Real-time Tumor Tracking Research Articles

Application of motion prediction based on a long short-term memory network for imaging dose reduction in real-time tumor-tracking radiation therapy

PurposeTo demonstrate the possibility of using a lower imaging rate while maintaining acceptable accuracy by applying motion prediction to minimize the imaging dose in real-time image-guided radiation therapy. MethodsTime-series of three-dimensional internal marker positions obtained from 98 patients in liver stereotactic body radiation therapy were used to train and test the long-short-term memory (LSTM) network. For real-time imaging, the root mean squared error (RMSE) of the prediction on three-dimensional marker position made by LSTM, the residual motion of the target under respiratory-gated irradiation, and irradiation efficiency were evaluated. In the evaluation of the residual motion, the system-specific latency was assumed to be 100 ms. ResultsExcept for outliers in the superior–inferior (SI) direction, the median/maximum values of the RMSE for imaging rates of 7.5, 5.0, and 2.5 frames per second (fps) were 0.8/1.3, 0.9/1.6, and 1.2/2.4 mm, respectively. The median/maximum residual motion in the SI direction at an imaging rate of 15.0 fps without prediction of the marker position, which is a typical clinical setting, was 2.3/3.6 mm. For rates of 7.5, 5.0, and 2.5 fps with prediction, the corresponding values were 2.0/2.6, 2.2/3.3, and 2.4/3.9 mm, respectively. There was no significant difference between the irradiation efficiency with and that without prediction of the marker position. The geometrical accuracy at lower frame rates with prediction applied was superior or comparable to that at 15 fps without prediction. In comparison with the current clinical setting for real-time image-guided radiation therapy, which uses an imaging rate of 15.0 fps without prediction, it may be possible to reduce the imaging dose by half or more. ConclusionsMotion prediction can effectively lower the frame rate and minimize the imaging dose in real-time image-guided radiation therapy.

Artificial intelligence-based motion tracking in cancer radiotherapy: A review.

Radiotherapy aims to deliver a prescribed dose to the tumor while sparing neighboring organs at risk (OARs). Increasingly complex treatment techniques such as volumetric modulated arc therapy (VMAT), stereotactic radiosurgery (SRS), stereotactic body radiotherapy (SBRT), and proton therapy have been developed to deliver doses more precisely to the target. While such technologies have improved dose delivery, the implementation of intra-fraction motion management to verify tumor position at the time of treatment has become increasingly relevant. Artificial intelligence (AI) has recently demonstrated great potential for real-time tracking of tumors during treatment. However, AI-based motion management faces several challenges, including bias in training data, poor transparency, difficult data collection, complex workflows and quality assurance, and limited sample sizes. This review presents the AI algorithms used for chest, abdomen, and pelvic tumor motion management/tracking for radiotherapy and provides a literature summary on the topic. We will also discuss the limitations of these AI-based studies and propose potential improvements.

Clinical application of real-time tumor-tracking for stereotactic volumetric modulated arc therapy for liver tumors

Real-time tumor-tracking volumetric modulated arc therapy (RT-VMAT) enabling beam-gating based on continuous X-ray tracking of the three-dimensional position of internal markers is relevant for moving tumors. Dose-volume characteristics and treatment time were evaluated in ten consecutive patients who underwent liver stereotactic body radiation therapy with RT-VMAT. Target dose conformity and sparing of the stomach and the intestine were improved comparing RT-VMAT with RT-3D conformal radiotherapy. The mean treatment time for each fraction was less than 10 min. RT-VMAT could be effective, especially for targets located adjacent to organs at risk.

1080: The first clinical implementation of real-time 6DoF tumour tracking for liver SABR in the LARK trial

1080: The first clinical implementation of real-time 6DoF tumour tracking for liver SABR in the LARK trial

1690: Comparing Transformer training strategies for real-time tumor tracking in MRI-guided radiotherapy

1690: Comparing Transformer training strategies for real-time tumor tracking in MRI-guided radiotherapy

Survival analysis after stereotactic ablative radiotherapy for early stage non-small cell lung cancer: a single-institution cohort study

