Mean Translational Error Research Articles

Fast and automatic periacetabular osteotomy fragment pose estimation using intraoperatively implanted fiducials and single-view fluoroscopy

Accurate and consistent mental interpretation of fluoroscopy to determine the position and orientation of acetabular bone fragments in 3D space is difficult. We propose a computer assisted approach that uses a single fluoroscopic view and quickly reports the pose of an acetabular fragment without any user input or initialization. Intraoperatively, but prior to any osteotomies, two constellations of metallic ball-bearings (BBs) are injected into the wing of a patient’s ilium and lateral superior pubic ramus. One constellation is located on the expected acetabular fragment, and the other is located on the remaining, larger, pelvis fragment. The 3D locations of each BB are reconstructed using three fluoroscopic views and 2D/3D registrations to a preoperative CT scan of the pelvis. The relative pose of the fragment is established by estimating the movement of the two BB constellations using a single fluoroscopic view taken after osteotomy and fragment relocation. BB detection and inter-view correspondences are automatically computed throughout the processing pipeline. The proposed method was evaluated on a multitude of fluoroscopic images collected from six cadaveric surgeries performed bilaterally on three specimens. Mean fragment rotation error was 2.4 ± 1.0 degrees, mean translation error was 2.1 ± 0.6 mm, and mean 3D lateral center edge angle error was 1.0 ± 0.5 degrees. The average runtime of the single-view pose estimation was 0.7 ± 0.2 s. The proposed method demonstrates accuracy similar to other state of the art systems which require optical tracking systems or multiple-view 2D/3D registrations with manual input. The errors reported on fragment poses and lateral center edge angles are within the margins required for accurate intraoperative evaluation of femoral head coverage.

Fast and efficient reconstruction of digitized frescoes

Abstract Virtually recomposing destroyed frescoes is of great importance for heritage conservation. Given a digitized fresco image and a digitized set of fragments, such a problem is challenging due to the potentially large number of fragments, their irregular shape, uniqueness and non-overlapping constraints, the possible absence of fragments and the possible presence of small, homogeneous, eroded and/or spurious fragments. To cope with these specific features, we propose in this paper a fast and efficient non-dense approach benefiting from previous developments in pattern matching. Preliminary experiments led on simulations exhibit a mean accuracy above 90% with a mean translation error of less than 4 pixels and a mean orientation error of about 1 degree. An analysis of fresco and fragment features impacting the algorithm is also provided. Compared to a dense approach and the recent DeepMatch approach, the proposed one remains competitive both in running time and accuracy.

Fast and Accurate Spacecraft Pose Estimation From Single Shot Space Imagery Using Box Reliability and Keypoints Existence Judgments

Real-time 6DOF (6 Degree of Freedom) pose estimation of an uncooperative spacecraft is an important part of proximity operations, e.g., space debris removal, spacecraft rendezvous and docking, on-orbit servicing, etc. In this article, a novel efficient deep learning based approach is proposed to estimate the 6DOF pose of uncooperative spacecraft using monocular-vision measurement. Firstly, we introduce a new lightweight YOLO-liked CNN to detect spacecraft and predict 2D locations of the projected keypoints of a prior reconstructed 3D model in real-time. Then, we design two novel models for predicting the bounding box (bbox) reliability scores and the probability of keypoints existence. The two models not only significantly reduce the false positive, but also speed up convergence. Finally, the 6DOF pose is estimated and refined using Perspective-n-Point and geometric optimizer. Results demonstrate that the proposed approach achieves 73.2% average precision and 77.6% average recall for spacecraft detection on the SPEED dataset after only 200 training epochs. For the pose estimation task, the mean rotational error is 0.6812°, and the mean translation error is 0.0320m. The proposed approach achieves competitive pose estimation performance and extreme lightweight ( $\sim ~0.89$ million learnable weights in total) on the SPEED dataset while being efficient for real-time applications.

Automatic Markerless Registration and Tracking of the Bone for Computer-Assisted Orthopaedic Surgery

To achieve a simple and less invasive registration procedure in computer-assisted orthopaedic surgery, we propose an automatic, markerless registration and tracking method based on depth imaging and deep learning. A depth camera is used to continuously capture RGB and depth images of the exposed bone during surgery, and deep neural networks are trained to first localise the surgical target using the RGB image, then segment the target area of the corresponding depth image, from which the surface geometry of the target bone can be extracted. The extracted surface is then compared to a pre-operative model of the same bone for registration. This process can be performed dynamically during the procedure at a rate of 5-6 Hz, without any need for surgeon intervention or invasive optical markers. Ex vivo registration experiments were performed on a cadaveric knee, and accuracy measurements against an optically tracked ground truth resulted in a mean translational error of 2.74 mm and a mean rotational error of 6.66°. Our results are the first to describe a promising new way to achieve automatic markerless registration and tracking in computer-assisted orthopaedic surgery, demonstrating that truly seamless registration and tracking of the limb is within reach. Our method reduces invasiveness by removing the need for percutaneous markers. The surgeon is also exempted from inserting markers and collecting registration points manually, which contributes to a more efficient surgical workflow and shorter procedure time in the operating room.

