Evaluation of a template-based algorithm for markerless lung tumour localization on single- and dual-energy kilovoltage images

Deep learning-based target decomposition for markerless lung tumor tracking in radiotherapy

BACKGROUND

In radiotherapy, real-time tumor tracking can verify tumor position during beam delivery, guide the radiation beam to target the tumor, and reduce the chance of a geometric miss. Markerless kV x-ray image-based tumor tracking is challenging due to the low tumor visibility caused by tumor-obscuring structures. Developing a new method to enhance tumor visibility for real-time tumor tracking is essential.

PURPOSE

To introduce a novel method for markerless kV image-based tracking of lung tumors via deep learning-based target decomposition.

METHODS

We utilized a conditional Generative Adversarial Network (cGAN), known as Pix2Pix, to build a patient-specific model and generate the synthetic decomposed target image (sDTI) to enhance tumor visibility on the real-time kV projection images acquired by the onboard kV imager equipped on modern linear accelerators. We used 4DCT simulation images to generate the digitally reconstructed radiograph (DRR) and DTI image pairs for model training. We augmented the training dataset by randomly shifting the 4DCT in the superior-inferior, anterior-posterior, and left-right directions during the DRR and DTI generation process. We performed real-time 2D tumor tracking via template matching between the DTI generated from the CT simulation and the sDTI generated from the real-time kV projection images. We validated the proposed method using nine patients' datasets with implanted beacons near the tumor.

RESULTS

The sDTI can effectively improve the image contrast around the lung tumors on the kV projection images for the nine patients. With the beacon motion as ground truth, the tracking errors were on average 0.8 ± 0.7 mm in the superior-inferior (SI) direction and 0.9 ± 0.8 mm in the in-plane left-right (IPLR) direction. The percentage of successful tracking, defined as a tracking error less than 2 mm in the SI direction, is 92.2% on the 4312 tested images. The patient-specific model took approximately 12 h to train. During testing, it took approximately 35 ms to generate one sDTI, and 13 ms to perform the tumor tracking using template matching.

CONCLUSIONS

Our method offers the potential solution for nearly real-time markerless lung tumor tracking. It achieved a high level of accuracy and an impressive tracking rate. Further development of 3D lung tumor tracking is warranted.

A novel bone suppression algorithm in intensity‐based 2D/3D image registration for real‐time tumour motion monitoring: development and phantom‐based validation

Background

Real‐time tumor motion monitoring (TMM) is a crucial process for intra‐fractional respiration management in lung cancer radiotherapy. Since the tumor can be partly or fully located behind the ribs, the TMM is challenging.

Purpose

The aim of this work was to develop a bone suppression (BS) algorithm designed for real‐time 2D/3D marker‐less TMM to increase the visibility of the tumor when overlapping with bony structures and consequently to improve the accuracy of TMM.

Method

A BS method was implemented in the in‐house developed software for ultrafast intensity‐based 2D/3D tumor registration (Fast Image‐based Registration [FIRE]). The method operates on both, digitally reconstructed radiograph (DRR) and intra‐fractional X‐ray images. The bony structures are derived from computed tomography data by thresholding during ray‐casting, and the resulting bone DRR is subtracted from intra‐fractional X‐ray images to obtain a soft‐tissue‐only image for subsequent tumor registration. The accuracy of TMM utilizing BS was evaluated within a retrospective phantom study with nine different 3D‐printed tumor phantoms placed in the in‐house developed Advanced Radiation DOSimetry (ARDOS) breathing phantom. A 24 mm craniocaudal tumor motion, including rib eclipses, was simulated, and X‐ray images were acquired on the Elekta Versa HD Linac in the lateral and posterior–anterior directions. An error assessment for BS images was evaluated with respect to the ground truth tumor position.

Results

A total error (root mean square error) of 0.87 ± 0.23 mm and 1.03 ± 0.26 mm was found for posterior–anterior and lateral imaging; the mean time for BS was 8.03 ± 1.54 ms. Without utilizing BS, TMM failed in all X‐ray images since the registration algorithm focused on the rib position due to the predominant intensity of this tissue within DRR and X‐ray images.

Conclusion

The BS algorithm developed and implemented improved the accuracy, robustness, and stability of real‐time TMM in lung cancer in a phantom study, even in the case of rib interlude where normal tumor registration fails.

