Tumor Motion Research Articles

A patient-specific deep learning framework for 3D motion estimation and volumetric imaging during lung cancer radiotherapy

Objective. Respiration introduces a constant source of irregular motion that poses a significant challenge for the precise irradiation of thoracic and abdominal cancers. Current real-time motion management strategies require dedicated systems that are not available in most radiotherapy centers. We sought to develop a system that estimates and visualises the impact of respiratory motion in 3D given the 2D images acquired on a standard linear accelerator. Approach. In this paper we introduce Voxelmap, a patient-specific deep learning framework that achieves 3D motion estimation and volumetric imaging using the data and resources available in standard clinical settings. Here we perform a simulation study of this framework using imaging data from two lung cancer patients. Main results. Using 2D images as input and 3D–3D Elastix registrations as ground-truth, Voxelmap was able to continuously predict 3D tumor motion with mean errors of 0.1 ± 0.5, −0.6 ± 0.8, and 0.0 ± 0.2 mm along the left–right, superior–inferior, and anterior–posterior axes respectively. Voxelmap also predicted 3D thoracoabdominal motion with mean errors of −0.1 ± 0.3, −0.1 ± 0.6, and −0.2 ± 0.2 mm respectively. Moreover, volumetric imaging was achieved with mean average error 0.0003, root-mean-squared error 0.0007, structural similarity 1.0 and peak-signal-to-noise ratio 65.8. Significance. The results of this study demonstrate the possibility of achieving 3D motion estimation and volumetric imaging during lung cancer treatments on a standard linear accelerator.

The Radiotherapy: Basic Concepts And Recent Techniques Advances For Thoracic Malignancies.

Radiation therapy (RT) is a clinical modality treatment that uses ionizing radiation in the approach of malignant tumors, (and occasionally benign diseases). Since its inception, the goal of RT has been to cure cancer without excessive side effects. The most important factors affecting the results of RT are the tumor histology, its location and regional extent, the anatomic area of involvement and the geometric accuracy with which a calculated radiation dose is delivered. Radiotherapy is one of the basic treatment modalities in thoracic malignancies and is used across all histology types and stages. Technical developments of radiotherapy have further strengthened and redefined its place in the management of lung cancer. High precision intensity-modulated radiation therapy, volumetric modulated arc therapy, and stereotactic body radiation therapy (SBRT), coupled with management of tumor motion and on-board imaging, increased efficacy and markedly decreased treatment-related toxicity. With this brief review, the authors pretend to present basic concepts and recent techniques advances in the use of radiation therapy for thoracic malignancies.

Cine MRI-based analysis of intrafractional motion in radiation treatment planning of head and neck cancer patients

Purpose/Objective(s)To investigate intrafraction motion of (HN) target volumes and to determine patient-specific planning target volume (PTV) margins. Materials/MethodsMR-cine imaging was performed for radiation treatment planning in HN cancer patients treated with definitive EBRT (n = 62) or SBRT (n = 4) on a 1.5 T MRI between 2017–2019. Dynamic MRI scans (sagittal orientation, 2 × 82 × 7 mm3 resolution), ranging from 3–5 min and 900–1500 images, were acquired. The position of the maximum tumor displacement along each direction in the anterior/posterior (A/P) and superior/inferior (S/I) position was recorded and analyzed to determine average PTV margins. ResultsPrimary tumor sites (n = 66) were oropharynx (n = 39), larynx (n = 24) and hypopharynx (n = 3). PTV margins for A/P/S/I positions were 4.1/4.4/5.0/6.2 mm and 4.9/4.3/6.7/7.7 mm for oropharyngeal and laryngeal/hypopharyngeal cancers when accounting for all motion. V100 for PTV was calculated and compared to the original plans. The mean drop in PTV coverage was in most cases under 5%. For a subset of patients with 3 mm plans available, V100 for PTV had more substantial decreases in coverage averaging 8.2% - and 14.3% for oropharyngeal and laryngeal/hypopharynx plans, respectively. ConclusionThe use of MR-cine in treatment planning allows for quantification of tumor motion during swallow and resting periods and should be accounted for during treatment planning. With motion considered, the derived margins may exceed the commonly used 3–5 mm PTV margins. Quantification and analysis of tumor and patient-specific PTV margins is a step towards real-time MRI guidance adaptive radiotherapy.