Mobilization and calibration of the HTC VIVE for virtual reality physical therapy.

AimsThe HTC VIVE virtual reality (VR) system is a potential tool for collecting kinematic data during inpatient and outpatient physical therapy (PT). When validated against research-grade systems, the VIVE has a reported translational error between 1.7 mm–2.0 cm. Our purpose was to portabilize the VIVE for room to room PT and validate the motion tracking software.MethodsThe VIVE was configured on a mobile cart. To validate the motion tracking software, the VIVE sensors (motion tracker, controller, headset) were mounted on a rigid linear track and driven through 10, one-meter translations in the X, Y, and Z axes.ResultsThe mean translational error for all three sensors was below 4.9 cm. While error is greater than that reported for research-grade systems, motion tracking software on the portable VIVE unit appears to be a valid means of tracking aggregate movement.ConclusionSome therapy may require more precise measurements, however, the advantages of portability and accessibility to patients may outweigh the limitation of reduced precision.

Augmented reality navigation for spinal pedicle screw instrumentation using intraoperative 3D imaging

BACKGROUND CONTEXTDue to recent developments in augmented reality with head-mounted devices, holograms of a surgical plan can be displayed directly in the surgeon's field of view. To the best of our knowledge, three dimensional (3D) intraoperative fluoroscopy has not been explored for the use with holographic navigation by head-mounted devices in spine surgery. PURPOSETo evaluate the surgical accuracy of holographic pedicle screw navigation by head-mounted device using 3D intraoperative fluoroscopy. STUDY DESIGNIn this experimental cadaver study, the accuracy of surgical navigation using a head-mounted device was compared with navigation with a state-of-the-art pose-tracking system. METHODSThree lumbar cadaver spines were embedded in nontransparent agar gel, leaving only commonly visible anatomy in sight. Intraoperative registration of preoperative planning was achieved by 3D fluoroscopy and fiducial markers attached to lumbar vertebrae. Trackable custom-made drill sleeve guides enabled real-time navigation. In total, 20 K-wires were navigated into lumbar pedicles using AR-navigation, 10 K-wires by the state-of-the-art pose-tracking system. 3D models obtained from postexperimental CT scans were used to measure surgical accuracy. MF is the founder and shareholder of Incremed AG, a Balgrist University Hospital start-up focusing on the development of innovative techniques for surgical executions. The other authors declare no conflict of interest concerning the contents of this study. No external funding was received for this study. RESULTSNo significant difference in accuracy was measured between AR-navigated drillings and the gold standard with pose-tracking system with mean translational errors between entry points (3D vector distance; p=.85) of 3.4±1.6 mm compared with 3.2±2.0 mm, and mean angular errors between trajectories (3D angle; p=.30) of 4.3°±2.3° compared with 3.5°±1.4°. CONCLUSIONSIn conclusion, holographic navigation by use of a head-mounted device achieve accuracy comparable to the gold standard of high-end pose-tracking systems. CLINICAL SIGNIFICANCEThese promising results could result in a new way of surgical navigation with minimal infrastructural requirements but now have to be confirmed in clinical studies.

Real-Time 6DOF Pose Estimation of Endoscopic Instruments Using Printable Markers

To accurately navigate surgical instruments around in vivo environments in minimally invasive surgeries through flexible endoscopes or rigid scopes, real-time tracking of surgical instruments is required. It is even challenging to track regular-shaped rigid bodies on a less-distorted general-purpose monocular vision, while most of the surgical instruments or devices used are cylindrical and slim imaged with endoscope distortion in nature. Thus, more dedicated and accurate rigid-body tracking approach for cylindrical endoscopic instruments will be more helpful in intra-operative guidance. In this paper, we present an endoscopic instrument tracking approach with printable markers that consists of a green band and a square with four white and one black circle inscribed in it. The proposed approach can estimate the six degrees of freedom pose, i.e., X, Y, Z, Roll, Pitch, and Yaw, of the imaged instrument with respect to the endoscope. Detailed experiments justify that our proposed marker detection framework allows translation motions between 42 and 100 mm along depth direction and rotational motions around X and Y up to ±180o, with 90% detection rate in dynamic condition, 99% in static condition, and accurate pose estimation with a mean translation error of 1.286 mm with a standard deviation of 0.673 mm and a mean rotational error of 1.497° with a standard deviation of 0.873°. Our proposed marker can be easily printed and wrapped to the existing endoscopic instruments, which potentially can assist in the precise guidance in minimally invasive surgeries without changing the existing workflow or incurring significant cost burdens.