Comprehensive characterization of ExacTrac stereoscopic image guidance system using Monte Carlo and Spektr simulations

The purpose of this work is to develop accurate computational methods to comprehensively characterize and model the clinical ExacTrac imaging system, which is used as an image guidance system for stereotactic treatment applications. The Spektr toolkit was utilized to simulate the spectral and imaging characterization of the system. Since Spektr only simulates the primary beam (ignoring scatter), a full model of ExacTrac was also developed in Monte Carlo (MC) to characterize the imaging system. To ensure proper performance of both simulation models, Spektr and MC data were compared to the measured spectral and half value layers (HVLs) values. To validate the simulation results, x-ray spectra of the ExacTrac system were measured for various tube potentials using a CdTe spectrometer with multiple added narrow collimators. The raw spectra were calibrated using a ⁵⁷Co source and corrected for the escape peaks and detector efficiency. HVLs in mm of Al for various energies were measured using a calibrated RaySafe detector. Spektr and MC HVLs were calculated and compared to the measured values. The patient surface dose was calculated for different clinical imaging protocols from the measured air kerma and HVL values following the TG-61 methodology. The x-ray focal spot was measured by slanted edge technique using gafchromic films. ExacTrac imaging system beam profiles were simulated for various energies by MC simulation and the results were benchmarked by experimentally acquired beam profiles using gafchromic films. The effect of 6D IGRT treatment couch on beam hardening, dynamic range of the flat panel detector and scatter effect were determined using both Spektr simulation and experimental measurements. The measured and simulated spectra (of both MC and Spektr) for various kVps were compared and agreed within acceptable error. As another validation, the measured HVLs agreed with the Spektr and MC simulated HVLs on average within 1.0% for all kVps. The maximum and minimum patient surface doses were found to be 1.06 mGy for shoulder (high) and 0.051 mGy for cranial (low) imaging protocols, respectively. The MC simulated beam profiles were well matched with experimental results and replicated the penumbral slopes, the heel effect, and out-of-field regions. Dynamic range of detector (in terms of air kerma at detector surface) was found to be in the range of [6.1 × 10^-6, 5.3 × 10^-3] mGy. Accurate MC and Spektr models of the ExacTrac image guidance system were successfully developed and benchmarked via experimental validation. While patient surface dose for available imaging protocols were reported in this study, the established MC model may be used to obtain 3D imaging dose distribution for real patient geometries.

Adaptive noise reduction for dual-energy x-ray imaging based on spatial variations in beam attenuation

PURPOSE

The main goal of this work is to improve the previously proposed patient-specific pixel-based dual-energy (PP-DE) algorithm by developing an adaptive anti-correlated noise reduction (ACNR) method, resulting in reduced image noise.

METHODS

Theoretical models of contrast-to-noise (CNR) and signal-to-noise (SNR) ratio were developed as functions of weighting factors for DE bone ω _Bn or soft tissue ω _ST cancellation. These analytical expressions describe CNR and SNR properties of dual-energy (DE) images, obtained with both simple log subtraction (SLS) and ACNR algorithms, and allow for a direct comparison between experimental and theoretical results. The theoretical models demonstrate the importance of ACNR weighting factor (ω _A ) optimization leading to the maximization of the SNR of the final image. A step phantom was constructed, which consisted of overlapping slabs of solid water (0-30 cm) and bone-mimicking material (0-6 cm), resulting in a total of 7 × 7 regions. High-energy (HE) and low-energy (LE) images were acquired at 140 kVp and 60 kVp with a clinical ExacTrac imaging system. The CNR and SNR were obtained for the DE images as functions of ω _Bn,ST and noise reduction weighting factor ω _A for different combinations of thicknesses. Weighting factors for bone cancellation were optimized for each region of interest (ROI) by finding zeros of CNR function for DE images between soft tissue only and soft tissue plus bone regions (and vice versa for soft tissue cancellation). The weighting factor for the ACNR algorithm ω _A was then optimized by maximizing the SNR function for each ROI. HE and LE images for an anthropomorphic Rando phantom were obtained with the same acquisition parameters as for the step phantom. DE images for bone only and soft tissue only were obtained with three algorithms: SLS and PP-DE with conventional ACNR (uniform ω _A ), and PP-DE with adaptive ACNR (region-varying ω _A ). Weighting factor maps for PP-DE and adaptive ACNR methods were obtained for Rando phantom geometry (which was determined from its CT scans) by interpolation (or extrapolation) of weighting factors for the step phantom. CNR values were calculated for different regions.

RESULTS

The CNR and SNR characteristics as functions of material cancellation and noise reduction weighting factors were obtained from theoretical models and experimental data from the step phantom. This showed a good qualitative validation of the models. For the ANCR algorithm, both the theory and experiment demonstrated that the material cancellation weighting factors (ω _Bn,ST ) can be optimized independently of the noise cancellation weighting factors (ω _A ), which can be optimized by maximizing SNR. For each ROI (with different overlapping bone and soft tissue thicknesses) the weighting factors ω _Bn,ST were determined as well as corresponding optimal weighting factors ω _A for noise reduction. For the Rando phantom, CNR values for regions representing different anatomical structures (ribs, spine, and tumor) were evaluated. It was shown that the proposed adaptive ACNR further improves image quality, compared to the conventional ACNR algorithm. The improvement is maximized for regions with bones (ribs or spine), where the largest attenuation is observed.

