3D Positions Of Markers Research Articles

Anatomical-Marker-Driven 3D Markerless Human Motion Capture.

Marker-based motion capture (mocap) is a conventional method used in biomechanics research to precisely analyze human movement. However, the time-consuming marker placement process and extensive post-processing limit its wider adoption. Therefore, markerless mocap systems that use deep learning to estimate 2D keypoint from images have emerged as a promising alternative, but annotation errors in training datasets used by deep learning models can affect estimation accuracy. To improve the precision of 2D keypoint annotation, we present a method that uses anatomical landmarks based on marker-based mocap. Specifically, we use multiple RGB cameras synchronized and calibrated with a marker-based mocap system to create a high-quality dataset (RRIS40) of images annotated with surface anatomical landmarks. A deep neural network is then trained to estimate these 2D anatomical landmarks and a ray-distance-based triangulation is used to calculate the 3D marker positions. We conducted extensive evaluations on our RRIS40 test set, which consists of 10 subjects performing various movements. Compared against a marker-based system, our method achieves a mean Euclidean error of 13.23 mm in 3D marker position, which is comparable to the precision of marker placement itself. By learning directly to predict anatomical keypoints from images, our method outperforms OpenCap's augmentation of 3D anatomical landmarks from triangulated wild keypoints. This highlights the potential of facilitating wider integration of markerless mocap into biomechanics research. The RRIS40 test set is made publicly available for research purposes at koonyook.github.io/rris40.

Functional popliteal angle tests improve identification of short hamstring muscle–tendon length in patients with a central neurological lesion

This study introduces a functional exercise protocol to improve the identification for short hamstring muscle–tendon length (HMTL), a common contributor to crouch gait in patients with central neurological lesions (CNL). The functional exercise protocol incorporates a knee extension movement with hip in a flexed position, while standing on one leg (functional popliteal angle test) and walking with large steps to the current standard protocol (walking at comfortable speed and as fast as possible). The main aim was to establish whether the new protocol allows better determination of maximum HMTLs and diagnostics of short HMTL in patients with a CNL. Lower limb 3D marker position data from 39 patient limbs and 10 healthy limbs performing the exercises were processed in OpenSim to extract HMTLs. The new protocol provoked significantly larger HMTLs compared to the current standard protocol. The total number of limbs classified as having too short HMTLs reduced from 16 to 4 out of a total of 30 limbs walking in crouch. The new protocol improves determination of maximum HMTL, thereby improving short HMTL diagnostics and identification of patients in need of lengthening treatment. Inter-individual variability observed among patients, indicating the need to include all exercises for comprehensive diagnosis.

Approaches for Hybrid Coregistration of Marker-Based and Markerless Coordinates Describing Complex Body/Object Interactions.

Full-body motion capture is essential for the study of body movement. Video-based, markerless, mocap systems are, in some cases, replacing marker-based systems, but hybrid systems are less explored. We develop methods for coregistration between 2D video and 3D marker positions when precise spatial relationships are not known a priori. We illustrate these methods on three-ball cascade juggling in which it was not possible to use marker-based tracking of the balls, and no tracking of the hands was possible due to occlusion. Using recorded video and motion capture, we aimed to transform 2D ball coordinates into 3D body space as well as recover details of hand motion. We proposed four linear coregistration methods that differ in how they optimize ball-motion constraints during hold and flight phases, using an initial estimate of hand position based on arm and wrist markers. We found that minimizing the error between ball and hand estimate was globally suboptimal, distorting ball flight trajectories. The best-performing method used gravitational constraints to transform vertical coordinates and ball-hold constraints to transform lateral coordinates. This method enabled an accurate description of ball flight as well as a reconstruction of wrist movements. We discuss these findings in the broader context of video/motion capture coregistration.

