Feasibility study of deep learning-based markerless real-time lung tumor tracking with orthogonal X-ray projection images.

PURPOSE: The feasibility of a deep learning-based markerless real-time tumor tracking (RTTT) method was retrospectively studied with orthogonal kV X-ray images and clinical tracking records acquired during lung cancer treatment. METHODS: Ten patients with lung cancer treated with marker-implanted RTTT were included. The prescription dose was 50 Gy in four fractions, using seven- to nine-port non-coplanar static beams. This corresponds to 14-18 X-ray tube angles for an orthogonal X-ray imaging system rotating with the gantry. All patients underwent 10 respiratory phases four-dimensional computed tomography. After a data augmentation approach, for each X-ray tube angle of a patient, 2250 digitally reconstructed radiograph (DRR) images with gross tumor volume (GTV) contour labeled were obtained. These images were adopted to train the patient and X-ray tube angle-specific GTV contour prediction model. During the testing, the model trained with DRR images predicted GTV contour on X-ray projection images acquired during treatment. The predicted three-dimensional (3D) positions of the GTV were calculated based on the centroids of the contours in the orthogonal images. The 3D positions of GTV determined by the marker-implanted RTTT during the treatment were considered as the ground truth. The 3D deviations between the prediction and the ground truth were calculated to evaluate the performance of the model. RESULTS: The median GTV volume and motion range were 7.42 (range, 1.18-25.74) cm3 and 22 (range, 11-28) mm, respectively. In total, 8993 3D position comparisons were included. The mean calculation time was 85 ms per image. The overall median value of the 3D deviation was 2.27 (interquartile range: 1.66-2.95) mm. The probability of the 3D deviation smaller than 5 mm was 93.6%. CONCLUSIONS: The evaluation results and calculation efficiency show the proposed deep learning-based markerless RTTT method may be feasible for patients with lung cancer.

Reducing variability among treatment machines using knowledge-based planning for head and neck, pancreatic, and rectal cancer.

PURPOSE: This study aimed to assess dosimetric indices of RapidPlan model-based plans for different energies (6, 8, 10, and 15 MV; 6- and 10-MV flattening filter-free), multileaf collimator (MLC) types (Millennium 120, High Definition 120, dual-layer MLC), and disease sites (head and neck, pancreatic, and rectal cancer) and compare these parameters with those of clinical plans. METHODS: RapidPlan models in the Eclipse version 15.6 were used with the data of 28, 42, and 20 patients with head and neck, pancreatic, and rectal cancer, respectively. RapidPlan models of head and neck, pancreatic, and rectal cancer were created for TrueBeam STx (High Definition 120) with 6 MV, TrueBeam STx with 10-MV flattening filter-free, and Clinac iX (Millennium 120) with 15 MV, respectively. The models were used to create volumetric-modulated arc therapy plans for a 10-patient test dataset using all energy and MLC types at all disease sites. The Holm test was used to compare multiple dosimetric indices in different treatment machines and energy types. RESULTS: The dosimetric indices for planning target volume and organs at risk in RapidPlan model-based plans were comparable to those in the clinical plan. Furthermore, no dose difference was observed among the RapidPlan models. The variability among RapidPlan models was consistent regardless of the treatment machines, MLC types, and energy. CONCLUSIONS: Dosimetric indices of RapidPlan model-based plans appear to be comparable to the ones based on clinical plans regardless of energies, MLC types, and disease sites. The results suggest that the RapidPlan model can generate treatment plans independent of the type of treatment machine.

Development of a physical geometric phantom for deformable image registration credentialing of radiotherapy centers for a clinical trial.

PURPOSE: This study aimed to develop a physical geometric phantom for the deformable image registration (DIR) credentialing of radiotherapy centers for a clinical trial and tested the feasibility of the proposed phantom at multiple domestic and international institutions. METHODS AND MATERIALS: The phantom reproduced tumor shrinkage, rectum shape change, and body shrinkage using several physical phantoms with custom inserts. We tested the feasibility of the proposed phantom using 5 DIR patterns at 17 domestic and 2 international institutions (21 datasets). Eight institutions used the MIM software (MIM Software Inc, Cleveland, OH); seven used Velocity (Varian Medical Systems, Palo Alto, CA), and six used RayStation (RaySearch Laboratories, Stockholm, Sweden). The DIR accuracy was evaluated using the Dice similarity coefficient (DSC) and Hausdorff distance (HD). RESULTS: The mean and one standard deviation (SD) values (range) of DSC were 0.909 ± 0.088 (0.434-0.984) and 0.909 ± 0.048 (0.726-0.972) for tumor and rectum proxies, respectively. The mean and one SD values (range) of the HD value were 5.02 ± 3.32 (1.53-20.35) and 5.79 ± 3.47 (1.22-21.48) (mm) for the tumor and rectum proxies, respectively. In three patterns evaluating the DIR accuracy within the entire phantom, 61.9% of the data had more than a DSC of 0.8 in both tumor and rectum proxies. In two patterns evaluating the DIR accuracy by focusing on tumor and rectum proxies, all data had more than a DSC of 0.8 in both tumor and rectum proxies. CONCLUSIONS: The wide range of DIR performance highlights the importance of optimizing the DIR process. Thus, the proposed method has considerable potential as an evaluation tool for DIR credentialing and quality assurance.