CONCLUSION

The ACNR weighting factors are dependent on the material thicknesses due to varying beam attenuation leading to different levels of quantum noise. This was shown with the derived theoretical expressions of the CNR and SNR functions and was validated by experimental data. The adaptive ANCR DE algorithm was developed, which allows for an increase in image quality by spatially varying weighting factors for noise reduction. This algorithm complements the previously developed PP-DE algorithm to obtain better quality DE images.

Characterization of Markerless Tumor Tracking Using the On-Board Imager of a Commercial Linear Accelerator Equipped With Fast-kV Switching Dual-Energy Imaging

Purpose

To describe and characterize fast-kV switching, dual-energy (DE) imaging implemented within the on-board imager of a commercial linear accelerator for markerless tumor tracking (MTT).

Methods and Materials

Fast-kV switching, DE imaging provides for rapid switching between programmed tube voltages (ie, 60 and 120 kVp) from one image frame to the next. To characterize this system, the weighting factor used for logarithmic subtraction and signal difference-to-noise ratio were analyzed as a function of time and frame rate. MTT was evaluated using a thorax motion phantom and fast kV, DE imaging was compared versus single energy (SE) imaging over 360 degrees of rotation. A template-based matching algorithm was used to track target motion on both DE and SE sequences. Receiver operating characteristics were used to compare tracking results for both modalities.

Results

The weighting factor was inversely related to frame rate and stable over time. After applying the frame rate–dependent weighting factor, the signal difference-to-noise ratio was consistent across all frame rates considered for simulated tumors ranging from 5 to 25 mm in diameter. An analysis of receiver operating characteristics curves showed improved tracking with DE versus SE imaging. The area under the curve for the 10-mm target ranged from 0.821 to 0.858 for SE imaging versus 0.968 to 0.974 for DE imaging. Moreover, the residual tracking errors for the same target size ranged from 2.02 to 2.18 mm versus 0.79 to 1.07 mm for SE and DE imaging, respectively.

Conclusions

Fast-kV switching, DE imaging was implemented on the on-board imager of a commercial linear accelerator. DE imaging resulted in improved MTT accuracy over SE imaging. Such an approach may have application for MTT of patients with lung cancer receiving stereotactic body radiation therapy, particularly for small tumors where MTT with SE imaging may fail.

Scatter imaging during lung stereotactic body radiation therapy characterized with phantom studies

By collecting photons scattered out of the therapy beam, scatter imaging creates images of the treated volume. Two phantoms were used to assess the possible application of scatter imaging for markerless tracking of lung tumors during stereotactic body radiation therapy (SBRT) treatment. A scatter-imaging camera was assembled with a CsI flat-panel detector and a 5 mm diameter pinhole collimator. Scatter images were collected during the irradiation of phantoms with megavoltage photons. To assess scatter image quality, spherical phantom lung tumors of 2.1-2.8 cm diameters were placed inside a static, anthropomorphic phantom. To show the efficacy of the technique with a moving target (3 cm diameter), the position of a simulated tumor was tracked in scatter images during sinusoidal motion (15 mm amplitude, 0.25 Hz frequency) in a dynamic lung phantom in open-field, dynamic conformal arc (DCA), and volumetric modulated arc therapy (VMAT) deliveries. Anatomical features are identifiable on static phantom scatter images collected with 10 MU of delivered dose (2.1 cm diameter lung tumor contrast-to-noise ratio of 4.4). The contrast-to-noise ratio increases with tumor size and delivered dose. During dynamic motion, the position of the 3.0 cm diameter lung tumor was identified with a root-mean-square error of 0.8, 1.2, and 2.9 mm for open field (0.3 s frame integration), DCA (0.5 s), and VMAT (0.5 s), respectively. Based on phantom studies, scatter imaging is a potential technique for markerless lung tumor tracking during SBRT without additional imaging dose. Quality scatter images may be collected at low, clinically relevant doses (10 MU). Scatter images are capable of sub-millimeter tracking precision, but modulation decreases accuracy.

Adaptive weighted log subtraction based on neural networks for markerless tumor tracking using dual-energy fluoroscopy

PURPOSE

To present a novel method, based on convolutional neural networks (CNN), to automate weighted log subtraction (WLS) for dual-energy (DE) fluoroscopy to be used in conjunction with markerless tumor tracking (MTT).