Enhancing the target visibility with synthetic target specific digitally reconstructed radiograph for intrafraction motion monitoring: A proof-of-concept study.

Intrafraction motion monitoring in External Beam Radiation Therapy (EBRT) is usually accomplished by establishing a correlation between the tumor and the surrogates such as an external infrared reflector, implanted fiducial markers, or patient skin surface. These techniques either have unstable surrogate-tumor correlation or are invasive. Markerless real-time onboard imaging is a noninvasive alternative that directly images the target motion. However, the low target visibility due to overlapping tissues along the X-ray projection path makes tumor tracking challenging. To enhance the target visibility in projection images, a patient-specific model was trained to synthesize the Target Specific Digitally Reconstructed Radiograph (TS-DRR). Patient-specific models were built using a conditional Generative Adversarial Network (cGAN) to map the onboard projection images to TS-DRR. The standard Pix2Pix network was adopted as our cGAN model. We synthesized the TS-DRR based on the onboard projection images using phantom and patient studies for spine tumors and lung tumors. Using previously acquired CT images, we generated DRR and its corresponding TS-DRR to train the network. For data augmentation, random translations were applied to the CT volume when generating the training images. For the spine, separate models were trained for an anthropomorphic phantom and a patient treated with paraspinal stereotactic body radiation therapy (SBRT). For lung, separate models were trained for a phantom with a spherical tumor insert and a patient treated with free-breathing SBRT. The models were tested using Intrafraction Review Images (IMR) for the spine and CBCT projection images for the lung. The performance of the models was validated using phantom studies with known couch shifts for the spine and known tumor deformation for the lung. Both the patient and phantom studies showed that the proposed method can effectively enhance the target visibility of the projection images by mapping them into synthetic TS-DRR (sTS-DRR). For the spine phantom with known shifts of 1 mm, 2 mm, 3 mm, and 4 mm, the absolute mean errors for tumor tracking were 0.11 ± 0.05 mm in the x direction and 0.25 ± 0.08 mm in the y direction. For the lung phantom with known tumor motion of 1.8 mm, 5.8 mm, and 9 mm superiorly, the absolute mean errors for the registration between the sTS-DRR and ground truth are 0.1 ± 0.3 mm in both the x and y directions. Compared to the projection images, the sTS-DRR has increased the image correlation with the ground truth by around 83% and increased the structural similarity index measure with the ground truth by around 75% for the lung phantom. The sTS-DRR can greatly enhance the target visibility in the onboard projection images for both the spine and lung tumors. The proposed method could be used to improve the markerless tumor tracking accuracy for EBRT.

Comparison of 3D and 4D robustly optimized proton treatment plans for non-small cell lung cancer patients with tumour motion amplitudes larger than 5mm.

There is no consensus about an ideal robust optimization (RO) strategy for proton therapy of targets with large intrafractional motion. We investigated the plan robustness of 3D and different 4D RO strategies. For eight non-small cell lung cancer patients with clinical target volume (CTV) motion >5mm, different RO approaches were investigated: 3DRO considering the average CT (AvgCT) with a target density override, 4DRO considering three/all 4DCT phases, and 4DRO considering the AvgCT and three/all 4DCT phases. Robustness against setup/range errors, interplay effects based on breathing and machine log file data for deliveries with/without rescanning, and interfractional anatomical changes were analyzed for target coverage and OAR sparing. All nominal plans fulfilled the clinical requirements with individual CTV coverage differences <2pp; 4DRO without AvgCT generated the most conformal dose distributions. Robustness against setup/range errors was best for 4DRO with AvgCT (18% more passed error scenarios than 3DRO). Interplay effects caused fraction-wise median CTV coverage loss of 3pp and missed maximum dose constraints for heart and esophagus in 18% of scenarios. CTV coverage and OAR sparing fulfilled requirements in all cases when accumulating four interplay scenarios. Interfractional changes caused less target misses for RO with AvgCT compared to 4DRO without AvgCT (≤42%/33% vs. ≥56%/44% failed single/accumulated scenarios). All RO strategies provided acceptable plans with equally low robustness against interplay effects demanding other mitigation than rescanning to ensure fraction-wise target coverage. 4DRO considering three phases and the AvgCT provided best compromise on planning effort and robustness.

