Clinical development of a failure detection-based online repositioning strategy for prostate IMRT--experiments, simulation, and dosimetry study

A new method to reconstruct intra-fractional prostate motion in volumetric modulated arc therapy

Intra-fractional motion is a concern during prostate radiation therapy, as it may cause deviations between planned and delivered radiation doses. Because accurate motion information during treatment delivery is critical to address dose deviation, we developed the projection marker matching method (PM³), a novel method for prostate motion reconstruction in volumetric modulated arc therapy. The purpose of this method is to reconstruct in-treatment prostate motion trajectory using projected positions of implanted fiducial markers measured in kV x-ray projection images acquired during treatment delivery. We formulated this task as a quadratic optimization problem. The objective function penalized the distance from the reconstructed 3D position of each fiducial marker to the corresponding straight line, defined by the x-ray projection of the marker. Rigid translational motion of the prostate and motion smoothness along the temporal dimension were assumed and incorporated into the optimization model. We tested the motion reconstruction method in both simulation and phantom experimental studies. We quantified the accuracy using 3D normalized root-mean-square (RMS) error defined as the norm of a vector containing ratios between the absolute RMS errors and corresponding motion ranges in three dimensions. In the simulation study with realistic prostate motion trajectories, the 3D normalized RMS error was on average [Formula: see text] (range from [Formula: see text] to [Formula: see text]). In an experimental study, a prostate phantom was driven to move along a realistic prostate motion trajectory. The 3D normalized RMS error was [Formula: see text]. We also examined the impact of the model parameters on reconstruction accuracy, and found that a single set of parameters can be used for all the tested cases to accurately reconstruct the motion trajectories. The motion trajectory derived by PM³ may be incorporated into novel strategies, including 4D dose reconstruction and adaptive treatment replanning to address motion-induced dose deviation.

Comparison of 2D and 3D modeled tumor motion estimation/prediction for dynamic tumor tracking during arc radiotherapy

Many real-time imaging techniques have been developed to localize a target in 3D space or in a 2D beam's eye view (BEV) plane for intrafraction motion tracking in radiation therapy. With tracking system latency, the 3D-modeled method is expected to be more accurate even in terms of 2D BEV tracking error. No quantitative analysis, however, has been reported. In this study, we simulated co-planar arc deliveries using respiratory motion data acquired from 42 patients to quantitatively compare the accuracy between 2D BEV and 3D-modeled tracking in arc therapy and to determine whether 3D information is needed for motion tracking. We used our previously developed low kV dose adaptive MV-kV imaging and motion compensation framework as a representative of 3D-modeled methods. It optimizes the balance between additional kV imaging dose and 3D tracking accuracy and solves the MLC blockage issue. With simulated Gaussian marker detection errors (zero mean and 0.39 mm standard deviation) and ~155/310/460 ms tracking system latencies, the mean percentage of time that the target moved  >2 mm from the predicted 2D BEV position are 1.1%/4.0%/7.8% and 1.3%/5.8%/11.6% for the 3D-modeled and 2D-only tracking, respectively. The corresponding average BEV RMS errors are 0.67/0.90/1.13 mm and 0.79/1.10/1.37 mm. Compared to the 2D method, the 3D method reduced the average RMS unresolved motion along the beam direction from ~3 mm to ~1 mm, resulting in on average only  <1% dosimetric advantage in the depth direction. Only for a small fraction of the patients, when tracking latency is long, the 3D-modeled method showed significant improvement of BEV tracking accuracy, indicating potential dosimetric advantage. However, if the tracking latency is short (~150 ms or less), those improvements are limited. Therefore, 2D BEV tracking has sufficient targeting accuracy for most clinical cases. The 3D technique is, however, still important in solving the MLC blockage problem during 2D BEV tracking.

Online updating of context-aware landmark detectors for prostate localization in daily treatment CT images

PURPOSE

In image guided radiation therapy, it is crucial to fast and accurately localize the prostate in the daily treatment images. To this end, the authors propose an online update scheme for landmark-guided prostate segmentation, which can fully exploit valuable patient-specific information contained in the previous treatment images and can achieve improved performance in landmark detection and prostate segmentation.