Prediction of the position of external markers using a recurrent neural network trained with unbiased online recurrent optimization for safe lung cancer radiotherapy

Background and Objective: During lung cancer radiotherapy, the position of infrared reflective objects on the chest can be recorded to estimate the tumor location. However, radiotherapy systems have a latency inherent to robot control limitations that impedes the radiation delivery precision. Prediction with online learning of recurrent neural networks (RNN) allows for adaptation to non-stationary respiratory signals, but classical methods such as real-time recurrent learning (RTRL) and truncated backpropagation through time are respectively slow and biased. This study investigates the capabilities of unbiased online recurrent optimization (UORO) to forecast respiratory motion and enhance safety in lung radiotherapy. Methods: We used nine observation records of the three-dimensional (3D) position of three external markers on the chest and abdomen of healthy individuals breathing during intervals from 73s to 222s. The sampling frequency was 10Hz, and the amplitudes of the recorded trajectories range from 6mm to 40mm in the superior-inferior direction. We forecast the 3D location of each marker simultaneously with a horizon value (the time interval in advance for which the prediction is made) between 0.1s and 2.0s, using an RNN trained with UORO. We compare its performance with an RNN trained with RTRL, least mean squares (LMS), and offline linear regression. We provide closed-form expressions for quantities involved in the gradient loss calculation in UORO, thereby making its implementation efficient. Training and cross-validation were performed during the first minute of each sequence. Results: On average over the horizon values considered and the nine sequences, UORO achieves the lowest root-mean-square (RMS) error and maximum error among the compared algorithms. These errors are respectively equal to 1.3mm and 8.8mm, and the prediction time per time step was lower than 2.8ms (Dell Intel core i9-9900K 3.60 GHz). Linear regression has the lowest RMS error for the horizon values 0.1s and 0.2s, followed by LMS for horizon values between 0.3s and 0.5s, and UORO for horizon values greater than 0.6s. Conclusions: UORO can accurately predict the 3D position of external markers for intermediate to high response times with an acceptable time performance. This will help limit unwanted damage to healthy tissues caused by radiotherapy.

Automated Intrafraction Motion Detection in 3D and Correction Using Monoscopic X-Ray Imaging Improves Accuracy and Efficiency of for High-Dose Prostate SBRT

Automated marker detection allows real time image guidance for treatment setup in 6D and intra-beam monitoring in 3D on a standard linear accelerator. This study retrospectively assessed the accuracy and efficiency of automated marker registration vs. manual match from a group of treated prostate SBRT patients.An automated marker detection scheme was developed by reconstructing a 3D marker model from orthogonal paired images for pre-treatment setup alignment using iterative closest point (ICP). The 3D marker model was back-projected to monoscopic x-ray images acquired during treatment. 3D marker motion was estimated from the consecutive (40o gantry angle apart) monoscopic x-ray images. Standard treatment was 2 arc Volumetric-modulated arc therapy (VMAT) plan prescribed to 8 Gy per fraction (x5). From 9 SBRT prostate patients, 135 paired images and 810 monoscopic x-ray images were analyzed using the automated scheme by comparing to the 3D marker model of planning image. A subgroup of the data set (5 patients, 12 fractions, 235 monoscopic images) were re-aligned to 3D marker position of setup images from the automated program to compare marker motion without realignment, which had minor residual setup error after manual match. Paired sample Wilcoxon ranked test was run to evaluate the improvement of using automated alignment, P-value < 0.05 was considered significant.Using automated marker match as reference, the residual setup error after manual match of the entire group (135 paired images) was (0.3 ± 0.6) mm, (0.1 ± 0.7) mm, (0.2 ± 0.8) mm in lateral (Lat), longitudinal (Long), vertical (Vert) directions, and (-2.6 ± 5.0)o, (-0.1 ± 2.8)o, (-1.3 ± 3.0)o in pitch, yaw, roll directions, respectively. The centroid of the marker position had an error of (0.2 ± 1.2) mm, (-0.6 ± 1.3) mm, (0.7 ± 1.4) mm in Lat, Long, Vert directions, respectively, during treatment (from 810 images). Treatment error at Lat, Long, Vert directions had stronger correlation with setup error in Lat & roll (-0.57 & -0.18), in Long & yaw (-0.49 & -0.25), and in Vert & roll (0.48 & 0.31), respectively. The magnitude of the improvement for individual fraction of the subgroup (235 images) using automated setup correction was 0.4 mm, 0.4 mm, 0.4 mm in Lat (P < 0.01), Long (P = 0.04), Vert (P = 0.39). The fraction of the treatment time that the marker centroid had a positional error of more than 2mm (a tolerance which triggered treatment interruption) was 9.5%, 18.5%, 12% in Lat, Long, Vert directions, respectively, using manual setup match, and 2.5%, 8.5%, 0.5% using automated setup match.This study demonstrated the ability of automated marker match to improve the accuracy and efficiency of prostate SBRT during treatment using monoscopic x-ray image. The improvement of treatment accuracy was 0.4 mm in each (X, Y, and Z) dimension. The improvement treatment efficiency was 50% reduction of treatment interruption which needs re-imaging. In addition, auto-registration saves manual markers alignment time.