Validation of the clinical applicability of knowledge-based planning models in single-isocenter volumetric-modulated arc therapy for multiple brain metastases.

PURPOSE: To validate the clinical applicability of knowledge-based (KB) planning in single-isocenter volumetric-modulated arc therapy (VMAT) for multiple brain metastases using the k-fold cross-validation (CV) method. METHODS: This study comprised 60 consecutive patients with multiple brain metastases treated with single-isocenter VMAT (28 Gy in five fractions). The patients were divided randomly into five groups (Groups 1-5). The data of Groups 1-4 were used as the training and validation dataset and those of Group 5 were used as the testing dataset. Four KB models were created from three of the training and validation datasets and then applied to the remaining Groups as the fourfold CV phase. As the testing phase, the final KB model was applied to Group 5 and the dose distributions were calculated with a single optimization process. The dose-volume indices (DVIs), modified Ian Paddick Conformity Index (mIPCI), modulation complexity scores for VMAT plans (MCSv), and the total number of monitor units (MUs) of the final KB plan were compared to those of the clinical plan (CL) using a paired Wilcoxon signed-rank test. RESULTS: In the fourfold CV phase, no significant differences were observed in the DVIs among the four KB plans (KBPs). In the testing phase, the final KB plan was statistically equivalent to the CL, except for planning target volumes (PTVs) D2% and D50% . The differences between the CL and KBP in terms of the PTV D99.5% , normal brain, and Dmax to all organs at risk (OARs) were not significant. The KBP achieved a lower total number of MUs and higher MCSv than the CL with no significant difference. CONCLUSIONS: We demonstrated that a KB model in a single-isocenter VMAT for multiple brain metastases was equivalent in dose distribution, MCSv, and total number of MUs to a CL with a single optimization.

Variation in accumulated dose of volumetric-modulated arc therapy for pancreatic cancer due to different beam starting phases.

PURPOSE: To assess the effects of different beam starting phases on dosimetric variations in the clinical target volume (CTV) and organs at risk (OARs), and to identify the relationship between plan complexity and the dosimetric impact of interplay effects in volumetric-modulated arc therapy (VMAT) plans for pancreatic cancer. METHODS: Single and double full-arc VMAT plans were generated for 11 patients. A dose of 50.4 Gy in 28 fractions was prescribed to cover 50% of the planning target volume. Patient-specific Digital Imaging and Communications in Medicine-Radiation Therapy plan files were divided into 10 files based on the respiratory phases in four-dimensional computed tomography (4DCT) simulations. The phase-divided VMAT plans were calculated in consideration of the beam starting phase for each arc and were then combined in the mid-ventilation phase of 4DCT (4D plans). The dose-volumetric parameters were compared with the calculated dose distributions without consideration of the interplay effects (3D plans). Additionally, relationships among plan parameters such as modulation complexity scores, monitor units (MUs), and dose-volumetric parameters were evaluated. RESULTS: Dosimetric differences in the median values associated with different beam starting phases were within ± 1.0% and ± 0.2% for the CTV and ± 0.5% and ± 0.9% for the OARs during single and double full-arc VMAT, respectively. Significant differences caused by variations in the beam starting phases were observed only for the dose-volumetric parameters of the CTV during single full-arc VMAT (P < 0.05), associated with moderate or strong correlations between the MUs and the dosimetric differences between the 4D and 3D plans. CONCLUSIONS: The beam starting phase affected CTV dosimetric variations of single full-arc VMAT. The use of double full-arc VMAT mitigated this problem. However, variation in the dose delivered to OARs was not dependent on the beam starting phase, even for single full-arc VMAT.

Five-year outcomes following hypofractionated stereotactic radiotherapy delivered in five fractions for acoustic neuromas: the mean cochlear dose may impact hearing preservation.