BackgroundStereotactic ablative radiotherapy (SABR) is the standard treatment for medically inoperable early-stage non-small cell lung cancer (ES-NSCLC), but which patients benefit from stereotactic radiotherapy is unclear. The aim of this study was to analyze prognostic factors for early mortality.MethodsFrom August 2010 to 2022, 617 patients with medically inoperable, peripheral or central ES-NSCLC were treated with SABR at our institution. We retrospectively evaluated the data from 172 consecutive patients treated from 2018 to 2020 to analyze the prognostic factors associated with overall survival (OS). The biological effective dose was > 100 Gy10 in all patients, and 60 Gy was applied in 3–5 fractions for a gross tumor volume (GTV) + 3 mm margin when the tumor diameter was < 1 cm; 30–33 Gy was delivered in one fraction. Real-time tumor tracking or an internal target volume approach was applied in 96% and 4% of cases, respectively. In uni- and multivariate analysis, a Cox model was used for the following variables: ventilation parameter FEV1, histology, age, T stage, central vs. peripheral site, gender, pretreatment PET, biologically effective dose (BED), and age-adjusted Charlson comorbidity index (AACCI).ResultsThe median OS was 35.3 months. In univariate analysis, no correlation was found between OS and ventilation parameters, histology, PET, or centrality. Tumor diameter, biological effective dose, gender, and AACCI met the criteria for inclusion in the multivariate analysis. The multivariate model showed that males (HR 1.51, 95% CI 1.01–2.28; p = 0.05) and AACCI > 5 (HR 1.56, 95% CI 1.06–2.31; p = 0.026) were significant negative prognostic factors of OS. However, the analysis of OS showed that the significant effect of AACCI > 5 was achieved only after 3 years (3-year OS 37% vs. 56%, p = 0.021), whereas the OS in one year was similar (1-year OS 83% vs. 86%, p = 0.58).ConclusionSABR of ES-NSCLC with precise image guidance is feasible for all medically inoperable patients with reasonable performance status. Early deaths were rare in our real-life cohort, and OS is clearly higher than would have been expected after best supportive care.

RT-SRTS: Angle-agnostic real-time simultaneous 3D reconstruction and tumor segmentation from single X-ray projection

Radiotherapy is one of the primary treatment methods for tumors, but the organ movement caused by respiration limits its accuracy. Recently, 3D imaging from a single X-ray projection has received extensive attention as a promising approach to address this issue. However, current methods can only reconstruct 3D images without directly locating the tumor and are only validated for fixed-angle imaging, which fails to fully meet the requirements of motion control in radiotherapy. In this study, a novel imaging method RT-SRTS is proposed which integrates 3D imaging and tumor segmentation into one network based on multi-task learning (MTL) and achieves real-time simultaneous 3D reconstruction and tumor segmentation from a single X-ray projection at any angle. Furthermore, the attention enhanced calibrator (AEC) and uncertain-region elaboration (URE) modules have been proposed to aid feature extraction and improve segmentation accuracy. The proposed method was evaluated on fifteen patient cases and compared with three state-of-the-art methods. It not only delivers superior 3D reconstruction but also demonstrates commendable tumor segmentation results. Simultaneous reconstruction and segmentation can be completed in approximately 70 ms, significantly faster than the required time threshold for real-time tumor tracking. The efficacies of both AEC and URE have also been validated in ablation studies. The code of work is available at https://github.com/ZywooSimple/RT-SRTS.

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

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. To introduce a novel method for markerless kV image-based tracking of lung tumors via deep learning-based target decomposition. 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. 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±0.7mm in the superior-inferior (SI) direction and 0.9±0.8mm in the in-plane left-right (IPLR) direction. The percentage of successful tracking, defined as a tracking error less than 2mm in the SI direction, is 92.2% on the 4312 tested images. The patient-specific model took approximately 12h to train. During testing, it took approximately 35ms to generate one sDTI, and 13ms to perform the tumor tracking using template matching. 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.