Generation of realistic saddle trajectories from captured horseback motion

Hippotherapy, riding a horse in the context of rehabilitation, is a medical treatment that successfully has been employed in various fields, e.g. for improving locomotion performance of patients with movement disorders. Robotic systems enable the application of hippotherapy in clinical environments with additional benefits, like adjustable speed and high repeatability. Fundamental for a therapy outcome equivalent to classical hippotherapy is that the trajectory of the robotic system is as realistic as possible. This paper introduces a method for the synthesis of horseback motions using motion capture data of various horse gaits. Based on complete gait cycles, an analytical, time-continuous and periodic motion description is constructed. Measured 3D marker positions are reconstructed with a mean error not exceeding 8.6 mm. If the motion capture data of several gait cycles are considered, a more generalized trajectory is generated. An adjustable time dilatation parameter enables the adaptation of the generated motion according to the physical abilities of the patient or the capabilities of the robotic system. This method allows for motion synthesis with arbitrary time span and time resolution, generating realistic trajectories effectively applicable to robotic systems for riding simulation in general and robotic hippotherapy in particular.

Simultaneous identification of human body model parameters and gait trajectory from 3D motion capture data

The analysis of human movements rests on a realistic human body model. Deducing model parameters from anthropomorphic data is challenging since these are inherently imprecise. An approach to improve model accuracy is the parameter adaptation based on motion data. 3D motion capture data are already being used for generating the trajectories of a human body model, so combining motion tracking and parameter identification seems most natural. This paper introduces a holistic approach to simultaneously identify the geometric parameters of a kinematic human lower limb model and the parameters defining a (cyclic) gait trajectory, based on 3D marker positions. The result is a time-continuous description of a physiologically compatible lower extremity movement along with optimal model parameters so to best reproduce the captured motion. The method takes into account restrictions such as the range of motion of human body joints and is robust against missing data due to marker occlusions or failures of the measurement system. Considering multiple gait cycles of a movement trial, we derive the characteristic motion pattern (CMP) of a specific subject walking at a specific speed. Our method further allows for motion analysis and assessment, but also for motion synthesis with arbitrary time span and time resolution and can thus be used for simulations and trajectory planning of rehabilitation and movement assistance systems, such as exoskeletons.

A new deep learning-based method for the detection of gait events in children with gait disorders: Proof-of-concept and concurrent validity

The stance and swing phases of the gait cycle are defined by foot strike (FS) and foot off (FO). Accurate determination of these events is thus an essential component of 3D motion recordings processing. Several methods have been developed for the automatic detection of these events (based on the heuristics of 3D marker position, velocity and acceleration), however the results may be inaccurate due to the high variability that is intrinsic to pathological gait. For this reason, gait events are still commonly determined manually, which is a tedious process. Here we propose a new application (DeepEvent) of a long short term memory recurrent neural network for the automatic detection of gait events. The 3D position and velocity of the markers on the heel, toe and lateral malleolus were used by the network to determine FS and FO. The method was developed from 10526 FS and 9375 FO from 226 children. DeepEvent predicted FS within 5.5 ms and FO within 10.7 ms of the gold standard (automatic determination using force platform data) and was more accurate than common heuristic marker trajectory-based methods proposed in the literature and another deep learning method.A sensitivity analysis showed that DeepEvent mainly used the toe and heel markers (z-axis (longitudinal) position and velocity) at the beginning and end of gait cycle to predict FS, and the toe marker (x-axis (anterior/posterior) velocity and z-axis position and velocity) at around 60% of the gait cycle to predict FO.