BACKGROUND: The aim of this study was to assess the clinical outcomes of acoustic neuromas (ANs) treated with hypofractionated stereotactic radiotherapy (hypo-FSRT) prescribed at a uniform dose. METHODS: Forty-seven patients with a unilateral AN were treated consecutively with hypo-FSRT between February 2007 and March 2012. Nineteen patients maintained a serviceable hearing status at the beginning of hypo-FSRT. The prescribed dose was 25 Gy delivered in five fractions per week to the isocenter, and the planning target volume was covered by the 80% isodose line. RESULTS: The median follow-up and audiometric follow-up periods were 61 and 52 months, respectively. The estimated tumor control rate at 5 years was 90% (95% CI 76-96). The existence of the cystic component before hypo-FSRT had a significantly worse impact on tumor control (p = 0.02). The estimated hearing preservation rates at 1, 3 and 5 years were 68% (95% CI 42-84), 41% (95% CI 20-62) and 36% (95% CI 15-57), respectively. A borderline significant difference was identified in the mean biological effective dose with an α/ß value of 3 Gy (BED3) to the ipsilateral cochlea between the preserved hearing and hearing loss groups (19 Gy vs. 28 Gy) (p = 0.08). CONCLUSIONS: Hypo-FSRT delivered in five fractions for unilateral ANs may achieve excellent tumor control with no severe facial or trigeminal complications. The mean BED3 in the cochlea may impact the hearing preservation rate. Therefore, the cochlear dose should be as low as possible.

Quality assurance of geometric accuracy based on an electronic portal imaging device and log data analysis for Dynamic WaveArc irradiation.

The purpose of this study was to develop a simple verification method for the routine quality assurance (QA) of Dynamic WaveArc (DWA) irradiation using electronic portal imaging device (EPID) images and log data analysis. First, an automatic calibration method utilizing the outermost multileaf collimator (MLC) slits was developed to correct the misalignment between the center of the EPID and the beam axis. Moreover, to verify the detection accuracy of the MLC position according to the EPID images, various positions of the MLC with intentional errors in the range 0.1-1 mm were assessed. Second, to validate the geometric accuracy during DWA irradiation, tests were designed in consideration of three indices. Test 1 evaluated the accuracy of the MLC position. Test 2 assessed dose output consistency with variable dose rate (160-400 MU/min), gantry speed (2.2-6°/s), and ring speed (0.5-2.7°/s). Test 3 validated dose output consistency with variable values of the above parameters plus MLC speed (1.6-4.2 cm/s). All tests were delivered to the EPID and compared with those obtained using a stationary radiation beam with a 0° gantry angle. Irradiation log data were recorded simultaneously. The 0.1-mm intentional error on the MLC position could be detected by the EPID, which is smaller than the EPID pixel size. In Test 1, the MLC slit widths agreed within 0.20 mm of their exposed values. The averaged root-mean-square error (RMSE) of the dose outputs was less than 0.8% in Test 2 and Test 3. Using log data analysis in Test 3, the RMSE between the planned and recorded data was 0.1 mm, 0.12°, and 0.07° for the MLC position, gantry angle, and ring angle, respectively. The proposed method is useful for routine QA of the accuracy of DWA.

Long-term stability assessment of a 4D tumor tracking system integrated into a gimbaled linear accelerator.

We assessed long-term stability of tracking accuracy using the Vero4DRT system. This metric was observed between September 2012 and March 2015. A programmable respiratory motion phantom, designed to move phantoms synchronously with respiratory surrogates, was used. The infrared (IR) markers moved in the anterior-posterior (AP) direction as respiratory surrogates, while a cube phantom with a steel ball at the center, representing the tumor, and with radiopaque markers around it moved in the superior-inferior (SI) direction with one-dimensional (1D) sinusoidal patterns. A correlation model between the tumor and IR marker motion (4D model) was created from the training data obtained for 20 s just before beam delivery. The irradiation field was set to 3 × 3 cm2 and 300 monitor units (MUs) of desired MV X-ray beam were delivered. The gantry and ring angles were set to 0° and 45°, respectively. During beam delivery, the system recorded approximately 60 electronic portal imaging device (EPID) images. We analyzed: 1) the predictive accuracy of the 4D model (EP), defined as the difference between the detected and predicted target positions during 4D model creation, and 2) the tracking accuracy (ET), defined as the difference between the center of the steel ball and the MV X-ray field on the EPID image. The median values of mean plus two standard deviations (SDs) for EP were 0.06, 0.35, and 0.06 mm in the left-right (LR), SI, and AP directions, respectively. The mean values of maximum deviation for ET were 0.38, 0.49, and 0.53 mm and the coefficients of variance (CV) were 0.16, 0.10, and 0.05 in lateral, longitudinal, and 2D directions, respectively. Consequently, the IR Tracking accuracy was consistent over a period of two years. Our proposed method assessed the overall tracking accuracy readily using real-time EPID images, and proved to be a useful QA tool for dynamic tumor tracking with the Vero4DRT system.

Dosimetric evaluation of the Acuros XB algorithm for a 4 MV photon beam in head and neck intensity-modulated radiation therapy.

