtained. In the clinical cases average standard deviations of 1.32 mm and 0.6 degrees were found. The average absolute mean errors were 1.09 mm and 0.39 degrees. Failures were registered in 2% of the phantom tests and in 3% of the clinical cases. The algorithm execution is approximately 5 s on a 168 MHz Sun Ultra 2 workstation. The automatic analysis tool is considered to be a very useful tool for on-line setup corrections.

Segmentation, registration, and measurement of shape variation via image object shape

A model of object shape by nets of medial and boundary primitives is justified as richly capturing multiple aspects of shape and yet requiring representation space and image analysis work proportional to the number of primitives. Metrics are described that compute an object representation's prior probability of local geometry by reflecting variabilities in the net's node and link parameter values, and that compute a likelihood function measuring the degree of match of an image to that object representation. A paradigm for image analysis of deforming such a model to optimize a posteriori probability is described, and this paradigm is shown to be usable as a uniform approach for object definition, object-based registration between images of the same or different imaging modalities, and measurement of shape variation of an abnormal anatomical object, compared with a normal anatomical object. Examples of applications of these methods in radiotherapy, surgery, and psychiatry are given.

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.

Comparison of computer workstation with light box for detecting setup errors from portal images

PURPOSE

Observer studies were conducted to test the hypothesis that radiation oncologists using a computer workstation for portal image analysis can detect setup errors at least as accurately as when following standard clinical practice of inspecting portal films on a light box.

METHODS AND MATERIALS

In a controlled observer study, nine radiation oncologists used a computer workstation, called PortFolio, to detect setup errors in 40 realistic digitally reconstructed portal radiograph (DRPR) images. PortFolio is a prototype workstation for radiation oncologists to display and inspect digital portal images for setup errors. PortFolio includes tools for image enhancement; alignment of crosshairs, field edges, and anatomic structures on reference and acquired images; measurement of distances and angles; and viewing registered images superimposed on one another. The test DRPRs contained known in-plane translation or rotation errors in the placement of the fields over target regions in the pelvis and head. Test images used in the study were also printed on film for observers to view on a light box and interpret using standard clinical practice. The mean accuracy for error detection for each approach was measured and the results were compared using repeated measures analysis of variance (ANOVA) with the Geisser-Greenhouse test statistic.

RESULTS

The results indicate that radiation oncologists participating in this study could detect and quantify in-plane rotation and translation errors more accurately with PortFolio compared to standard clinical practice.

CONCLUSIONS

Based on the results of this limited study, it is reasonable to conclude that workstations similar to PortFolio can be used efficaciously in clinical practice.

Portfolio: a prototype workstation for development and evaluation of tools for analysis and management of digital portal images

PURPOSE

The purpose of this investigation was to design and implement a prototype physician workstation, called PortFolio, as a platform for developing and evaluating, by means of controlled observer studies, user interfaces and interactive tools for analyzing and managing digital portal images. The first observer study was designed to measure physician acceptance of workstation technology, as an alternative to a view box, for inspection and analysis of portal images for detection of treatment setup errors.

METHODS AND MATERIALS

The observer study was conducted in a controlled experimental setting to evaluate physician acceptance of the prototype workstation technology exemplified by PortFolio. PortFolio incorporates a windows user interface, a compact kit of carefully selected image analysis tools, and an object-oriented data base infrastructure. The kit evaluated in the observer study included tools for contrast enhancement, registration, and multimodal image visualization. Acceptance was measured in the context of performing portal image analysis in a structured protocol designed to simulate clinical practice. The acceptability and usage patterns were measured from semistructured questionnaires and logs of user interactions.

RESULTS

Radiation oncologists, the subjects for this study, perceived the tools in PortFolio to be acceptable clinical aids. Concerns were expressed regarding user efficiency, particularly with respect to the image registration tools.