A method to reconstruct intra-fractional liver motion in rotational radiotherapy using linear fiducial markers

Complex intra-fractional motion and deformation of the liver significantly impacts the accuracy of delivered dose in radiotherapy. It limits margin reduction, dose escalation and normal tissue sparing. A critical component of motion management is to accurately reconstruct tumor motion. In this study, we developed a six degrees of freedom projection marker matching method (6-DoF PM3) to reconstruct translational and rotational liver tumor motion in a rotational treatment delivery, such as volumetric modulated arc therapy (VMAT). Specifically, we modeled the use of two gold markers implanted in a linear form. The four endpoints of the two gold linear markers were used as tracking surrogates. During delivery, kV x-ray projection images were acquired. A method was developed to automatically identify the 2D marker-endpoints on the projection images. 3D marker positions were determined by solving an optimization problem with the objective function penalizing the distance from the reconstructed 3D position of each fiducial marker endpoint to the corresponding straight line defined by the kV x-ray projection of the endpoints. We performed a series of tests to evaluate different components of the method. For 2D marker endpoints identification, 99.9% of the marker endpoints were identified with an error (1 pixel) along both u and v directions. For 3D reconstruction of motion in simulation studies, error of rotational angle was ° without considering the 2D marker identification error. The rotational angle error was relatively sensitive to the accuracy of 2D marker identification. When the 2D error raised from 0.22 mm to 0.776 mm, the error of 3D rotational angle increased from 0.5° to 2.5°. In the experimental end-to-end tests, the mean root-mean square error of the 3D reconstructed marker positions was 0.75 mm and the mean error of rotational angle was within 1.7°. Our method can accurately determine intra-fractional liver tumor motion in rotational radiotherapy using kV projections of only two linear fiducial markers.

On optical data-guided optimal control simulations of human motion

This work addresses the synergistic fusion of optimal control simulations and marker-based optical measurements of human motion. The latter is a widespread capturing technology in biomechanics and movement science. In the context of optimal control simulations, the idea is to improve the computational performance by using a realistic initial guess and to increase the realism of the simulated motion through data-guiding. In the context of motion capturing, the idea is to use biomechanical simulations in order to maintain accurate capturings also with reduced measurement frequencies and points. This would greatly improve the usability of such systems in terms of setup time and wearing comfort. In this work, we investigate different methods for combining physical laws, 3D marker positions obtained from the optical system, and physiologically motivated objectives in an optimal control framework. Moreover, we explore the potential of obtaining reasonable results—in terms of motion trajectories and torques that are close to reference obtained from using all available information—with a reduced measurement frequency and a reduced number of markers. The tests are performed on a human steering and throwing motion, where a human arm was captured with seven retroreflective markers at $120\text{ Hz}$. Our results show, that a significant reduction of exploited measurements still provides feasible simulation results in our proposed method, given that the physiologically motivated objective reflects the actual movement. Furthermore, it turns out that neglecting markers close to the shoulder has less influence on the simulation results than neglecting markers close to the hand.

Directional scatter imaging for the stereoscopic tracking of fiducial markers in a single kV exposure.

To demonstrate, via Monte Carlo simulation, that an image obtained from the patient-generated scattered radiation forced to impinge on the detector from a known direction by means of parallel-focused grids, can be used to complement the information conveyed by the primary image, such that accurate stereoscopic three-dimensional localization of fiducial markers can be achieved in a single kV x-ray exposure. A voxelized Zubal phantom was used to model the process of fiducial marker localization. The markers were represented as made of gold and cylindrical in shape with dimensions of 5mm in length and 1mm in diameter. Three such markers were placed in the Zubal phantom at the prostate level. Two gantry-mounted image acquisition geometries were modeled: a single kV imaging system and a dual kV-MV imaging system. The detector was modeled as a 30cm×40cm Gd2 O2 S screen with a thickness of 0.2cm and a resolution of 768×1024 pixels. The PENELOPE Monte Carlo code was used to calculate the absorbed dose in this detector imparted by the transmitted primary and directional scatter radiation. A grayscale conversion was then applied to obtain an image from which the positions of the markers were determined. Two parallel-focused grid geometries were modeled, one based on the standard lead-carbon fiber grids and a proposed modification using tungsten as the shielding material. Absorbed dose in the patient model was also calculated. It is shown that the combination of primary and directional scatter images provides the means for an accurate stereoscopic fiducial marker 3D localization in a single x-ray exposure, provided the antiscatter grids are made thick enough to allow radiation traveling only in a particular direction to reach the detector. For the proposed tungsten grid and the x-ray spectrum used in this work, grid ratios of 20 and thickness of 0.2cm, provide the necessary shielding while for the standard lead grids, a ratio of at least 166 and a thickness of 2cm are needed to obtain discernible directional scatter images. We have shown that it is in principle possible to determine the 3D position of fiducial markers in a single exposure by making use of the radiation scattered by the patient to form an image that complements the information obtained with the primary beam. The method here proposed requires minimal modification of existing clinical hardware.