METHODS

To localize the prostate in the daily treatment images, the authors first automatically detect six anatomical landmarks on the prostate boundary by adopting a context-aware landmark detection method. Specifically, in this method, a two-layer regression forest is trained as a detector for each target landmark. Once all the newly detected landmarks from new treatment images are reviewed or adjusted (if necessary) by clinicians, they are further included into the training pool as new patient-specific information to update all the two-layer regression forests for the next treatment day. As more and more treatment images of the current patient are acquired, the two-layer regression forests can be continually updated by incorporating the patient-specific information into the training procedure. After all target landmarks are detected, a multiatlas random sample consensus (multiatlas RANSAC) method is used to segment the entire prostate by fusing multiple previously segmented prostates of the current patient after they are aligned to the current treatment image. Subsequently, the segmented prostate of the current treatment image is again reviewed (or even adjusted if needed) by clinicians before including it as a new shape example into the prostate shape dataset for helping localize the entire prostate in the next treatment image.

RESULTS

The experimental results on 330 images of 24 patients show the effectiveness of the authors' proposed online update scheme in improving the accuracies of both landmark detection and prostate segmentation. Besides, compared to the other state-of-the-art prostate segmentation methods, the authors' method achieves the best performance.

CONCLUSIONS

By appropriate use of valuable patient-specific information contained in the previous treatment images, the authors' proposed online update scheme can obtain satisfactory results for both landmark detection and prostate segmentation.

Incremental learning with selective memory (ILSM): towards fast prostate localization for image guided radiotherapy

Image-guided radiotherapy (IGRT) requires fast and accurate localization of the prostate in 3-D treatment-guided radiotherapy, which is challenging due to low tissue contrast and large anatomical variation across patients. On the other hand, the IGRT workflow involves collecting a series of computed tomography (CT) images from the same patient under treatment. These images contain valuable patient-specific information yet are often neglected by previous works. In this paper, we propose a novel learning framework, namely incremental learning with selective memory (ILSM), to effectively learn the patient-specific appearance characteristics from these patient-specific images. Specifically, starting with a population-based discriminative appearance model, ILSM aims to "personalize" the model to fit patient-specific appearance characteristics. The model is personalized with two steps: backward pruning that discards obsolete population-based knowledge and forward learning that incorporates patient-specific characteristics. By effectively combining the patient-specific characteristics with the general population statistics, the incrementally learned appearance model can localize the prostate of a specific patient much more accurately. This work has three contributions: 1) the proposed incremental learning framework can capture patient-specific characteristics more effectively, compared to traditional learning schemes, such as pure patient-specific learning, population-based learning, and mixture learning with patient-specific and population data; 2) this learning framework does not have any parametric model assumption, hence, allowing the adoption of any discriminative classifier; and 3) using ILSM, we can localize the prostate in treatment CTs accurately (DSC  ∼ 0.89 ) and fast (  ∼ 4 s), which satisfies the real-world clinical requirements of IGRT.

Real-time automatic fiducial marker tracking in low contrast cine-MV images

PURPOSE

To develop a real-time automatic method for tracking implanted radiographic markers in low-contrast cine-MV patient images used in image-guided radiation therapy (IGRT).

METHODS

Intrafraction motion tracking using radiotherapy beam-line MV images have gained some attention recently in IGRT because no additional imaging dose is introduced. However, MV images have much lower contrast than kV images, therefore a robust and automatic algorithm for marker detection in MV images is a prerequisite. Previous marker detection methods are all based on template matching or its derivatives. Template matching needs to match object shape that changes significantly for different implantation and projection angle. While these methods require a large number of templates to cover various situations, they are often forced to use a smaller number of templates to reduce the computation load because their methods all require exhaustive search in the region of interest. The authors solve this problem by synergetic use of modern but well-tested computer vision and artificial intelligence techniques; specifically the authors detect implanted markers utilizing discriminant analysis for initialization and use mean-shift feature space analysis for sequential tracking. This novel approach avoids exhaustive search by exploiting the temporal correlation between consecutive frames and makes it possible to perform more sophisticated detection at the beginning to improve the accuracy, followed by ultrafast sequential tracking after the initialization. The method was evaluated and validated using 1149 cine-MV images from two prostate IGRT patients and compared with manual marker detection results from six researchers. The average of the manual detection results is considered as the ground truth for comparisons.

RESULTS

The average root-mean-square errors of our real-time automatic tracking method from the ground truth are 1.9 and 2.1 pixels for the two patients (0.26 mm/pixel). The standard deviations of the results from the 6 researchers are 2.3 and 2.6 pixels. The proposed framework takes about 128 ms to detect four markers in the first MV images and about 23 ms to track these markers in each of the subsequent images.