Deep inspiration breath hold real-time tumor-tracking radiation therapy (DBRT) as a novel stereotactic body radiation therapy approach for lung tumors

Radiotherapy with deep inspiration breath hold (DIBH) reduces doses to the lungs and organs at risk. The stability of breath holding and reproducibility of tumor location are higher during expiration than during inspiration; therefore, we developed an irradiation method combining DIBH and real-time tumor-tracking radiotherapy (RTRT) (DBRT). Nine patients were enrolled in this study. Fiducial markers were placed near tumors using bronchoscopy. Treatment planning computed tomography (CT) was performed thrice during DIBH, assisted by spirometer-based device. Each CT scan was fused using fiducial markers. Gross tumor volume (GTV) was contoured for each dataset and summed to create GTVsum; adding a 5-mm margin around GTVsum generated the planning target volume. The prescribed dose was mainly 42 Gy in four fractions. The treatment plan was created using DIBH CT (DBRT-plan), with a similar treatment plan created for expiratory CT for cases for which DBRT could not be performed (conv-plan). Vx defined as the volume of the lung received x Gy, and the mean lung dose, V20, V10, and V5 were evaluated. DBRT was completed in all patients. Mean dose, V20, and V10 were significantly lower in the DBRT-plan than in the conv-plan (all p = 0.003). Mean rates of decrease for mean dose, V20, and V10 were 14.0%, 27.6%, and 19.1%, respectively. No significant difference was observed in V5. We developed DBRT, a stereotactic body radiation therapy performed with the DIBH technique; it combines a spirometer-based breath-hold support system with an RTRT system. All patients who underwent DBRT completed the procedure without any technical or mechanical complications. This is a promising methodology that may significantly reduce lung doses.

Radiological imaging protection: a study on imaging dose used while planning computed tomography for external radiotherapy in Japan.

Previous studies have primarily focused on quality of imaging in radiotherapy planning computed tomography (RTCT), with few investigations on imaging doses. To our knowledge, this is the first study aimed to investigate the imaging dose in RTCT to determine baseline data for establishing national diagnostic reference levels (DRLs) in Japanese institutions. A survey questionnaire was sent to domestic RT institutions between 10 October and 16 December 2021. The questionnaire items were volume computed tomography dose index (CTDIvol), dose-length product (DLP), and acquisition parameters, including use of auto exposure image control (AEC) or image-improving reconstruction option (IIRO) for brain stereotactic irradiation (brain STI), head and neck (HN) intensity-modulated radiotherapy (IMRT), lung stereotactic body radiotherapy (lung SBRT), breast-conserving radiotherapy (breast RT), and prostate IMRT protocols. Details on the use of motion-management techniques for lung SBRT were collected. Consequently, we collected 328 responses. The 75th percentiles of CTDIvol were 92, 33, 86, 23, and 32mGy and those of DLP were 2805, 1301, 2416, 930, and 1158mGy·cm for brain STI, HN IMRT, lung SBRT, breast RT, and prostate IMRT, respectively. CTDIvol and DLP values in institutions that used AEC or IIRO were lower than those without use for almost all sites. The 75th percentiles of DLP in each treatment technique for lung SBRT were 2541, 2034, 2336, and 2730mGy·cm for free breathing, breath holding, gating technique, and real-time tumor tracking technique, respectively. Our data will help in establishing DRLs for RTCT protocols, thus reducing imaging doses in Japan.

Preliminary study of cine-MRI compression in MR-guided radiotherapy.

Cine-magnetic resonance imaging (MRI) is currently used in real-time tumor tracking during magnetic resonance (MR)-guided radiotherapy. As a type of MRI specified for motion tracking, a few minutes' acquisition results in thousands of 2-dimensional (2D) images. For MR-guided radiotherapy consisting of multiple treatment fractions, the large number of cine-MRI images would be disproportionate to the tight clinical data storage available. To alleviate this issue, the feasibility of compression of cine-MRI via video encoders was investigated in this study. The cine-MRI images were first sorted into 3 sequences according to their plane orientations. Then, each sequence was reordered according to their acquisition times [time-based (TB)] or content similarities [similarity-based (SB)]. As a result, 3 sequences were obtained for 3 plan orientations. Next, the obtained sequences were processed by a video encoder and the corresponding 3 video files were achieved. We employed 3 popular video encoders: Motion JPEG (M-JPEG), Advanced Video Coding (AVC), and High Efficiency Video Coding (HEVC). The performances of the sequence reordering methods and video encoders were evaluated based on a total of 150 image sets. The mean correlation quantities for SB sequences were higher than those for TB sequences by 3% (sagittal), 2% (coronal), and 1% (transverse), respectively. The average compression ratio (CR) yielded by the SB sequences was higher than that achieved by the TB sequences. Comparing with M-JPEG, the CRs obtained by AVC and HEVC were increased by 58% and 62% (sagittal), 16% and 23% (coronal), and 48% and 56% (transverse), respectively. Among the 3 video encoders, the highest CRs and restoration accuracy were achieved by HEVC. HEVC with inter-frame coding is more effective in reducing the redundant information in consecutive images. It is feasible to implement the video encoder for high-performance cine-MRI compression.