Online 4D ultrasound guidance for real-time motion compensation by MLC tracking.

With the trend in radiotherapy moving toward dose escalation and hypofractionation, the need for highly accurate targeting increases. While MLC tracking is already being successfully used for motion compensation of moving targets in the prostate, current real-time target localization methods rely on repeated x-ray imaging and implanted fiducial markers or electromagnetic transponders rather than direct target visualization. In contrast, ultrasound imaging can yield volumetric data in real-time (3D + time = 4D) without ionizing radiation. The authors report the first results of combining these promising techniques-online 4D ultrasound guidance and MLC tracking-in a phantom. A software framework for real-time target localization was installed directly on a 4D ultrasound station and used to detect a 2 mm spherical lead marker inside a water tank. The lead marker was rigidly attached to a motion stage programmed to reproduce nine characteristic tumor trajectories chosen from large databases (five prostate, four lung). The 3D marker position detected by ultrasound was transferred to a computer program for MLC tracking at a rate of 21.3 Hz and used for real-time MLC aperture adaption on a conventional linear accelerator. The tracking system latency was measured using sinusoidal trajectories and compensated for by applying a kernel density prediction algorithm for the lung traces. To measure geometric accuracy, static anterior and lateral conformal fields as well as a 358° arc with a 10 cm circular aperture were delivered for each trajectory. The two-dimensional (2D) geometric tracking error was measured as the difference between marker position and MLC aperture center in continuously acquired portal images. For dosimetric evaluation, VMAT treatment plans with high and low modulation were delivered to a biplanar diode array dosimeter using the same trajectories. Dose measurements with and without MLC tracking were compared to a static reference dose using 3%/3 mm and 2%/2 mm γ-tests. The overall tracking system latency was 172 ms. The mean 2D root-mean-square tracking error was 1.03 mm (0.80 mm prostate, 1.31 mm lung). MLC tracking improved the dose delivery in all cases with an overall reduction in the γ-failure rate of 91.2% (3%/3 mm) and 89.9% (2%/2 mm) compared to no motion compensation. Low modulation VMAT plans had no (3%/3 mm) or minimal (2%/2 mm) residual γ-failures while tracking reduced the γ-failure rate from 17.4% to 2.8% (3%/3 mm) and from 33.9% to 6.5% (2%/2 mm) for plans with high modulation. Real-time 4D ultrasound tracking was successfully integrated with online MLC tracking for the first time. The developed framework showed an accuracy and latency comparable with other MLC tracking methods while holding the potential to measure and adapt to target motion, including rotation and deformation, noninvasively.

SU‐F‐I‐11: Software Development for 4D‐CBCT Research of Real‐Time‐Image Gated Spot Scanning Proton Therapy