Association of Internal and External Motion Based on Cine MR Images Acquired During Real-Time Treatment on MRI–Guided Linear Accelerator for Patients With Lung Cancer

With the recent clinical implementation of magnetic resonance imaging (MRI)-guided linear accelerators, a large number of real-time planar MR images has been acquired during lung cancer treatment as a standard of care. In this study, associations among lung tumor, diaphragm, and external skin movement were studied based on MR cine imaging during the entire duration of each treatment fraction. This retrospective study used 181,798 planar MRI frames acquired over 55 treatment/imaging sessions of 13 patients with lung cancer treated on 2 MRI-guided linear accelerators. From each planar MR image frame, in-house software automatically extracted 9 features: the superior-interior (SI) and posterior-anterior (PA) positions of a lung tumor; the area of the lung (Lung_Area); the posterior (Dia_Post), dome/apex (Dia_Dome), and anterior (Dia_Ant) points of a diaphragmatic curve; the diaphragm curve point (Dia_Max); and the chest (Chest) and belly (Belly) skin points experienced the maximum range of motions. Correlation analyses were performed among the 9 features for every session. Lung tumor motion range and standard deviations were calculated based on positions obtained in cine images and compared with motion ranges obtained from 4-dimensional computed tomography images. In the study, 177,009 frames of images were successfully analyzed. For all patients, correlation coefficients were as follows: 0.91 ± 0.10 between any 2 features among Lung_Area, Dia_Post, Dia_Dome, and Dia_Max; 0.82 ± 0.21 between SI and any feature among Lung_Area, Dia_Post, Dia_Dome, and Dia_Max; 0.75 ± 0.24 between SI and Belly. Six of 13 patients were considered large amplitude motion (patients with lung tumor SI motion standard deviation >5 mm). Furthermore, 92,956 frames of images were analyzed for the 6 large-amplitude motion patients. For this set, correlation coefficients were 0.93 ± 0.07 between any 2 features among Lung_Area, Dia_Post, Dia_Dome, and Dia_Max; 0.94 ± 0.06 between SI and any feature among Lung_Area, Dia_Post, Dia_Dome, and Dia_Max; and 0.90 ± 0.09 between SI and Belly. Both belly and diaphragmatic motions as assessed by cine MRI are highly correlated with large amplitude lung tumor motion in the longitudinal axis.

PO-1339 Comparing intra- and inter-fraction motion of tumors with fiducial markers for lung SBRT

PO-1339 Comparing intra- and inter-fraction motion of tumors with fiducial markers for lung SBRT

PO-1907 comparison of dosimetric error due to the tumor motion between TomoTherapy and LINAC-based VMAT

PO-1907 comparison of dosimetric error due to the tumor motion between TomoTherapy and LINAC-based VMAT

PD-0250 Intrafraction tumor motion monitoring and dose reconstruction for liver spot scanning proton therapy

PD-0250 Intrafraction tumor motion monitoring and dose reconstruction for liver spot scanning proton therapy

PD-0742 Accuracy of four models and update strategies to estimate liver tumor motion from external motion

PD-0742 Accuracy of four models and update strategies to estimate liver tumor motion from external motion

Respiratory-Correlated 4-Dimensional Magnetic Resonance Fingerprinting for Liver Cancer Radiation Therapy Motion Management