In this study, we assessed the differences in the dose distribution of a 4 MV photon beam among different calculation algorithms: the Acuros XB (AXB) algorithm, the analytic anisotropic algorithm (AAA), and the pencil beam convolution (PBC) algorithm (ver. 11.0.31), in phantoms and in clinical intensity-modulated radiation therapy (IMRT) plans. Homogeneous and heterogeneous, including middle-, low-, and high-density, phantoms were combined to assess the percentage depth dose and lateral dose profiles among AXB, AAA, and PBC. For the phantom containing the low-density area, AXB was in agreement with measurement within 0.5%, while the greatest differences between the AAA and PBC calculations and measurement were 2.7% and 3.6%, respectively. AXB showed agreement with measurement within 2.5% at the high-density area, while AAA and PBC overestimated the dose by more than 4.5% and 4.0%, respectively. Furthermore, 15 IMRT plans, calculated using AXB, for oropharyngeal, hypopharyngeal, and laryngeal carcinomas were analyzed. The dose prescription was 70 Gy to 50% of the planning target volume (PTV70). Subsequently, each plan was recalculated using AAA and PBC while maintaining the AXB-calculated monitor units, leaf motion, and beam arrangement. Additionally, nine hypopharyngeal and laryngeal cancer patients were analyzed in terms of PTV70 for cartilaginous structures (PTV(70_cartilage)). The doses covering 50% to PTV70 calculated by AAA and PBC were 2.1% ± 1.0% and 3.7% ± 0.8% significantly higher than those using AXB, respectively (p < 0.01). The increases in doses to PTV(70_cartilage) calculated by AAA and PBC relative to AXB were 3.9% and 5.3% on average, respectively, and were relatively greater than those in the entire PTV70. AXB was found to be in better agreement with measurement in phantoms in heterogeneous areas for the 4 MV photon beam. Considering AXB as the standard, AAA and PBC overestimated the IMRT dose for head and neck cancer. The dosimetric differences should not be ignored, particularly with cartilaginous structures in PTV.

Baseline correction of a correlation model for improving the prediction accuracy of infrared marker-based dynamic tumor tracking.

We previously found that the baseline drift of external and internal respiratory motion reduced the prediction accuracy of infrared (IR) marker-based dynamic tumor tracking irradiation (IR Tracking) using the Vero4DRT system. Here, we proposed a baseline correction method, applied immediately before beam delivery, to improve the prediction accuracy of IR Tracking. To perform IR Tracking, a four-dimensional (4D) model was constructed at the beginning of treatment to correlate the internal and external respiratory signals, and the model was expressed using a quadratic function involving the IR marker position (x) and its velocity (v), namely function F(x,v). First, the first 4D model, F1st(x,v), was adjusted by the baseline drift of IR markers (BDIR) along the x-axis, as function F'(x,v). Next, BDdetect, that defined as the difference between the target positions indicated by the implanted fiducial markers (Pdetect) and the predicted target positions with F'(x,v) (Ppredict) was determined using orthogonal kV X-ray images at the peaks of the Pdetect of the end-inhale and end-exhale phases for 10 s just before irradiation. F'(x,v) was corrected with BDdetect to compensate for the residual error. The final corrected 4D model was expressed as Fcor(x,v) = F1st{(x-BDIR),v}-BDdetect. We retrospectively applied this function to 53 paired log files of the 4D model for 12 lung cancer patients who underwent IR Tracking. The 95th percentile of the absolute differences between Pdetect and Ppredict (|Ep|) was compared between F1st(x,v) and Fcor(x,v). The median 95th percentile of |Ep| (units: mm) was 1.0, 1.7, and 3.5 for F1st(x,v), and 0.6, 1.1, and 2.1 for Fcor(x,v) in the left-right, anterior-posterior, and superior-inferior directions, respectively. Over all treatment sessions, the 95th percentile of |Ep| peaked at 3.2 mm using Fcor(x,v) compared with 8.4 mm using F1st(x,v). Our proposed method improved the prediction accuracy of IR Tracking by correcting the baseline drift immediately before irradiation.

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

BACKGROUND: 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. PURPOSE: 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. METHODS: A total of 74 705, 73 235, and 164 968 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. RESULTS: The margins required for RTTT were at most 6.2, 4.6, and 3.9 mm 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. CONCLUSIONS: 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.

Multi-institutional questionnaire-based survey on online adaptive radiotherapy performed using commercial systems in Japan in 2023.