Purpose:To acquire correct information for inside the body in patient positioning of Real‐time‐image Gated spot scanning Proton Therapy (RGPT), utilization of tomographic image at exhale phase of patient respiration obtained from 4‐dimensional Cone beam CT (4D‐CBCT) has been desired. We developed software named “Image Analysis Platform” for 4D‐CBCT researches which has technique to segment projection‐images based on 3D marker position in the body. The 3D marker position can be obtained by using two axes CBCT system at Hokkaido University Hospital Proton Therapy Center. Performance verification of the software was implemented.Methods:The software calculates 3D marker position retrospectively by using matching positions on pair projection‐images obtained by two axes fluoroscopy mode of CBCT system. Log data of 3D marker tracking are outputted after the tracking. By linking the Log data and gantry‐angle file of projection‐image, all projection‐images are equally segmented to spatial five‐phases according to marker 3D position of SI direction and saved to specified phase folder. Segmented projection‐images are used for CBCT reconstruction of each phase. As performance verification of the software, test of segmented projection‐images was implemented for sample CT phantom (Catphan) image acquired by two axes fluoroscopy mode of CBCT. Dummy marker was added on the images. Motion of the marker was modeled to move in 3D space. Motion type of marker is sin4 wave function has amplitude 10.0 mm/5.0 mm/0 mm, cycle 4 s/4 s/0 s for SI/AP/RL direction.Results:The marker was tracked within 0.58 mm accuracy in 3D for all images, and it was confirmed that all projection‐images were segmented and saved to each phase folder correctly.Conclusion:We developed software for 4D‐CBCT research which can segment projection‐image based on 3D marker position. It will be helpful to create high quality of 4D‐CBCT reconstruction image for RGPT.

TH-AB-202-05: BEST IN PHYSICS (JOINT IMAGING-THERAPY): First Online Ultrasound-Guided MLC Tracking for Real-Time Motion Compensation in Radiotherapy

Purpose: While MLC tracking has been successfully used for motion compensation of moving targets, current real-time target localization methods rely on correlation models with x-ray imaging or implanted electromagnetic transponders rather than direct target visualization. In contrast, ultrasound imaging yields volumetric data in real-time (4D) without ionizing radiation. We report the first results of online 4D ultrasound-guided MLC tracking in a phantom. Methods: A real-time tracking framework was installed on a 4D ultrasound station (Vivid7 dimension, GE) and used to detect a 2mm spherical lead marker inside a water tank. The volumetric frame rate was 21.3Hz (47ms). The marker was rigidly attached to a motion stage programmed to reproduce nine tumor trajectories (five prostate, four lung). The 3D marker position from ultrasound was used for real-time MLC aperture adaption. The tracking system latency was measured and compensated by prediction for lung trajectories. To measure geometric accuracy, anterior and lateral conformal fields with 10cm circular aperture were delivered for each trajectory. The tracking error was measured as the difference between marker position and MLC aperture in continuous portal imaging. For dosimetric evaluation, 358° VMAT fields were delivered to a biplanar diode array dosimeter using the same trajectories. Dose measurements with and without MLC tracking were compared to a static reference dose using a 3%/3 mm γ-test. Results: The tracking system latency was 170ms. The mean root-mean-square tracking error was 1.01mm (0.75mm prostate, 1.33mm lung). Tracking reduced the mean γ-failure rate from 13.9% to 4.6% for prostate and from 21.8% to 0.6% for lung with high-modulation VMAT plans and from 5% (prostate) and 18% (lung) to 0% with low modulation. Conclusion: Real-time ultrasound tracking was successfully integrated with MLC tracking for the first time and showed similar accuracy and latency as other methods while holding the potential to measure target motion non-invasively. SI was supported by the Graduate School for Computing in Medicine and Life Science, German Excellence Initiative [grant DFG GSC 235/1].

Clinical applicability of a new 3D portable stereo-photogrammetric opto-electronic device