Biomedical advances and future prospects of high-precision three-dimensional radiotherapy and four-dimensional radiotherapy.

Biomedical advances of external-beam radiotherapy (EBRT) with improvements in physical accuracy are reviewed. High-precision (±1 mm) three-dimensional radiotherapy (3DRT) can utilize respective therapeutic open doors in the tumor control probability curve and in the normal tissue complication probability curve instead of the one single therapeutic window in two-dimensional EBRT. High-precision 3DRT achieved higher tumor control and probable survival rates for patients with small peripheral lung and liver cancers. Four-dimensional radiotherapy (4DRT), which can reduce uncertainties in 3DRT due to organ motion by real-time (every 0.1-1 s) tumor-tracking and immediate (0.1-1 s) irradiation, have achieved reduced adverse effects for prostate and pancreatic tumors near the digestive tract and with similar or better tumor control. Particle beam therapy improved tumor control and probable survival for patients with large liver tumors. The clinical outcomes of locally advanced or multiple tumors located near serial-type organs can theoretically be improved further by integrating the 4DRT concept with particle beams.

Ultra-fast multi-parametric 4D-MRI image reconstruction for real-time applications using a downsampling-invariant deformable registration (D2R) model

Background and purposeMotion estimation from severely downsampled 4D-MRI is essential for real-time imaging and tumor tracking. This simulation study developed a novel deep learning model for simultaneous MR image reconstruction and motion estimation, named the Downsampling-Invariant Deformable Registration (D2R) model. Materials and methodsForty-three patients undergoing radiotherapy for liver tumors were recruited for model training and internal validation. Five prospective patients from another center were recruited for external validation. Patients received 4D-MRI scans and 3D MRI scans. The 4D-MRI was retrospectively down-sampled to simulate real-time acquisition. Motion estimation was performed using the proposed D2R model. The accuracy and robustness of the proposed D2R model and baseline methods, including Demons, Elastix, the parametric total variation (pTV) algorithm, and VoxelMorph, were compared. High-quality (HQ) 4D-MR images were also constructed using the D2R model for real-time imaging feasibility verification. The image quality and motion accuracy of the constructed HQ 4D-MRI were evaluated. ResultsThe D2R model showed significantly superior and robust registration performance than all the baseline methods at downsampling factors up to 500. HQ T1-weighted and T2-weighted 4D-MR images were also successfully constructed with significantly improved image quality, sub-voxel level motion error, and real-time efficiency. External validation demonstrated the robustness and generalizability of the technique. ConclusionIn this study, we developed a novel D2R model for deformation estimation of downsampled 4D-MR images. HQ 4D-MR images were successfully constructed using the D2R model. This model may expand the clinical implementation of 4D-MRI for real-time motion management during liver cancer treatment.

Deep-Inspiration Breath-Hold Stereotactic Body Radiation Therapy by Combining Spirometer-Guided Breath-Hold and a Real-Time Tumor Tracking System: A Novel Approach