In this study, we aimed to conduct a survey on the current clinical practice of, staffing for, commissioning of, and staff training for online adaptive radiotherapy (oART) in the institutions that installed commercial oART systems in Japan, and to share the information with institutions that will implement oART systems in future. A web-based questionnaire, containing 107 questions, was distributed to nine institutions in Japan. Data were collected from November to December 2023. Three institutions each with the MRIdian (ViewRay, Oakwood Village, OH, USA), Unity (Elekta AB, Stockholm, Sweden), and Ethos (Varian Medical Systems, Palo Alto, CA, USA) systems completed the questionnaire. One institution (MRIdian) had not performed oART by the response deadline. Each institution had installed only one oART system. Hypofractionation, and moderate hypofractionation or conventional fractionation were employed in the MRIdian/Unity and Ethos systems, respectively. The elapsed time for the oART process was faster with the Ethos than with the other systems. All institutions added additional staff for oART. Commissioning periods differed among the oART systems owing to provision of beam data from the vendors. Chambers used during commissioning measurements differed among the institutions. Institutional training was provided by all nine institutions. To the best of our knowledge, this was the first survey about oART performed using commercial systems in Japan. We believe that this study will provide useful information to institutions that installed, are installing, or are planning to install oART systems.

Effect of audio instruction on tracking errors using a four-dimensional image-guided radiotherapy system.

The Vero4DRT (MHI-TM2000) is capable of performing X-ray image-based tracking (X-ray Tracking) that directly tracks the target or fiducial markers under continuous kV X-ray imaging. Previously, we have shown that irregular respiratory patterns increased X-ray Tracking errors. Thus, we assumed that audio instruction, which generally improves the periodicity of respiration, should reduce tracking errors. The purpose of this study was to assess the effect of audio instruction on X-ray Tracking errors. Anterior-posterior abdominal skin-surface displacements obtained from ten lung cancer patients under free breathing and simple audio instruction were used as an alternative to tumor motion in the superior-inferior direction. First, a sequential predictive model based on the Levinson-Durbin algorithm was created to estimate the future three-dimensional (3D) target position under continuous kV X-ray imaging while moving a steel ball target of 9.5 mm in diameter. After creating the predictive model, the future 3D target position was sequentially calculated from the current and past 3D target positions based on the predictive model every 70 ms under continuous kV X-ray imaging. Simultaneously, the system controller of the Vero4DRT calculated the corresponding pan and tilt rotational angles of the gimbaled X-ray head, which then adjusted its orientation to the target. The calculated and current rotational angles of the gimbaled X-ray head were recorded every 5 ms. The target position measured by the laser displacement gauge was synchronously recorded every 10 msec. Total tracking system errors (ET) were compared between free breathing and audio instruction. Audio instruction significantly improved breathing regularity (p < 0.01). The mean ± standard deviation of the 95th percentile of ET (E95T ) was 1.7 ± 0.5 mm (range: 1.1-2.6mm) under free breathing (E95T,FB) and 1.9 ± 0.5 mm (range: 1.2-2.7 mm) under audio instruction (E95T,AI). E95T,AI was larger than E95T,FB for five patients; no significant difference was found between E95T,FB and E95T,AI (p = 0.21). Correlation analysis revealed that the rapid respiratory velocity significantly increased E95T. Although audio instruction improved breathing regularity, it also increased the respiratory velocity, which did not necessarily reduce tracking errors.

Clinical features and risk factors for interstitial lung disease spreading in low-dose irradiated areas after definitive radiotherapy with or without durvalumab consolidation therapy for patients with non-small cell lung cancer.

BACKGROUND: The current standard of care for patients with unresectable locally advanced non-small cell lung cancer (NSCLC) is chemoradiotherapy (CRT) combined with durvalumab consolidation therapy. However, radiotherapy (RT) always carries the risk of radiation pneumonitis (RP), which can preclude durvalumab continuation. In particular, the spread of interstitial lung disease (ILD) in low-dose areas or extending beyond the RT field often makes it difficult to determine the safety of continuation or rechallenging of durvalumab. Thus, we retrospectively analyzed ILD/RP after definitive RT with and without durvalumab, with assessment of radiologic features and dose distribution in RT. METHODS: We retrospectively evaluated the clinical records, CT imaging, and radiotherapy planning data of 74 patients with NSCLC who underwent definitive RT at our institution between July 2016 and July 2020. We assessed the risk factors for recurrence within one year and occurrence of ILD/RP. RESULTS: Kaplan-Meier method showed that ≥ 7 cycles of durvalumab significantly improved 1-year progression free survival (PFS) (p < 0.001). Nineteen patients (26%) were diagnosed with ≥ Grade 2 and 7 (9.5%) with ≥ Grade 3 ILD/RP after completing RT. There was no significant correlation between durvalumab administration and ≥ Grade 2 ILD/RP. Twelve patients (16%) developed ILD/RP that spread outside the high-dose (> 40 Gy) area, of whom 8 (67%) had ≥ Grade 2 and 3 (25%) had Grade 3 symptoms. In unadjusted and multivariate Cox proportional-hazards models adjusted for V20 (proportion of the lung volume receiving ≥ 20 Gy), high HbA1c level was significantly correlated with ILD/RP pattern spreading outside the high-dose area (hazard ratio, 1.842; 95% confidence interval, 1.35-2.51). CONCLUSIONS: Durvalumab improved 1-year PFS without increasing the risk of ILD/RP. Diabetic factors were associated with ILD/RP distribution pattern spreading in the lower dose area or outside RT fields, with a high rate of symptoms. Further study of the clinical background of patients including diabetes is needed to safely increase the number of durvalumab doses after CRT.