s / Gait & Posture 42S (2015) S1–S27 S5 Fig. 1. (a) Reference image. Model generated after the manual identification of the relevant anatomical landmarks. (b) Tracked model in a dynamic frame. They wore an underwear garment with a coloured patch applied, and uniformly coloured ankle socks used as segmentalmarkers (for pelvis and feet, respectively). A RGB video camera (Vicon Bonita 720C, 60 fps) was placed laterally to the walkway. Synchronous 3D marker positions were captured with a Vicon system (Bonita10). Markers were positioned according to the Davis protocol [3] (GS). The proposed ML methodology requires a preliminary model calibration step, performed on a reference image of the subject standing upright (Fig. 1a), from which a foot contour model template is automatically identified, while the lateral malleolus, lateral epicondyle and greater trochanter are manually identified [2]. Model templates of the shank and thigh segments are defined bymeans of technical reference points identified on the silhouette contour. The pelvis orientation is assumed to coincidewith the orientation of the segment obtained by interpolating the points extracted from the pelvis patch. For eachgait trial, the silhouette contourof themoving subject was extracted using a background subtraction technique, and the foot and pelvis segmental markers were identified using colour filters. The foot orientation was determined by matching the foot template with the foot current image. The shank and thigh reference points on the deformed silhouette contour were identified for each frame, and the corresponding templates fitted using a SVD procedure (Fig. 1b) [2]. The kinematic estimations provided by theMLmethodwere compared to the GS in terms of correlation coefficient (CC) and root mean square deviation (RMSD). Results: The CC (and RMSD) between the ML and the GS estimations ranged from 0.91 (4.9◦) to 0.98 (7.3◦) for the hip flexextension, from0.94 (5.4◦) to 0.97 (8.3◦) for the knee flex-extension and from0.74 (5.9◦) to0.84 (9.9◦) for theankledorsi-plantarflexion. Discussion: Joint kinematics estimatedusingML showed a good correlation with the GS joint kinematics (0.74≤CC≤0.98), however the relevant errors are clearly larger than those found in healthy adults [2] (up to 9.9◦). This difference can be attributed partly to the children’s body segments small size (particularly for the foot) and in some cases to the presence of the arm swing that reduces the visibility of the pelvis segmental marker. Therefore, the extraction of the segmental markers will have to be improved in order for this method to be applied in clinical settings.

Application of a markerless gait analysis methodology in children with cerebral palsy: Preliminary results

s / Gait & Posture 42S (2015) S1–S27 S5 Fig. 1. (a) Reference image. Model generated after the manual identification of the relevant anatomical landmarks. (b) Tracked model in a dynamic frame. They wore an underwear garment with a coloured patch applied, and uniformly coloured ankle socks used as segmentalmarkers (for pelvis and feet, respectively). A RGB video camera (Vicon Bonita 720C, 60 fps) was placed laterally to the walkway. Synchronous 3D marker positions were captured with a Vicon system (Bonita10). Markers were positioned according to the Davis protocol [3] (GS). The proposed ML methodology requires a preliminary model calibration step, performed on a reference image of the subject standing upright (Fig. 1a), from which a foot contour model template is automatically identified, while the lateral malleolus, lateral epicondyle and greater trochanter are manually identified [2]. Model templates of the shank and thigh segments are defined bymeans of technical reference points identified on the silhouette contour. The pelvis orientation is assumed to coincidewith the orientation of the segment obtained by interpolating the points extracted from the pelvis patch. For eachgait trial, the silhouette contourof themoving subject was extracted using a background subtraction technique, and the foot and pelvis segmental markers were identified using colour filters. The foot orientation was determined by matching the foot template with the foot current image. The shank and thigh reference points on the deformed silhouette contour were identified for each frame, and the corresponding templates fitted using a SVD procedure (Fig. 1b) [2]. The kinematic estimations provided by theMLmethodwere compared to the GS in terms of correlation coefficient (CC) and root mean square deviation (RMSD). Results: The CC (and RMSD) between the ML and the GS estimations ranged from 0.91 (4.9◦) to 0.98 (7.3◦) for the hip flexextension, from0.94 (5.4◦) to 0.97 (8.3◦) for the knee flex-extension and from0.74 (5.9◦) to0.84 (9.9◦) for theankledorsi-plantarflexion. Discussion: Joint kinematics estimatedusingML showed a good correlation with the GS joint kinematics (0.74≤CC≤0.98), however the relevant errors are clearly larger than those found in healthy adults [2] (up to 9.9◦). This difference can be attributed partly to the children’s body segments small size (particularly for the foot) and in some cases to the presence of the arm swing that reduces the visibility of the pelvis segmental marker. Therefore, the extraction of the segmental markers will have to be improved in order for this method to be applied in clinical settings.