CONCLUSIONS

The unified framework for tracking of multiple markers presented here can achieve marker detection accuracy similar to manual detection even in low-contrast cine-MV images. It can cope with shape deformations of fiducial markers at different gantry angles. The fast processing speed reduces the image processing portion of the system latency, therefore can improve the performance of real-time motion compensation.

Incremental learning with selective memory (ILSM): towards fast prostate localization for image guided radiotherapy

Image-guided radiotherapy (IGRT) requires fast and accurate localization of prostate in treatment CTs, which is challenging due to low tissue contrast and large anatomical variations across patients. On the other hand, in IGRT workflow, a series of CT images is acquired from the same patient under treatment, which contains valuable patient-specific information yet is often neglected by previous works. In this paper, we propose a novel learning framework, namely incremental learning with selective memory (ILSM), to effectively learn the patient-specific appearance characteristics from these patient-specific images. Specifically, starting with a population-based discriminative appearance model, ILSM aims to "personalize" the model to fit patient-specific appearance characteristics. Particularly, the model is personalized with two steps, backward pruning that discards obsolete population-based knowledge, and forward learning that incorporates patient-specific characteristics. By effectively combining the patient-specific characteristics with the general population statistics, the incrementally learned appearance model can localize the prostate of the specific patient much more accurately. Validated on a large dataset (349 CT scans), our method achieved high localization accuracy (DSC approximately 0.87) in 4 seconds.

Hybrid MV-kV 3D respiratory motion tracking during radiation therapy with low imaging dose

A novel real-time adaptive MV-kV imaging framework for image-guided radiation therapy is developed to reduce the thoracic and abdominal tumor targeting uncertainty caused by respiration-induced intrafraction motion with ultra-low patient imaging dose. In our method, continuous stereoscopic MV-kV imaging is used at the beginning of a radiation therapy delivery for several seconds to measure the implanted marker positions. After this stereoscopic imaging period, the kV imager is switched off except for the times when no fiducial marker is detected in the cine-MV images. The 3D time-varying marker positions are estimated by combining the MV 2D projection data and the motion correlations between directional components of marker motion established from the stereoscopic imaging period and updated afterwards; in particular, the most likely position is assumed to be the position on the projection line that has the shortest distance to the first principal component line segment constructed from previous trajectory points. An adaptive windowed auto-regressive prediction is utilized to predict the marker position a short time later (310 ms and 460 ms in this study) to allow for tracking system latency. To demonstrate the feasibility and evaluate the accuracy of the proposed method, computer simulations were performed for both arc and fixed-gantry deliveries using 66 h of retrospective tumor motion data from 42 patients treated for thoracic or abdominal cancers. The simulations reveal that using our hybrid approach, a smaller than 1.2 mm or 1.5 mm root-mean-square tracking error can be achieved at a system latency of 310 ms or 460 ms, respectively. Because the kV imaging is only used for a short period of time in our method, extra patient imaging dose can be reduced by an order of magnitude compared to continuous MV-kV imaging, while the clinical tumor targeting accuracy for thoracic or abdominal cancers is maintained. Furthermore, no additional hardware is required with the proposed method.

Incorporation of treatment plan spatial and temporal dose patterns into a prostate intrafractional motion management strategy

PURPOSE

Periodic MV∕KV radiographs taken during volumetric modulated arc therapy (VMAT) for hypofractionated treatment provide guidance in intrafractional motion management. The choice of imaging frequency and timing are key components in delivering the desired dose while reducing associated overhead such as imaging dose, preparation, and processing time. In this project the authors propose a paradigm with imaging timing and frequency based on the spatial and temporal dose patterns of the treatment plan.

METHODS

A number of control points are used in treatment planning to model VMAT delivery. For each control point, the sensitivity of individual target or organ-at-risk dose to motion can be calculated as the summation of dose degradations given the organ displacements along a number of possible motion directions. Instead of acquiring radiographs at uniform time intervals, MV∕KV image pairs are acquired indexed to motion sensitivity. Five prostate patients treated via hypofractionated VMAT are included in this study. Intrafractional prostate motion traces from the database of an electromagnetic tracking system are used to retrospectively simulate the VMAT delivery and motion management. During VMAT delivery simulation patient position is corrected based on the radiographic findings via couch movement if target deviation violates a patient-specific 3D threshold. The violation rate calculated as the percentage of traces failing the clinical dose objectives after motion correction is used to evaluate the efficacy of this approach.