CONCLUSIONS

The results of our observer study indicate that workstation technology is acceptable to radiation oncologists as an alternative to a view box for clinical detection of setup errors from digital portal images. Improvements in implementation, including more tools and a greater degree of automation in the image analysis tasks, are needed to make PortFolio more clinically practical.

Automatic setup deviation measurements with electronic portal images for pelvic fields

The purpose of this work was to develop a fully automatic tool for the detection of setup deviation for small pelvic field using, in external beam radiotherapy, an electronic portal imaging device (EPID). The algorithm processes electronic portal images of prostate cancer patients. No fiducial points or user interventions are needed. Deviation measurements are based on bone edge detection performed with Laplacian of a Gaussian (LoG) operator. Two bone edge images are then correlated, one of which is a reference image taken as the first fraction image for the purpose of this study. The electronic portal images (EPI) also show band artefacts which are removed using the morphological top-hat transform. The algorithm was first validated with 59 phantom images acquired in clinical treatment conditions with known displacements. The algorithm was then validated with 79 clinical images where bone contours were delineated manually. For the phantom images, the setup deviations were measured with a absolute mean error of 0.59 mm and 0.47 mm with a standard deviation of 0.64 mm and 0.42 mm, horizontally and vertically, respectively. A second validation was performed using clinical prostate cancer images. The measured patient displacements have an absolute mean error of 0.48 mm and 1.41 mm with a standard deviation of 0.58 mm and 1.30 mm in the X and Y directions, respectively. The algorithm execution time on a SUN workstation is 5 s. This algorithm shows good potential as a setup deviation measurement tool in clinical practice. The possibility of using this algorithm combined with decision rules based on statistical observations is very promising.

A simple algorithm for planar image registration in radiation therapy

A simple algorithm is presented for planar image registration and the method is applied to the simulator and portal image registration for patient setup verification in radiation therapy. Basically, the algorithm follows the concept proposed by Balter et al. [Med. Phys. 19, 329-334 (1992)], which converts the problem of open curve registration into matching a series of points along the curves. Balter's algorithm consists of three steps: (1) to determine a common starting point for each curve pair, (2) acquire two corresponding point sets along each curve, and (3) obtain a global transform matrix by matching two point sets. We integrate all three steps into one simple procedure which fits the sampled points along the intended curve pair by taking the relative path length shift as an independent fitting parameter. After being modified, the algorithm is able to take the different magnification factors of images into account, and it avoids curvature calculations. Numerical simulation as well as clinical and phantom images have been utilized to test the accuracy of the algorithm. The typical errors are less than 1 mm in translation and 1 degree in rotation. We also made a comparison study with the chamfer method. The results of the two methods agree to within 0.5 mm in translation and 0.5 degree in rotation.

Determination of rotations in three dimensions using two-dimensional portal image registration

The relative relationships among anatomic features visualized on planar radiographic images change due to rotations of the patient out of the imaging plane. These changes can be predicted a priori from a three-dimensional radiographic model of the patient. In this study we assess the feasibility of using that information together with a planar image feature alignment tool to account for out-of-plane rotations in the evaluation of subsequent clinical patient images. A series of digitally reconstructed radiographs (DRRs) with known patient rotations was generated from a computed tomography scan of an anthropomorphic head phantom. Fixed anatomic features were extracted, as seen in the DRRs of rotated anatomy and entered into a database. Alignment of features from test radiographs with those from an entry in this database yielded an estimate of rotation out of plane (database entry that resulted in the best fit via planar transformation) along with the planar components of setup errors in the rotated plane. Tests using DRRs and films show that it is possible to select anatomic features in AP skull radiographs with position and orientation sensitive to out-of-plane rotation.

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.