The objective of this study was to develop a respiratory-correlated (RC) 4-dimensional (4D) imaging technique based on magnetic resonance fingerprinting (MRF) (RC-4DMRF) for liver tumor motion management in radiation therapy. Thirteen patients with liver cancer were prospectively enrolled in this study. k-space MRF signals of the liver were acquired during free-breathing using the fast acquisition with steady-state precession sequence on a 3T scanner. The signals were binned into 8 respiratory phases based on respiratory surrogates, and interphase displacement vector fields were estimated using a phase-specific low-rank optimization method. Hereafter, the tissue property maps, including T1 and T2 relaxation times, and proton density, were reconstructed using a pyramid motion-compensated method that alternatively optimized interphase displacement vector fields and subspace images. To evaluate the efficacy of RC-4DMRF, amplitude motion differences and Pearson correlation coefficients were determined to assess measurement agreement in tumor motion between RC-4DMRF and cine magnetic resonance imaging (MRI); mean absolute percentage errors of the RC-4DMRF-derived tissue maps were calculated to reveal tissue quantification accuracy using digital human phantom; and tumor-to-liver contrast-to-noise ratio of RC-4DMRF images was compared with that of planning CT and contrast-enhanced MRI (CE-MRI) images. A paired Student t test was used for statistical significance analysis with a P value threshold of .05. RC-4DMRF achieved excellent agreement in motion measurement with cine MRI, yielding the mean (± standard deviation) Pearson correlation coefficients of 0.95 ± 0.05 and 0.93 ± 0.09 and amplitude motion differences of 1.48 ± 1.06 mm and 0.81 ± 0.64 mm in the superior-inferior and anterior-posterior directions, respectively. Moreover, RC-4DMRF achieved high accuracy in tissue property quantification, with mean absolute percentage errors of 8.8%, 9.6%, and 5.0% for T1, T2, and proton density, respectively. Notably, the tumor contrast-to-noise ratio in RC-4DMRI-derived T1 maps (6.41 ± 3.37) was found to be the highest among all tissue property maps, approximately equal to that of CE-MRI (6.96 ± 1.01, P=.862), and substantially higher than that of planning CT (2.91 ± 1.97, P=.048). RC-4DMRF demonstrated high accuracy in respiratory motion measurement and tissue properties quantification, potentially facilitating tumor motion management in liver radiation therapy.

Evaluating motion of pancreatic tumors and anatomical surrogates using cine MRI in 0.35T MRgRT under free breathing conditions.

Treatment tolerability is a significant limitation to pancreatic cancer treatment with radiotherapy due to proximity to highly radiosensitive organs and respiratory motion necessitating expanded target margins. Further, pancreatic tumors are difficult to visualize on conventional radiotherapy systems. Surrogates are often used to locate the tumor but are often inconsistent and do not provide strong positional relations throughout the respiratory cycle. This work utilizes a retrospective dataset of 45 pancreatic cancer patients treated on an MR-Linac system with cine MRI acquired for real-time target tracking. We investigated intra-fraction motion of tumors and two abdominal surrogates, leading to prediction models between the tumor and surrogate. Patient specific motion evaluation and prediction models were generated from 225 cine MRI series acquired during treatment. Tumor contours were used to evaluate the pancreatic tumor motion. Linear regression and principal component analysis (PCA) based models were used to predict tumor position from the anterior-posterior (AP) motion of the abdominal surface, the superior-inferior (SI) motion of the diaphragm, or a combination. Models were evaluated using mean squared error (MSE) and mean absolute error (MAE). Contour analysis showed the average pancreatic tumor motion range was 7.4±2.7mm and 14.9±5.8mm in the AP and SI directions, respectively. The PCA model had MSE of 1.4mm2 and 0.6mm2 , for the SI and AP directions, respectively, with both surrogates as inputs for the models. When only the abdomen surrogate was used, MSE was 1.3mm2 and 0.4mm2 in the SI and AP directions, while it was 0.4mm2 and 1.3mm2 when only the diaphragm surrogate was used. We evaluated intra-fraction pancreatic tumor motion and demonstrated prediction models between the tumor and surrogate. The models calculated the pancreatic tumor position from diaphragm, abdominal, or both contours within standard pancreatic cancer target margin, and the process could be applied to other disease sites in the abdominothoracic cavity.