There are several methods used against respiratory motion (RM). Expiratory breath-hold (BH) is considered more stable and reproducible than inspiratory BH; therefore, BH with spirometry is often used for expiration. The real-time tumor tracking radiotherapy (RTRT) system is a highly effective method for reducing the margin of RM. This system ambushes and irradiates tumors during the expiratory phase when tumors move slowly. Although these methods usually involve expiration, it is advantageous to expand the lungs with inspiration to reduce the risk of adverse events. Here, we developed a new approach of performing stereotactic body radiation therapy (SBRT) under deep-inspiration BH (DIBH) by combining these two methods. Lung tumors with respiratory motion ≥ 1 cm were included. Three or four fiducial markers were placed near the tumor via bronchoscopy. DIBH CT (CT-IN) was performed under the guidance of spirometer. The PTV was obtained by adding a 5-mm margin to the GTV delineated on CT-IN. The prescribed dose was 42 Gy in four fractions for the D95 of the PTV. An error of 2.0 mm around the planned position of the fiducial marker on CT-IN was permitted along each orthogonal axis as a gating box. In preparation for cases in which the reproducibility of DIBH is low and treatment cannot be performed, light expiration BH CT (CT-EX) was also performed, and a radiotherapy plan was prepared for the conventional RTRT system so that it could be switched at any time. Lung volumes and doses (mean dose, V20 Gy, V10 Gy, and V5 Gy) on CT-EX and CT-IN were compared. Five patients underwent SBRT with DIBH, and all completed the planned irradiation course. The median treatment time per fraction was 27.86 min (range, 25.5-40.6). Four tumors were located in the left lower lobe and one in the right lower lobe. The median volume of PTV was 12.4 (range, 5.2-26.2) mL. The lung volumes and doses on CT-EX and CT-IN are shown in the Table. The lung volume on CT-IN was 1.6 times larger than that on CT-EX. The PTV-to-lung ratio on CT-IN was significantly lower than that on CT-EX. V20 Gy and V10 Gy on CT-IN were significantly lower than those on CT-EX. SBRT with DIBH was achieved by combining the spirometer and RTRT system. This can help to eliminate concerns about reproducibility and high-speed tumor movement during inspiration, which are weaknesses of spirometer-guided breath-hold and the RTRT system, respectively, while ensuring the accuracy of the RTRT system. DIBH SBRT is a promising method that can reduce lung dose.

Dosimetric Analysis of Patients Treated with Stereotactic Body Radiation Therapy for Malignant Portal Vein Tumor Thrombosis Using MRI-Guided vs. CT-Guided Linear Accelerator

The use of MRI-guided linear accelerators (MRI-LINAC) for radiation therapy has allowed for increased treatment precision and real-time tumor tracking. We hypothesized that the use of an MRI-LINAC, recently acquired by our institution, would allow us to treat malignant portal vein tumor thromboses (PVTT) closer to the origin of the portal vein with similar doses to the liver and nearby GI organs and with similar treatment toxicities compared to patients treated using a CT-guided linear accelerator (CT-LINAC). We retrospectively assessed patients who received SBRT for treatment of PVTT within the UVA health system using the electronic medical record. We contoured the origin of the portal vein on the simulation scans (MRI or CT) for each patient. The shortest distance from the GTV to the portal vein origin and to organs at risk (OAR; duodenum, stomach or "other intestine") were measured. The PTV, volume of the liver receiving less than 18 Gy during treatment, and the Dmax to the portal vein origin and OAR were determined. The means for each variable were then compared between CT-LINAC and MRI-LINAC using independent samples t-tests. Treatment toxicities were graded using CTCAE version 5.0 guidelines. All plans met institutional dose constraints for SBRT. A total of 6 patients treated on CT-LINAC and 3 patients treated on MRI-LINAC were identified (mean dose CT 37.7 Gy [range 21-50] in 2-5 fractions; MRI 48.3 Gy [range 45-50] in 5 fractions). The GTV was significantly closer to the portal vein origin in the MRI-LINAC group (mean CT 92.75 mm ± 45.89; MRI 32.00 mm ± 29.46; p = 0.04). The GTV trended towards being closer to OAR in the MRI-LINAC group (mean CT 43.81 mm ± 37.09; MRI 4.36 mm ± 1.85; p = 0.059). The Dmax to both the portal vein origin (mean CT 0.58 Gy ± 0.33; MRI 22.49 Gy ±24.16; p = 0.024) and OAR (mean CT 13.38 Gy ± 5.98; MRI 36.23 Gy ± 5.38; p = <0.001) was significantly higher in the MRI-LINAC group. The PTV (mean CT 152.11 cc ± 169.94; MRI 122.47 cc ± 114.65; p = 0.398) or volume of liver receiving less than 18 Gy during treatment (mean CT 1287.87 cc ± 285.20; MRI 1743.88 cc ±961.94 = 3; p = 0.147) were not significantly different between the two groups. No patients in either the MRI-LINAC or CT-LINAC group reported acute toxicities greater than Grade 1. We found that patients with PVTT treated with MRI-LINAC SBRT at our institution were more likely to have central tumors near dose-limiting OAR. However, due to superior tumor delineation and dynamic tumor tracking, these patients were able to be treated without decreasing treatment dose or increasing acute GI toxicity compared to patients treated with CT-LINAC. A larger sample size and longer follow-up period is warranted to corroborate these initial findings.