Positional accuracy of novel x-ray-image-based dynamic tumor-tracking irradiation using a gimbaled MV x-ray head of a Vero4DRT (MHI-TM2000).

PURPOSE: To verify the positional accuracy of a novel x-ray-image-based dynamic tumor-tracking (DTT) irradiation technique using the gimbaled MV x-ray head of a Vero4DRT (MHI-TM2000). METHODS: Verification of the x-ray-image-based DTT was performed using three components: a three-dimensional moving phantom with a steel ball target, a laser displacement gauge, and an orthogonal kV x-ray imaging subsystem with a gimbaled MV x-ray head and the system controller of the Vero4DRT. The moving phantom was driven based on seven periodic patterns [peak-to-peak amplitude (A): 20-40 mm, breathing period (T): 2-5 s] and 15 patients' aperiodic respiratory patterns (A: 6.5-22.9 mm, T: 1.9-5.8 s). The target position was detected in real time with the orthogonal kV x-ray imaging subsystem using the stereo vision technique. Subsequently, the Vero4DRT predicted the next position of the target, and then the gimbaled MV x-ray head tracked the corresponding orientation of the target. The displacements of the target were measured synchronously using the laser displacement gauge. The difference between the target positions predicted by the Vero4DRT and those measured by the laser displacement gauge was computed as the prediction error (E(P)), and the difference between the target positions tracked by the gimbaled MV x-ray head and predicted target positions was computed as the mechanical error (E(M)). Total tracking system error (E(T)) was defined as the difference between the tracked and measured target positions. RESULTS: The root mean squares (RMSs) of E(P), E(M), and E(T) were up to 0.8, 0.3, and 0.7 mm, respectively, for the periodic patterns. Regarding the aperiodic patterns, the median RMSs of E(P), E(M), and E(T) were 1.2 (range, 0.9-1.8) mm, 0.1 (range, 0.1-0.5) mm, and 1.2 (range, 0.9-1.8) mm, respectively. From the results of principal component analysis, tracking efficiency, defined as the ratio of twice the RMS of E(T) to A, was improved for patients with high respiratory function (R = 0.91; p < 0.01). CONCLUSIONS: The present study demonstrated that the Vero4DRT is capable of high-accuracy x-ray-image-based DTT. E(T) was caused primarily by E(P), and E(M) was negligible. Furthermore, principal component analysis showed that tracking efficiency could be improved with this system, especially for patients with high respiratory function.

Optimization of the x-ray monitoring angle for creating a correlation model between internal and external respiratory signals.

PURPOSE: To perform dynamic tumor tracking irradiation with the Vero4DRT (MHI-TM2000), a correlation model [four dimensional (4D) model] between the displacement of infrared markers on the abdominal wall and the three-dimensional position of a tumor indicated by a minimum of three implanted gold markers is required. However, the gold markers cannot be detected successfully on fluoroscopic images under the following situations: (1) overlapping of the gold markers; and (2) a low intensity ratio of the gold marker to its surroundings. In the present study, the authors proposed a method to readily determine the optimal x-ray monitoring angle for creating a 4D model utilizing computed tomography (CT) images. METHODS: The Vero4DRT mounting two orthogonal kV x-ray imaging subsystems can separately rotate the gantry along an O-shaped guide-lane and the O-ring along its vertical axis. The optimal x-ray monitoring angle was determined on CT images by minimizing the root-sum-square of water equivalent path lengths (WEPLs) on the orthogonal lines passing all of the gold markers while rotating the O-ring and the gantry. The x-ray monitoring angles at which the distances between the gold markers were within 5 mm at the isocenter level were excluded to prevent false detection of the gold markers in consideration of respiratory motions. First, the relationship between the WEPLs (unit: mm) and the intensity ratios of the gold markers was examined to assess the validity of our proposed method. Second, our proposed method was applied to the 4D-CT images at the end-expiration phase for 11 lung cancer patients who had four to five gold markers. To prove the necessity of the x-ray monitoring angle optimization, the intensity ratios of the least visible markers (minimum intensity ratios) that were estimated from the WEPLs were compared under the following conditions: the optimal x-ray monitoring angle and the angles used for setup verification. Additionally, the intra- and interfractional variations in the intensity ratio were examined from the optimal x-ray monitoring angle. RESULTS: A negative strong correlation was observed between the WEPL (x) and the intensity ratio (y) (y = 6.57 exp[-0.0125x] + 1, R = -0.88 [95% confidence interval: -0.85 to -0.90], p < 0.01). Our proposed method effectively avoided having the x-ray beam pass through high-density structures, although there were large interpatient variations in the optimal x-ray monitoring angle because of the geometric arrangement between the gold markers and the anatomical structures. The minimum intensity ratios that were estimated from the WEPLs at the optimal x-ray monitoring angle ranged from 1.43 to 2.48, which was an average of 1.27 times (range, 1.02-1.66) higher than the angles used for setup verification. The maximum intra- and interfractional decreases in the intensity ratio were 0.23 and 0.17, respectively. CONCLUSIONS: The authors demonstrated that the optimal x-ray monitoring angle for creating a 4D model can improve the visibility of gold markers.