Endobronchially Implanted Real-Time Electromagnetic Transponder Beacon–Guided, Respiratory-Gated SABR for Moving Lung Tumors: A Prospective Phase 1/2 Cohort Study

Endobronchial electromagnetic transponder beacons (EMT) provide real-time, precise positional data of moving lung tumors. We report results of a phase 1/2, prospective, single-arm cohort study evaluating the treatment planning effects of EMT-guided SABR for moving lung tumors. Eligible patients were adults, Eastern Cooperative Oncology Group 0 to 2, with T1-T2N0 non-small cell lung cancer or pulmonary metastasis ≤4 cm with motion amplitude ≥5 mm. Three EMTs were endobronchially implanted using navigational bronchoscopy. Four-dimensional free-breathing computed tomography simulation scans were obtained, and end-exhalation phases were used to define the gating window internal target volume. A 3-mm expansion of gating window internal target volume defined the planning target volume (PTV). EMT-guided, respiratory-gated (RG) SABR was delivered (54 Gy/3 fractions or 48 Gy/4 fractions) using volumetric modulated arc therapy. For each RG-SABR plan, a 10-phase image-guided SABR plan was generated for dosimetric comparison. PTV/organ-at-risk (OAR) metrics were tabulated and analyzed using the Wilcoxon signed-rank pair test. Treatment outcomes were evaluated using RECIST (Response Evaluation Criteria in Solid Tumours; version 1.1). Of 41 patients screened, 17 were enrolled and 2 withdrew from the study. Median age was 73 years, with 7 women. Sixty percent had T1/T2 non-small cell lung cancer and 40% had M1 disease. Median tumor diameter was 1.9 cm with 73% of targets located peripherally. Mean respiratory tumor motion was 1.25 cm (range, 0.53-4.04 cm). Thirteen tumors were treated with EMT-guided SABR and 47% of patients received 48 Gy in 4 fractions while 53% received 54 Gy in 3 fractions. RG-SABR yielded an average PTV reduction of 46.9% (P < .005). Lung V5, V10, V20, and mean lung dose had mean relative reductions of 11.3%, 20.3%, 31.1%, and 20.3%, respectively (P < .005). Dose to OARs was significantly reduced (P < .05) except for spinal cord. At 6 months, mean radiographic tumor volume reduction was 53.5% (P < .005). EMT-guided RG-SABR significantly reduced PTVs of moving lung tumors compared with image-guided SABR. EMT-guided RG-SABR should be considered for tumors with large respiratory motion amplitudes or those located in close proximity to OARs.

Application of four-dimensional cone beam computed tomography in lung cancer radiotherapy

ObjectiveThis study explored the application of four-dimensional cone beam computed tomography (4D CBCT) in lung cancer patients, seeking to improve the accuracy of radiotherapy and to establish a uniform protocol for the application of 4D CBCT in radiotherapy for lung cancer.Methods4D CBCT was applied to evaluate tumor volume response (TVR), motion, and center coordinates during radiotherapy in 67 eligible individuals with lung cancer diagnoses. The differences between 4D CBCT and 3D CBCT in different registration methods were compared.ResultsTVR was observed during treatment in 41% of patients (28/67), with a mean volume reduction of 41.7% and a median time to TVR of 19 days. Tumor motion was obvious in 16 patients, with a mean value of 0.52 cm (0.22 to 1.34 cm), and in 3 of 6 tumors close to the diaphragm (0.28 to 0.66 cm). Gray value registration based on mean density projection could still achieve close results to the 4D gray value registration. However, when the registration was based on bone alone, partial off-targeting occurred in the treatment in 41.8% of cases. The off-target rate was 19.0% when the tumor motion was ≤ 0.5 cm and 52.2% when the motion was > 0.5 cm.ConclusionTumor volume and motion of intrapulmonary lesions in individuals diagnosed with lung cancer varied significantly in the third week of radiotherapy. 4D CBCT may be more advantageous for isolated lesions without reference to relative anatomical structures or those near the diaphragm. Grayscale registration based on mean density projection is feasible.

Realistic CT data augmentation for accurate deep-learning based segmentation of head and neck tumors in kV images acquired during radiation therapy.