Risk-Adapted SABR for Ultra-Central Lung Tumors Attaining High BED10 to Targets in a Steep Dose-Response Window of 72-105 Gy under Tumor Tracking: A Planning Study

Higher rate of durable local control in stereotactic ablative radiotherapy (SABR) for ultra-central lung tumors (UCLTs) strongly correlates to a BED10 ≥ 100 Gy to the tumors. However, higher rates of grade 5 toxicities are reported to be directly associated with higher dose to the organs at risk (OARs) abutting or in close distance to the UCLTs, including trachea, proximal bronchial tree (PBT) and esophagus. This study evaluated two risk-adapted SABR schemes in a steep dose-response window of BED10 at 72 -105 Gy with isotoxic optimization and dynamic tumor tracking to safely attain high BED10 to the tumors while keeping dose to the OARs within evidence-based tolerances. A total of 12-16 patients are included for planning under two risk-adapted-SABR schemes of 60 Gy in 8 (scheme A) and15 (scheme B) fractions to attain high BED10 of ≥100 Gy to the distal sections of the PTV and yet potent BED10 at 72-84 Gy to the proximal sections of the PTV. Such inhomogeneous dose plans use a 3-5 mm GTV to PTV margins under a scenario that patients have at least one fiducial marker placed in or near the GTV for real-time tumor tracking on a robotic SRS-SABR system. All plans used fixed cone collimators and a planning tool's dose calculation to reach highest BED10 dose coverage achievable to the target volumes while respecting the OAR dose tolerances, including V105 Gy, V100 Gy, V84 Gy and V72 Gy in BED10 for target volumes and the maximum EQD2 dose tolerances (α/β = 3 Gy) in D0.03cc for the trachea/PBT ≤80.5-82.5 Gy and esophagus ≤64.0-77.6 Gy in scheme-A; and trachea/PBT≤97.7 Gy and esophagus ≤64.3 Gy in scheme-B. Median and range of plan dosimetry metrics are compared between the two schemes. For scheme-A, mean BED10 to PTV are 118.0 Gy (median) in105.5-133.5 Gy (range), and mean BED10 dose to GTV are 132.2 (118.3-149.4 Gy). EQD2 dose to trachea are 80.5 Gy (68.0-81.7 Gy), PBT 80.5 Gy (49.1-82.2 Gy), esophagus 67.2 Gy (51.6-77.3 Gy). PTV coverage by BED10 of 72 Gy, 84 Gy, 100 Gy and 105 Gy are 98.9% (87.5-100%), 97.1% (81.8-99.8%), 91.9% (72.5-98.6%) and 85.3% (68.5-97.9%). For scheme-B, mean BED10 to PTV are 102.2 Gy (93.4-106.0 Gy), and mean BED10 dose to GTV are 115.4 (103.4-117.3 Gy). EQD2 dose to trachea are 73.3 Gy (67.5-96.3 Gy), PBT 90.9 Gy (74.4-97.2 Gy), esophagus 61.4 Gy (41.9-64.1 Gy). PTV coverage by BED10 of 72 Gy, 84 Gy, 100 Gy and 105 Gy are 100% (94.9-100%), 95.5% (83.9-99.0%), 54.8% (31.2-65.4%) and 42.8% (17.9-53.2%). Two risk-adapted SABR of 7.5 Gyx8 and 4 Gyx15 are implemented for treating ultra-central lung tumors with BED10 ≥100 Gy to the distal sections of PTV and yet attain potent BED10 ≥72-84 Gy to the proximal sections of PTV that abut or overlap with trachea, proximal bronchial tree or esophagus. Accurate dose calculation by a planning tool and real-time tumor tracking are essential for safe and accurate delivery of such high-risk SABR treatments.

Simultaneous object detection and segmentation for patient-specific markerless lung tumor tracking in simulated radiographs with deep learning.