Bayesian registration of models using finite element eigenmodes

This paper is concerned with registering three-dimensional wire-frame organ models. This involves finding correspondences between points on the models of two different examples of the same organ. Such registration is widely used in the processing of medical data; for example in segmentation, or to superimpose functional information on a more detailed structural map. The algorithm described in this paper is based on matching the modes of deformation of organ shapes. Modes with lower spatial frequency characterise large scale organ features whereas small scale variations determine the high frequency modes. First, the organ sizes are normalised using a generalised version of the centroid size metric. The axes of the fundamental frequency modes are then aligned to provide initial rigid-body registration. The registration is refined by matching increasingly high frequency modes using the 'Highest confidence first' algorithm. The matches are evaluated using a Bayesian combination of local prior and likelihood functions. The prior is derived from the Gompertz metric of biological growth and ensures that physically impossible matches are not accepted. The likelihood function is a measure of the similarity between local modal deformation components. The registration algorithm has been applied by the authors in the analysis of three dimensional ultrasound data. Results are presented showing the registration of two liver models derived from 3D ultrasound.

Benchmark test cases for evaluation of computer-based methods for detection of setup errors: realistic digitally reconstructed electronic portal images with known setup errors

PURPOSE

The purpose of this investigation was to develop methods and software for computing realistic digitally reconstructed electronic portal images with known setup errors for use as benchmark test cases for evaluation and intercomparison of computer-based methods for image matching and detecting setup errors in electronic portal images.

METHODS AND MATERIALS

An existing software tool for computing digitally reconstructed radiographs was modified to compute simulated megavoltage images. An interface was added to allow the user to specify which setup parameter(s) will contain computer-induced random and systematic errors in a reference beam created during virtual simulation. Other software features include options for adding random and structured noise, Gaussian blurring to simulate geometric unsharpness, histogram matching with a "typical" electronic portal image, specifying individual preferences for the appearance of the "gold standard" image, and specifying the number of images generated. The visible male computed tomography data set from the National Library of Medicine was used as the planning image.

RESULTS

Digitally reconstructed electronic portal images with known setup errors have been generated and used to evaluate our methods for automatic image matching and error detection. Any number of different sets of test cases can be generated to investigate setup errors involving selected setup parameters and anatomic volumes. This approach has proved to be invaluable for determination of error detection sensitivity under ideal (rigid body) conditions and for guiding further development of image matching and error detection methods. Example images have been successfully exported for similar use at other sites.

CONCLUSIONS

Because absolute truth is known, digitally reconstructed electronic portal images with known setup errors are well suited for evaluation of computer-aided image matching and error detection methods. High-quality planning images, such as the visible human CT scans from the National Library of Medicine, are essential for producing realistic images. Sets of test cases with systematic and random errors in selected setup parameters and anatomic volumes are suitable for use as standard benchmarks by the radiotherapy community. In addition to serving as an aid to research and development, benchmark images may also be useful for evaluation of commercial systems and as part of a quality assurance program for clinical systems. Test cases and software are available upon request.

A portal image alignment and patient setup verification procedure using moments and correlation techniques

The objective of this study was to develop an integrated field shape and patient setup verification procedure for portal images. The proposed procedure used one of the approved portal images as the reference image for automated comparison with subsequent portal images. The first step of the procedure used a moments method to align treatment field boundaries. This step was necessary to compensate for the repositioning error of an image detector and to create a common frame of reference for comparing anatomical shifts relative to the field boundary. At the end of the moments alignment, a moments figure of merit was computed and compared with a pre-established threshold. This verified whether there was a potential shape change in the treatment field. To measure anatomical misalignment, the last step in the procedure was to use a grey-scale image correlation method to align translations and in-plane rotations relative to the anatomy of the reference image. The procedure was shown in phantom studies to err by less than 1 mm when detecting translational shifts and less than 1 degree when detection in-plane rotations. The moments verification method showed a sensitivity of detecting a placement error of 6 mm for a single leaf in a controlled experiment where a multileaf collimator was used for field shaping. The alignment procedure was fast and could be done in less than 12 s on an IBM-compatible 486 personal computer.