Using radiation therapy (RT) to treat head and neck (H&N) cancers requires precise targeting of the tumor to avoid damaging the surrounding healthy organs. Immobilisation masks and planning target volume margins are used to attempt to mitigate patient motion during treatment, however patient motion can still occur. Patient motion during RT can lead to decreased treatment effectiveness and a higher chance of treatment related side effects. Tracking tumor motion would enable motion compensation during RT, leading to more accurate dose delivery. The purpose of this paper is to develop a method to detect and segment the tumor in kV images acquired during RT. Unlike previous tumor segmentation methods for kV images, in this paper a process for generating realistic and synthetic CT deformations was developed to augment the training data, and make the segmentation method robust to patient motion. Detecting the tumor in 2D kV images is a necessary step towards 3D tracking of the tumor position during treatment. In this paper a conditional generative adversarial network (cGAN) is presented that can detect and segment the gross tumor volume (GTV) in kV images acquired during H&N RT. Retrospective data from 15 H&N cancer patients obtained from the Cancer Imaging Archive were used to train and test patient-specific cGANs. The training data consisted of digitally reconstructed radiographs (DRRs) generated from each patient's planning CT and contoured GTV. Training data was augmented by using synthetically deformed CTs to generate additional DRRs (in total 39600 DRRs per patient or 25200 DRRs for nasopharyngeal patients) containing realistic patient motion. The method for deforming the CTs was a novel deformation method based on simulating head rotation and internal tumor motion. The testing dataset consisted of 1080 DRRs for each patient, obtained by deforming the planning CT and GTV at different magnitudes to the training data. The accuracy of the generated segmentations was evaluated by measuring the segmentation centroid error, Dice similarity coefficient (DSC) and mean surface distance (MSD). This paper evaluated the hypothesis that when patient motion occurs, using a cGAN to segment the GTV would create a more accurate segmentation than no-tracking segmentations from the original contoured GTV, the current standard-of-care. This hypothesis was tested using the 1-tailed Mann-Whitney U-test. The magnitude of our cGAN segmentation centroid error was (mean ± standard deviation) 1.1 ± 0.8 millimeters and the DSC and MSD values were 0.90 ± 0.03 and 1.6 ± 0.5 millimeters respectively. Our cGAN segmentation method reduced the segmentation centroid error (p<0.001), and MSD (p = 0.031) when compared to the no-tracking segmentation, but did not significantly increase the DSC (p = 0.294). The accuracy of our cGAN segmentation method demonstrates the feasibility of this method for H&N cancer patients during RT. Accurate tumor segmentation of H&N tumors would allow for intrafraction monitoring methods to compensate for tumor motion during treatment, ensuring more accurate dose delivery and enabling better H&N cancer patient outcomes. This article is protected by copyright. All rights reserved.

Accuracy of abdominal organ motion estimation in radiotherapy using the right hemidiaphragm top as a surrogate during prolonged breath-holds quantified with MRI.

Respiratory motion presents a challenge in radiotherapy of thoracic and upper abdominal tumors. Techniques to account for respiratory motion include tracking. Using magnetic resonance imaging (MRI) guided radiotherapy systems, tumors can be tracked continuously. Using conventional linear accelerators, tracking of lung tumors is possible by determining tumor motion on kilo voltage (kV) imaging. But tracking of abdominal tumors with kV imaging is hampered by limited contrast. Therefore, surrogates for the tumor are used. One of the possible surrogates is the diaphragm. However, there is no universal method for establishing the error when using a surrogate and there are particular challenges in establishing such errors during free breathing (FB). Prolonged breath-holding might address these challenges. The aim of this study was to quantify the error when using the right hemidiaphragm top (RHT) as surrogate for abdominal organ motion during prolonged breath-holds (PBH) for possible application in radiation treatments. Fifteen healthy volunteers were trained to perform PBHs in two subsequent magnetic resonance imaging (MRI) sessions (PBH-MRI1 and PBH-MRI2). From each MRI acquisition, we selected seven images (dynamics) to determine organ displacement during PBH by using deformable image registration (DIR). On the first dynamic, the RHT, right and left hemidiaphragm, liver, spleen and right and left kidney were segmented. We used the deformation vector fields (DVF), generated by DIR, to determine the displacement of each organ between two dynamics in inferior-superior (IS), anterior-posterior (AP), left-right (LR) direction and we calculated the 3D vector magnitude (|d|). The displacements of the RHT, both hemidiaphragms and the abdominal organs were compared using a linear fit to determine the correlation (R2 of the fit) and the displacement ratio (DR, slope of the fit) between displacements of the RHT and each organ. We quantified the median difference between the DRs of PBH-MRI1 and PBH-MRI2 for each organ. Additionally, we estimated organ displacement in the second PBH by applying the DR from the first PBH to the displacement of the RHT measured during the second PBH. We compared the estimated organ displacement to the measured organ displacement during the second PBH. The difference between the two values was defined as the estimation error of using the RHT as a surrogate and assuming a constant DR over MRI sessions. The linear relationships were confirmed by the high R2 values of the linear fit between the displacements of the RHT and the abdominal organs (R2 > 0.96) in the IS and AP direction and |d|, and high to moderate correlations in the LR direction (0.93 > R2 > 0.64). The median DR difference between PBH-MRI1 and PBH-MRI2 varied between 0.13 to 0.31 for all organs. The median estimation error of the RHT as a surrogate varied between 0.4 and 0.8 mm/min for all organs. The RHT could serve as an accurate surrogate for abdominal organ motion during radiation treatments, e.g. in tracking, provided the error of the RHT as motion surrogate is taken into account in the margins. The study was registered in the Netherlands Trial Register (NL7603). This article is protected by copyright. All rights reserved.