METHODS

A CNN was developed to automate WLS (aWLS) of DE fluoroscopy to enhance soft tissue visibility. Briefly, this algorithm consists of two phases: training a CNN architecture to predict pixel-wise weighting factors followed by application of WLS subtraction to reduce anatomical noise. To train the CNN, a custom phantom was built consisting of aluminum (Al) and acrylic (PMMA) step wedges. Per-pixel ground truth (GT) weighting factors were calculated by minimizing the contrast of Al in the step wedge phantom to train the CNN. The pretrained model was then utilized to predict pixel-wise weighting factors for use in WLS. For comparison, the weighting factor was manually determined in each projection (mWLS). A thorax phantom with five simulated spherical targets (5-25 mm) embedded in a lung cavity, was utilized to assess aWLS performance. The phantom was imaged with fast-kV dual-energy (120 and 60 kVp) fluoroscopy using the on-board imager of a commercial linear accelerator. DE images were processed offline to produce soft tissue images using both WLS methods. MTT was compared using soft tissue images produced with both mWLS and aWLS techniques.

RESULTS

Qualitative evaluation demonstrated that both methods achieved soft tissue images with similar quality. The use of aWLS increased the number of tracked frames by 1-5% compared to mWLS, with the largest increase observed for the smallest simulated tumors. The tracking errors for both methods produced agreement to within 0.1 mm.

CONCLUSIONS

A novel method to perform automated WLS for DE fluoroscopy was developed. Having similar soft tissue quality as well as bone suppression capability as mWLS, this method allows for real-time processing of DE images for MTT.

Markerless tumor tracking using fast-kV switching dual-energy fluoroscopy on a benchtop system

PURPOSE

To evaluate markerless tumor tracking (MTT) using fast-kV switching dual-energy (DE) fluoroscopy on a bench top system.

METHODS

Fast-kV switching DE fluoroscopy was implemented on a bench top which includes a turntable stand, flat panel detector, and x-ray tube. The customized generator firmware enables consecutive x-ray pulses that alternate between programmed high and low energies (e.g., 60 and 120 kVp) with a maximum frame rate of 15 Hz. In-house software was implemented to perform weighted DE subtraction of consecutive images to create an image sequence that removes bone and enhances soft tissues. The weighting factor was optimized based on gantry angle. To characterize this system, a phantom was used that simulates the chest anatomy and tumor motion in the lung. Five clinically relevant tumor sizes (5-25 mm diameter) were considered. The targets were programmed to move in the inferior-superior direction of the phantom, perpendicular to the x-ray beam, using a cos⁴ waveform to mimic respiratory motion. Target inserts were then tracked with MTT software using a template matching method. The optimal computed tomography (CT) slice thickness for template generation was also evaluated. Tracking success rate and accuracy were calculated in regions of the phantom where the target overlapped ribs vs spine, to compare the performance of single energy (SE) and DE imaging methods.

RESULTS

For the 5 mm target, a CT slice thickness of 0.75 mm resulted in the lowest tracking error. For the larger targets (≥10 mm) a CT slice thickness ≤2 mm resulted in comparable tracking errors for SE and DE images. Overall DE imaging improved MTT accuracy, relative to SE imaging, for all tumor targets in a rotational acquisition. Compared to SE, DE imaging increased tracking success rate of small target inserts (5 and 10 mm). For fast motion tracking, success rates improved from 23% to 64% and 74% to 90% for 5 and 10 mm targets inserts overlapping ribs, respectively. For slow moving targets success rates improved from 19% to 59% and 59% to 91% in 5 and 10 mm targets overlapping the ribs, respectively. Similar results were observed when the targets overlapped the spine. For larger targets (≥15 mm) tracking success rates were comparable using SE and DE imaging.

CONCLUSION

This work presents the first results of MTT using fast-kV switching DE fluoroscopy. Using DE imaging has improved the tracking accuracy of MTT, especially for small targets. The results of this study will guide the future implementation of fast-kV switching DE imaging using the on-board imager of a linear accelerator.

Evaluation of Radiomics to Predict the Accuracy of Markerless Motion Tracking of Lung Tumors: A Preliminary Study

Template-based matching algorithms are currently being considered for markerless motion tracking of lung tumors. These algorithms use tumor templates derived from the planning CT scan, and track the motion of the tumor on single energy fluoroscopic images obtained at the time of treatment. In cases where bone may obstruct the view of the tumor, dual energy fluoroscopy may be used to enhance soft tissue contrast. The goal of this study is to predict which tumors will have a high degree of accuracy for markerless motion tracking based on radiomic features obtained from the planning CT scan, using peak-to-sidelobe ratio (PSR) as a surrogate of tracking accuracy. In this study, CT imaging data of 8 lung cancer patients were obtained and analyzed through the open source IBEX program to generate 2,287 radiomic features. Agglomerative hierarchical clustering was used to narrow down these features into 145 clusters comprised of the highest correlation to PSR. The features among the clusters with the least inter-correlation were then chosen to limit redundancy in the data. The results of this study demonstrated a number of radiomic features that are positively correlated to PSR. The features with the highest degree of correlation included complexity, orientation and range. This approach may be used to determine patients for whom markerless motion tracking would be beneficial.