Development of a deep learning-based patient-specific target contour prediction model for markerless tumor positioning.

PURPOSE: For pancreatic cancer patients, image guided radiation therapy and real-time tumor tracking (RTTT) techniques can deliver radiation to the target accurately. Currently, for the radiation therapy machine with kV X-ray imaging systems, internal markers must be implemented to facilitate tumor tracking. The purpose of this study was to develop a markerless deep learning-based pancreatic tumor positioning procedure for real-time tumor tracking with a kV X-ray imaging system. METHODS AND MATERIALS: Fourteen pancreatic cancer patients treated with intensity-modulated radiation therapy from six fixed gantry angles with a gimbal-head radiotherapy system were included in this study. For a gimbal-head radiotherapy system, the three-dimensional (3D) intrafraction target position can be determined using an orthogonal kV X-ray imaging system. All patients underwent four-dimensional computed tomography (4DCT) simulations for treatment planning, which were divided into 10 respiratory phases. After a patient's 4DCT was acquired, for each X-ray tube angle, 10 digitally reconstructed radiograph (DRR) images were obtained. Then, a data augmentation procedure was conducted. The data augmentation procedure first rotated the CT volume around the superior-inferior and anterior-posterior directions from -3° to 3° in 1.5° intervals. Then, the Super-SloMo model was adapted to interpolate 10 frames between respiratory phases. In total, the data augmentation procedure expanded the data scale 250-fold. In this study, for each patient, 12 datasets containing the DRR images from each specific X-ray tube angle based on the radiation therapy plan were obtained. The augmented dataset was randomly divided into training and testing datasets. The training dataset contained 2000 DRR images with clinical target volume (CTV) contours labeled for fine-tuning the pre-trained target contour prediction model. After the fine-tuning, the patient and X-ray tube angle-specific CTV contour prediction model was acquired. The testing dataset contained the remaining 500 images to evaluate the performance of the CTV contour prediction model. The dice similarity coefficient (DSC) between the area enclosed by the CTV contour and predicted contour was calculated to evaluate the model's contour prediction performance. The 3D position of the CTV was calculated based on the centroid of the contour in the orthogonal DRR images, and the 3D error of the prediction position was calculated to evaluate the CTV positioning performance. For each patient, the DSC results from 12 X-ray tube angles and 3D error from 6 gantry angles were calculated, representing the novelty of this study. RESULTS: The mean and standard deviation (SD) of all patients' DSCs were 0.98 and 0.015, respectively. The mean and SD of the 3D error were 0.29 mm and 0.14 mm, respectively. The global maximum 3D error was 1.66 mm, and the global minimum DSC was 0.81. The mean calculation time for CTV contour prediction was 55 ms per image. This fulfills the requirement of RTTT. CONCLUSIONS: Regarding the positioning accuracy and calculation efficiency, the presented procedure can provide a solution for markerless real-time tumor tracking for pancreatic cancer patients.

Preoperative chemoradiotherapy for locally advanced low rectal cancer using intensity-modulated radiotherapy to spare the intestines: a single-institutional pilot trial.