RESULTS

Imaging indexed to a 10 s equitime interval and correcting patient position accordingly reduces the violation rate to 19.5% with intervention from 44.5% without intervention. Imaging indexed to the motion sensitivity further reduces the violation rate to 12.1% with the same number of images. To achieve the same 5% violation rate, the imaging incidence can be reduced by 40% by imaging indexed to motion sensitivity instead of time.

CONCLUSIONS

The simulation results suggest that image scheduling according to the characteristics of the treatment plan can improve the efficiency of intrafractional motion management. Using such a technique, the accuracy of delivered dose during image-guided hypofractionated VMAT treatment can be improved.

IL-1 generated subsequent to radiation-induced tissue injury contributes to the pathogenesis of radiodermatitis

Radiation injury in the skin causes radiodermatitis, a condition in which the skin becomes inflamed and the epidermis can break down. This condition causes significant morbidity and if severe it can be an independent factor that contributes to radiation mortality. Radiodermatitis is seen in some settings of radiotherapy for cancer and is also of concern as a complication post-radiation exposure from accidents or weapons, such as a "dirty bomb". The pathogenesis of this condition is incompletely understood. Here we have developed a murine model of radiodermatitis wherein the skin is selectively injured by irradiation with high-energy electrons. Using this model we showed that the interleukin-1 (IL-1) pathway plays a significant role in the development of radiodermatitis. Mice that lack either IL-1 or the IL-1 receptor developed less inflammation and less severe pathological changes in their skin, especially at later time-points. These findings suggest that IL-1 pathway may be a potential therapeutic target for reducing the severity of radiodermatitis.

Determination of action thresholds for electromagnetic tracking system-guided hypofractionated prostate radiotherapy using volumetric modulated arc therapy

PURPOSE

Hypofractionated prostate radiotherapy may benefit from both volumetric modulated are therapy (VMAT) due to shortened treatment time and intrafraction real-time monitoring provided by implanted radiofrequency(RF) transponders. The authors investigate dosimetrically driven action thresholds (whether treatment needs to be interrupted and patient repositioned) in VMAT treatment with electromagnetic (EM) tracking.

METHODS

VMAT plans for five patients are generated for prescription doses of 32.5 and 42.5 Gy in five fractions. Planning target volume (PTV) encloses the clinical target volume (CTV) with a 3 mm margin at the prostate-rectal interface and 5 mm elsewhere. The VMAT delivery is modeled using 180 equi-spaced static beams. Intrafraction prostate motion is simulated in the plan by displacing the beam isocenter at each beam assuming rigid organ motion according to a previously recorded trajectory of the transponder centroid. The cumulative dose delivered in each fraction is summed over all beams. Two sets of 57 prostate motion trajectories were randomly selected to form a learning and a testing dataset. Dosimetric end points including CTV D95%, rectum wall D1cc, bladder wall D1cc, and urethra Dmax, are analyzed against motion characteristics including the maximum amplitude of the anterior-posterior (AP), superior-inferior (SI), and left-right components. Action thresholds are triggered when intrafraction motion causes any violations of dose constraints to target and organs at risk (OAR), so that treatment is interrupted and patient is repositioned.

RESULTS

Intrafraction motion has a little effect on CTV D95%, indicating PTV margins are adequate. Tight posterior and inferior action thresholds around 1 mm need to be set in a patient specific manner to spare organs at risk, especially when the prescription dose is 42.5 Gy. Advantages of setting patient specific action thresholds are to reduce false positive alarms by 25% when prescription dose is low, and increase the sensitivity of detecting dose limits violations by 30% when prescription dose is high, compared to a generic 2 mm action box. The sensitivity and specificity calculated from the testing dataset are consistent to the learning set, which indicates that the patient specific approach is reliable and reproducible within the scope of the prostate database.

CONCLUSIONS

This work introduces a formalism for ensuring a VMAT delivery meets the most clinically important dose requirements by using patient specific and dosimetric-driven action thresholds to hold the beam and reposition the patient when necessary. Such methods can provide improved sensitivity and specificity compared to conventional methods, which assume directionally symmetric action thresholds.