Real-time respiratory motion prediction using photonic reservoir computing

Respiration induced motion is a well-recognized challenge in many clinical practices including upper body imaging, lung tumor motion tracking and radiation therapy. In this work, we present a recurrent neural network algorithm that was implemented in a photonic delay-line reservoir computer (RC) for real-time respiratory motion prediction. The respiratory motion signals are quasi-periodic waveforms subject to a variety of non-linear distortions. In this work, we demonstrated for the first time that RC can be effective in predicting short to medium range of respiratory motions within practical timescales. A double-sliding window technology is explored to enable the real-time establishment of an individually trained model for each patient and the real-time processing of live-streamed respiratory motion data. A breathing dataset from a total of 76 patients with breathing speeds ranging from 3 to 20 breaths per minute (BPM) is studied. Motion prediction of look-ahead times of 66.6, 166.6, and 333 ms are investigated. With a 333 ms look-ahead time, the real-time RC model achieves an average normalized mean square error (NMSE) of 0.025, an average mean absolute error (MAE) of 0.34 mm, an average root mean square error (RMSE) of 0.45 mm, an average therapeutic beam efficiency (TBE) of 94.14% for an absolute error (AE) < 1 mm, and 99.89% for AE < 3 mm. This study demonstrates that real-time RC is an efficient computing framework for high precision respiratory motion prediction.

Reproducibility of surface-based deep inspiration breath-hold technique for lung stereotactic body radiotherapy on a closed-bore gantry linac.

Tumor motion and delivery efficiency are two main challenges of lung stereotactic body radiotherapy (SBRT). The present work implemented the deep inspiration breath hold technique (DIBH) with surface guided radiation therapy (SGRT) on closed-bore linacs and investigated the correlation between SGRT data and internal target position. Thirteen lung SBRT patients treated in DIBH using a closed-bore gantry linac and a ring-mounted SGRT system were retrospectively analysed. Visual coaching was used to achieve DIBH with a±1mm threshold window in the anterior-posterior direction. Three kV-CBCTs were added to the treatment workflow and examined offline to verify intra-fraction tumor position. Surface-based DIBH was analysed using SGRT treatment reports and an in-house python script. Data from 73 treatment sessions and 175kV-CBCTs were studied. Correlations between target and surface positions were studied with Linear Mixed Models. Median intra-fraction tumor motion was 0.8mm (range: 0.7-1.3 mm) in the anterior-posterior direction, 1.2mm (range: 1-1.7 mm) in the superior-inferior direction, and 1mm (range: 0.7-1.1 mm) in the left-right direction, with rotations of <1° (range: 0.6°-1.1°) degree in all three directions. Planned target volumes and healthy lung volumes receiving 12.5Gy and 13.5Gy were reduced on average by 67% and 54%, respectively. Lung SBRT in DIBH with the ring-mounted SGRT system proved reproducible. The surface monitoring provided by SGRT was found to be a reliable surrogate for internal target motion. Moreover, the implementation of DIBH technique helped reduce target volumes and lung doses.