The irradiated volume of intestines is associated with gastrointestinal toxicity in preoperative chemoradiotherapy for rectal cancer. The current trial prospectively explored how much of the irradiated volume of intestines was reduced by intensity-modulated radiotherapy (IMRT) compared with 3-dimensional conformal radiotherapy (3DCRT) and whether IMRT might alleviate the acute gastrointestinal toxicity in this population. The treatment protocol encompassed preoperative chemoradiotherapy using IMRT plus surgery for patients with clinical T3-4, N0-2 low rectal cancer. IMRT delivered 45 Gy per 25 fractions for gross tumors, mesorectal and lateral lymph nodal regions, and tried to reduce the volume of intestines receiving 15 Gy (V15 Gy) < 120 cc and V45 Gy ≤ 0 cc, respectively, while keeping target coverage. S-1 and irinotecan were concurrently administered. Acute gastrointestinal toxicity, rates of clinical downstaging, sphincter preservation, local regional control (LRC) and overall survival (OS) were evaluated. Twelve enrolled patients completed the chemoradiotherapy protocol. The volumes of intestines receiving medium to high doses were reduced by the current IMRT protocol compared to 3DCRT; however, the predefined constraint of V15 Gy was met only in three patients. The rate of ≥ grade 2 gastrointestinal toxicity excluding anorectal symptoms was 17%. The rates of clinical downstaging, sphincter preservation, three-year LRC and OS were 75%, 92%, 92% and 92%, respectively. In conclusion, preoperative chemoradiotherapy using IMRT for this population might alleviate acute gastrointestinal toxicity, achieving high LRC and sphincter preservation; although further advancement is required to reduce the irradiated volume of intestines, especially those receiving low doses.

Multi-institutional phase II study on the safety and efficacy of dynamic tumor tracking-stereotactic body radiotherapy for lung tumors.

BACKGROUND AND PURPOSE: This study aimed to evaluate the safety and efficacy of dynamic tumor tracking-stereotactic body radiotherapy (DTT-SBRT) for lung tumors. MATERIALS AND METHODS: Patients with cStage I primary lung cancer or metastatic lung cancer with an expected range of respiratory motion of ≥10 mm were eligible for the study. The prescribed dose was 50 Gy in four fractions. A gimbal-mounted linac was used for DTT-SBRT delivery. The primary endpoint was local control at 2 years. RESULTS: Forty-eight patients from four institutions were enrolled in this study. Forty-two patients had primary non-small-cell lung cancer, and six had metastatic lung tumors. DTT-SBRT was delivered for 47 lesions in 47 patients with a median treatment time of 28 min per fraction. The median respiratory motion during the treatment was 13.7 mm (range: 4.5-28.1 mm). The motion-encompassing method was applied for the one remaining patient due to the poor correlation between the abdominal wall and tumor movement. The median follow-up period was 32.3 months, and the local control at 2 years was 95.2% (lower limit of the one-sided 85% confidence interval [CI]: 90.3%). The overall survival and progression-free survival at 2 years were 79.2% (95% CI: 64.7%-88.2%) and 75.0% (95% CI: 60.2%-85.0%), respectively. Grade 3 toxicity was observed in one patient (2.1%) with radiation pneumonitis. Grade 4 or 5 toxicity was not observed. CONCLUSION: DTT-SBRT achieved excellent local control with low incidences of severe toxicities in lung tumors with respiratory motion.

Development of AI-driven prediction models to realize real-time tumor tracking during radiotherapy.

BACKGROUND: In infrared reflective (IR) marker-based hybrid real-time tumor tracking (RTTT), the internal target position is predicted with the positions of IR markers attached on the patient's body surface using a prediction model. In this work, we developed two artificial intelligence (AI)-driven prediction models to improve RTTT radiotherapy, namely, a convolutional neural network (CNN) and an adaptive neuro-fuzzy inference system (ANFIS) model. The models aim to improve the accuracy in predicting three-dimensional tumor motion. METHODS: From patients whose respiration-induced motion of the tumor, indicated by the fiducial markers, exceeded 8 mm, 1079 logfiles of IR marker-based hybrid RTTT (IR Tracking) with the gimbal-head radiotherapy system were acquired and randomly divided into two datasets. All the included patients were breathing freely with more than four external IR markers. The historical dataset for the CNN model contained 1003 logfiles, while the remaining 76 logfiles complemented the evaluation dataset. The logfiles recorded the external IR marker positions at a frequency of 60 Hz and fiducial markers as surrogates for the detected target positions every 80-640 ms for 20-40 s. For each logfile in the evaluation dataset, the prediction models were trained based on the data in the first three quarters of the recording period. In the last quarter, the performance of the patient-specific prediction models was tested and evaluated. The overall performance of the AI-driven prediction models was ranked by the percentage of predicted target position within 2 mm of the detected target position. Moreover, the performance of the AI-driven models was compared to a regression prediction model currently implemented in gimbal-head radiotherapy systems. RESULTS: The percentage of the predicted target position within 2 mm of the detected target position was 95.1%, 92.6% and 85.6% for the CNN, ANFIS, and regression model, respectively. In the evaluation dataset, the CNN, ANFIS, and regression model performed best in 43, 28 and 5 logfiles, respectively. CONCLUSIONS: The proposed AI-driven prediction models outperformed the regression prediction model, and the overall performance of the CNN model was slightly better than that of the ANFIS model on the evaluation dataset.