Real-time tumor tracking is one motion management method to address motion-induced uncertainty. To date, fiducial markers are often required to reliably track lung tumors with X-ray imaging, which carries risks of complications and leads to prolonged treatment time. A markerless tracking approach is thus desirable. Deep learning-based approaches have shown promise for markerless tracking, but systematic evaluation and procedures to investigate applicability in individual cases are missing. Moreover, few efforts have been made to provide bounding box prediction and mask segmentation simultaneously, which could allow either rigid or deformable multi-leaf collimatortracking. The purpose of this study was to implement a deep learning-based markerless lung tumor tracking model exploiting patient-specific training which outputs both a bounding box and a mask segmentation simultaneously. We also aimed to compare the two kinds of predictions and to implement a specific procedure to understand the feasibility of markerless tracking on individualcases. We first trained a Retina U-Net baseline model on digitally reconstructed radiographs (DRRs) generated from a public dataset containing 875 CT scans and corresponding lung nodule annotations. Afterwards, we used an independent cohort of 97 lung patients to develop a patient-specific refinement procedure. In order to determine the optimal hyperparameters for automatic patient-specific training, we selected 13 patients for validation where the baseline model predicted a bounding box on planning CT (PCT)-DRR with intersection over union (IoU) with the ground-truth higher than 0.7. The final test set contained the remaining 84 patients with varying PCT-DRR IoU. For each testing patient, the baseline model was refined on the PCT-DRR to generate a patient-specific model, which was then tested on a separate 10-phase 4DCT-DRR to mimic the intrafraction motion during treatment. A template matching algorithm served as benchmark model. The testing results were evaluated by four metrics: the center of mass (COM) error and the Dice similarity coefficient (DSC) for segmentation masks, and the center of box (COB) error and the DSC for bounding box detections. Performance was compared to the benchmark model including statistical testing forsignificance. A PCT-DRR IoU value of 0.2 was shown to be the threshold dividing inconsistent (68%) and consistent (100%) success (defined as mean bounding box DSC > 0.6) of PS models on 4DCT-DRRs. Thirty-seven out of the eighty-four testing cases had a PCT-DRR IoU above 0.2. For these 37 cases, the mean COM error was 2.6 mm, the mean segmentation DSC was 0.78, the mean COB error was 2.7 mm, and the mean box DSC was 0.83. Including the validation cases, the model was applicable to 50 out of 97 patients when using the PCT-DRR IoU threshold of 0.2. The inference time per frame was 170 ms. The model outperformed the benchmark model on all metrics, and the comparison was significant (p < 0.001) over the 37 PCT-DRR IoU > 0.2 cases, but not over the undifferentiated 84 testingcases. The implemented patient-specific refinement approach based on a pre-trained baseline model was shown to be applicable to markerless tumor tracking in simulated radiographs for lungcases.

Population-based asymmetric margins for moving targets in real-time tumor tracking.

Both geometric and dosimetric components are commonly considered when determining the margin for planning target volume (PTV). As dose distribution is shaped by controlling beam aperture in peripheral dose prescription and dose-escalated simultaneously integrated boost techniques, adjusting the margin by incorporating the variable dosimetric component into the PTV margin is inappropriate; therefore, geometric components should be accurately estimated for margin calculations. We introduced an asymmetric margin-calculation theory using the guide to the expression of uncertainty in measurement (GUM) and intra-fractional motion. The margins in fiducial marker-based real-time tumor tracking (RTTT) for lung, liver, and pancreatic cancers were calculated and were then evaluated using Monte Carlo (MC) simulations. A total of 74705, 73235, and 164968 sets of intra- and inter-fractional positional data were analyzed for 48 lung, 48 liver, and 25 pancreatic cancer patients, respectively, in RTTT clinical trials. The 2.5th and 97.5th percentiles of the positional error were considered representative values of each fraction of the disease site. The population-based statistics of the probability distributions of these representative positional errors (PD-RPEs) were calculated in six directions. A margin covering 95% of the population was calculated using the proposed formula. The content rate in which the clinical target volume (CTV) was included in the PTV was calculated through MC simulations using the PD-RPEs. The margins required for RTTT were at most 6.2, 4.6, and 3.9mm for lung, liver, and pancreatic cancer, respectively. MC simulations revealed that the median content rates using the proposed margins satisfied 95% for lung and liver cancers and 93% for pancreatic cancer, closer to the expected rates than the margins according to van Herk's formula. Our proposed formula based on the GUM and motion probability distributions (MPD) accurately calculated the practical margin size for fiducial marker-based RTTT. This was verified through MC simulations.