Real-time liver tumor localization via combined surface imaging and a single x-ray projection

Objective. Real-time imaging, a building block of real-time adaptive radiotherapy, provides instantaneous knowledge of anatomical motion to drive delivery adaptation to improve patient safety and treatment efficacy. The temporal constraint of real-time imaging (<500 milliseconds) significantly limits the imaging signals that can be acquired, rendering volumetric imaging and 3D tumor localization extremely challenging. Real-time liver imaging is particularly difficult, compounded by the low soft tissue contrast within the liver. We proposed a deep learning (DL)-based framework (Surf-X-Bio), to track 3D liver tumor motion in real-time from combined optical surface image and a single on-board x-ray projection. Approach. Surf-X-Bio performs mesh-based deformable registration to track/localize liver tumors volumetrically via three steps. First, a DL model was built to estimate liver boundary motion from an optical surface image, using learnt motion correlations between the respiratory-induced external body surface and liver boundary. Second, the residual liver boundary motion estimation error was further corrected by a graph neural network-based DL model, using information extracted from a single x-ray projection. Finally, a biomechanical modeling-driven DL model was applied to solve the intra-liver motion for tumor localization, using the liver boundary motion derived via prior steps. Main results. Surf-X-Bio demonstrated higher accuracy and better robustness in tumor localization, as compared to surface-image-only and x-ray-only models. By Surf-X-Bio, the mean (±s.d.) 95-percentile Hausdorff distance of the liver boundary from the ‘ground-truth’ decreased from 9.8 (±4.5) (before motion estimation) to 2.4 (±1.6) mm. The mean (±s.d.) center-of-mass localization error of the liver tumors decreased from 8.3 (±4.8) to 1.9 (±1.6) mm. Significance. Surf-X-Bio can accurately track liver tumors from combined surface imaging and x-ray imaging. The fast computational speed (<250 milliseconds per inference) allows it to be applied clinically for real-time motion management and adaptive radiotherapy.

Evaluation of deep learning based implanted fiducial markers tracking in pancreatic cancer patients

Real-time target position verification during pancreas stereotactic body radiation therapy (SBRT) is important for the detection of unplanned tumour motions. Fast and accurate fiducial marker segmentation is a Requirement of real-time marker-based verification. Deep learning (DL) segmentation techniques are ideal because they don’t require additional learning imaging or prior marker information (e.g., shape, orientation). In this study, we evaluated three DL frameworks for marker tracking applied to pancreatic cancer patient data. The DL frameworks evaluated were (1) a convolutional neural network (CNN) classifier with sliding window, (2) a pretrained you-only-look-once (YOLO) version-4 architecture, and (3) a hybrid CNN-YOLO. Intrafraction kV images collected during pancreas SBRT treatments were used as training data (44 fractions, 2017 frames). All patients had 1-4 implanted fiducial markers. Each model was evaluated on unseen kV images (42 fractions, 2517 frames). The ground truth was calculated from manual segmentation and triangulation of markers in orthogonal paired kV/MV images. The sensitivity, specificity, and area under the precision-recall curve (AUC) were calculated. In addition, the mean-absolute-error (MAE), root-mean-square-error (RMSE) and standard-error-of-mean (SEM) were calculated for the centroid of the markers predicted by the models, relative to the ground truth. The sensitivity and specificity of the CNN model were 99.41% and 99.69%, respectively. The AUC was 0.9998. The average precision of the YOLO model for different values of recall was 96.49%. The MAE of the three models in the left-right, superior-inferior, and anterior-posterior directions were under 0.88 ± 0.11 mm, and the RMSE were under 1.09 ± 0.12 mm. The detection times per frame on a GPU were 48.3, 22.9, and 17.1 milliseconds for the CNN, YOLO, and CNN-YOLO, respectively. The results demonstrate submillimeter accuracy of marker position predicted by DL models compared to the ground truth. The marker detection time was fast enough to meet the requirements for real-time application.