SU-E-J-61: Monitoring Tumor Motion in Real-Time with EPID Imaging During Cervical Cancer Treatment

Purpose: Traditional external beam radiotherapy for cervical cancer requires setup by external skin marks. In order to improve treatment accuracy and reduce planning margin for more conformal therapy, it is essential to monitor tumor positions interfractionally and intrafractionally. We demonstrate feasibility of monitoring cervical tumor motion online using EPID imaging from Beam’s Eye View. Methods: Prior to treatment, 1∼2 cylindrical radio opaque markers were implanted into inferior aspect of cervix tumor. During external beam treatments on a Varian 2100C by 4-field 3D plans, treatment beam images were acquired continuously by an EPID. A Matlab program was developed to locate internal markers on MV images. Based on 2D marker positions obtained from different treatment fields, their 3D positions were estimated for every treatment fraction. Results: There were 398 images acquired during different treatment fractions of three cervical cancer patients. Markers were successfully located on every frame of image at an analysis speed of about 1 second per frame. Intrafraction motions were evaluated by comparing marker positions relative to the position on the first frame of image. The maximum intrafraction motion of the markers was 1.6 mm. Interfraction motions were evaluated by comparing 3D marker positions at different treatment fractions. The maximummore » interfraction motion was up to 10 mm. Careful comparison found that this is due to patient positioning since the bony structures shifted with the markers. Conclusion: This method provides a cost-free and simple solution for online tumor tracking for cervical cancer treatment since it is feasible to acquire and export EPID images with fast analysis in real time. This method does not need any extra equipment or deliver extra dose to patients. The online tumor motion information will be very useful to reduce planning margins and improve treatment accuracy, which is particularly important for SBRT treatment with long delivery time.« less

Six Degrees-of-Freedom Prostate and Lung Tumor Motion Measurements Using Kilovoltage Intrafraction Monitoring

Tumor positional uncertainty has been identified as a major issue that deteriorates the efficacy of radiation therapy. Tumor rotational movement, which is not well understood, can result in significant geometric and dosimetric inaccuracies. The objective of this study was to measure 6 degrees-of-freedom (6 DoF) prostate and lung tumor motion, focusing on the more novel rotation, using kilovoltage intrafraction monitoring (KIM). Continuous kilovoltage (kV) projections of tumors with gold fiducial markers were acquired during radiation therapy for 267 fractions from 10 prostate cancer patients and immediately before or after radiation therapy for 50 fractions from 3 lung cancer patients. The 6 DoF motion measurements were determined from the individual 3-dimensional (3D) marker positions, after using methods to reject spurious and smooth noisy data, using an iterative closest point algorithm. There were large variations in the magnitude of the tumor rotation among different fractions and patients. Various rotational patterns were observed. The average prostate rotation angles around the left-right (LR), superior-inferior (SI), and anterior-posterior (AP) axes were 1.0 ± 5.0°, 0.6 ± 3.3°, and 0.3 ± 2.0°, respectively. For 35% of the time, the prostate rotated more than 5° about the LR axis, indicating the need for intrafractional adaptation during radiation delivery. For lung patients, the average LR, SI, and AP rotation angles were 0.8 ± 4.2°, -0.8 ± 4.5°, and 1.7 ± 3.1°, respectively. For about 30% of the time, the lung tumors rotated more than 5° around the SI axis. Respiration-induced rotation was detected in 2 of the 3 lung patients. The prostate and lung tumors were found to undergo rotations of more than 5° for about a third of the time. The lung tumor data represent the first 6 DoF tumor motion measured by kV images. The 6 DoF KIM method can enable rotational and translational adaptive radiation therapy and potentially reduce treatment margins.

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

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

Real-time marker prediction and CoR estimation in optical motion capture

Optical motion capture systems suffer from marker occlusions resulting in loss of useful information. This paper addresses the problem of real-time joint localisation of legged skeletons in the presence of such missing data. The data is assumed to be labelled 3d marker positions from a motion capture system. An integrated framework is presented which predicts the occluded marker positions using a Variable Turn Model within an Unscented Kalman filter. Inferred information from neighbouring markers is used as observation states; these constraints are efficient, simple, and real-time implementable. This work also takes advantage of the common case that missing markers are still visible to a single camera, by combining predictions with under-determined positions, resulting in more accurate predictions. An Inverse Kinematics technique is then applied ensuring that the bone lengths remain constant over time; the system can thereby maintain a continuous data-flow. The marker and Centre of Rotation (CoR) positions can be calculated with high accuracy even in cases where markers are occluded for a long period of time. Our methodology is tested against some of the most popular methods for marker prediction and the results confirm that our approach outperforms these methods in estimating both marker and CoR positions.