Deep learning-based markerless lung tumor tracking in stereotactic radiotherapy using Siamese networks.

Radiotherapy (RT) is involved in about 50% of all cancer patients, making it a very important treatment modality. The most common type of RT is external beam RT, which consists of delivering the radiation to the tumor from outside the body. One novel treatment delivery method is volumetric modulated arc therapy (VMAT), where the gantry continuously rotates around the patient during the radiation delivery. Accurate tumor position monitoring during stereotactic body radiotherapy (SBRT) for lung tumors can help to ensure that the tumor is only irradiated when it is inside the planning target volume. This can maximize tumor control and reduce uncertainty margins, lowering organ-at-risk dose. Conventional tracking methods are prone to errors, or have a low tracking rate, especially for small tumors that are in close vicinity to bonystructures. We investigated patient-specific deep Siamese networks for real-time tumor tracking, during VMAT. Due to lack of ground truth tumor locations in the kilovoltage (kV) images, each patient-specific model was trained on synthetic data (DRRs), generated from the 4D planning CT scans, and evaluated on clinical data (x-rays). Since there are no annotated datasets with kV images, we evaluated the model on a 3D printed anthropomorphic phantom but also on six patients by computing the correlation coefficient with the breathing-related vertical displacement of the surface-mounted marker (RPM). For each patient/phantom, we used 80% of DRRs for training and 20% forvalidation. The proposed Siamese model outperformed the conventional benchmark template matching-based method (RTR): (1) when evaluating both methods on the 3D phantom, the Siamese model obtained a 0.57-0.79-mm mean absolute distance to the ground truth tumor locations, compared to 1.04-1.56 mm obtained by RTR; (2) on patient data, the Siamese-determined longitudinal tumor position had a correlation coefficient of 0.71-0.98 with the RPM, compared to 0.07-0.85 for RTR; (3) the Siamese model had a 100% tracking rate, compared to 62%-82% forRTR. Based on these results, we argue that Siamese-based real-time 2D markerless tumor tracking during radiation delivery is possible. Further investigation and development of 3D tracking iswarranted.

A novel external/internal tumor tracking approach to compensate for respiratory motion baseline drifts

Objective. Real-time respiratory tumor tracking as implemented in a robotic treatment unit is based on continuous optical measurement of the position of external markers and a correlation model between them and internal target positions, which are established with X-ray imaging of the tumor, or fiducials placed in or around the tumor. Correlation models are created with fifteen simultaneously measured external/internal marker position pairs divided over the respiratory cycle. Every 45–150 s, the correlation model is updated by replacing the three first acquired data pairs with three new pairs. Tracking simulations for >120.000 computer-generated respiratory tracks demonstrated that this tracking approach resulted in relevant inaccuracies in internal target position predictions, especially in case of presence of respiratory motion baseline drifts. Approach. To better cope with drifts, we introduced a novel correlation model with an explicit time dependence, and we proposed to replace the currently applied linear-motion tracking (LMT) by mixed-model tracking (MMT). In MMT, the linear correlation model is extended with an explicit time dependence in case of a detected baseline drift. MMT prediction accuracies were then established for the same >120.000 computer-generated patients as used for LMT. Main results. For 150 s update intervals, MMT outperformed LMT in internal target position prediction accuracy for 93.7 ∣ 97.2% of patients with 0.25 ∣ 0.5 mm min−1 linear respiratory motion baseline drifts with similar numbers of X-ray images and similar treatment times. For the upper 25% of patients, mean 3D internal target position prediction errors reduced by 0.7 ∣ 1.8 mm, while near maximum reductions (upper 10% of patients) were 0.9 ∣ 2.0 mm. Significance. For equal numbers of acquired X-ray images, MMT greatly improved tracking accuracy compared to LMT, especially in the presence of baseline drifts. Even with almost 50% less acquired X-ray images, MMT still outperformed LMT in internal target position prediction accuracy.