Feasibility of using a fully immersive virtual reality system for kinematic data collection

Commercially-available Virtual Reality (VR) systems have the potential to be effective tools for simultaneous visual manipulation and kinematic data collection. Previously, these systems have been integrated with research-grade motion capture systems to provide both functionalities; however, they are yet to be used as stand-alone systems for kinematic data collection. The present study aimed to validate the HTC VIVE VR system for kinematic data collection by evaluating the accuracy of its position and orientation signals. The VIVE controller and tracker were each compared to a Polhemus Liberty magnetic tracking system sensor for angular and translational measurement error and signal drift. A sensor from each system was mounted to opposite ends of a rigid segment which was driven through fifty rotations and fifty translations. Mean angular errors for both the VIVE tracker and controller were below 0.4°. Mean translational error for both sensors was below 3 mm. Drift in the Liberty signal components was consistently lower than drift in VIVE components. However, all mean rotational drift measures were below 0.1° and all mean translational measures were below 0.35 mm. These data indicate that the HTC VIVE system has the potential to be a valid and reliable means of kinematic data collection. However, further investigation is necessary to determine the VIVE’s suitability for capturing extremely minute or high-volume movements.

Investigation into the accuracy and measurement methods of sequential 3D dental scan alignment

ObjectivesAlignment procedures have yet to be standardised and may influence the measurement outcome. This investigation assessed the accuracy of commonly used alignment techniques and their impact on measurement metrics. MethodsDatasets of 10 natural molar teeth were created with a structured-light model-scanner (Rexcan DS2, Europac 3D, Crewe). A 300μm depth layer was then digitally removed from the occlusal surface creating a defect of known size. The datasets were duplicated, randomly repositioned and re-alignment attempted using a “best-fit” alignment, landmark-based alignment or reference alignment in Geomagic Control (3D Systems, Darmstadt, Germany). The re-alignment accuracy was mathematically assessed using the mean angular and translation differences between the original alignment and the re-aligned datasets. The effect of the re-alignment on conventional measurement metrics was calculated by analysing differences between the known defect size and defect size after re-alignment. Data were analysed in SPSS v24(ANOVA, post hoc Games Howell test, p<0.05). ResultsThe mean translation error (SD) was 139μm (42) using landmark alignment, 130μm (26) for best-fit and 22μm (9) for reference alignment (p<0.001). The mean angular error (SD) between the datasets was 2.52 (1.18) degrees for landmark alignment, 0.56 (0.38) degrees for best-fit alignment and 0.26 (0.12) degrees for reference alignment (p<0.001). Using a reference alignment statistically reduced the mean profilometric change, volume change and percentage of surface change errors (p<0.001). SignificanceReference alignment produced significantly lower alignment errors and truer measurements. Best-fit and landmark-based alignment algorithms significantly underestimated the size of the defect. Challenges remain in identifying reference surfaces in a robust, clinically relevant method.

Volumetric 3D stitching of optical coherence tomography volumes

Abstract Optical coherence tomography (OCT) is a noninvasive medical imaging modality, which provides highresolution transectional images of biological tissue. However, its potential is limited due to a relatively small field of view. To overcome this drawback, we describe a scheme for fully automated stitching of multiple 3D OCT volumes for panoramic imaging. The voxel displacements between two adjacent images are calculated by extending the Lucas-Kanade optical flow a lgorithm to dense volumetric images. A RANSAC robust estimator is used to obtain rigid transformations out of the resulting flow v ectors. T he i mages a re t ransformed into the same coordinate frame and overlapping areas are blended. The accuracy of the proposed stitching scheme is evaluated on two datasets of 7 and 4 OCT volumes, respectively. By placing the specimens on a high-accuracy motorized translational stage, ground truth transformations are available. This results in a mean translational error between two adjacent volumes of 16.6 ± 0.8 μm (2.8 ± 0.13 voxels). To the author’s knowledge, this is the first reported stitching of multiple 3D OCT volumes by using dense voxel information in the registration process. The achieved results are sufficient for providing high accuracy OCT panoramic images. Combined with a recently available high-speed 4D OCT, our method enables interactive stitching of hand-guided acquisitions.

Accuracy in identifying the elbow rotation axis on simulated fluoroscopic images using a new anatomical landmark

External fixation of the elbow requires identification of the elbow rotation axis, but the accuracy of traditional landmarks (capitellum and trochlea) on fluoroscopy is limited. The relative distance (RD) of the humerus may be helpful as additional landmark. The first aim of this study was to determine the optimal RD that corresponds to an on-axis lateral image of the elbow. The second aim was to assess whether the use of the optimal RD improves the surgical accuracy to identify the elbow rotation axis on fluoroscopy. CT scans of elbows from five volunteers were used to simulate fluoroscopy; the actual rotation axis was calculated with CT-based flexion–extension analysis. First, three observers measured the optimal RD on simulated fluoroscopy. The RD is defined as the distance between the dorsal part of the humerus and the projection of the posteromedial cortex of the distal humerus, divided by the anteroposterior diameter of the humerus. Second, eight trauma surgeons assessed the elbow rotation axis on simulated fluoroscopy. In a preteaching session, surgeons used traditional landmarks. The surgeons were then instructed how to use the optimal RD as additional landmark in a postteaching session. The deviation from the actual rotation axis was expressed as rotational and translational error (±SD). Measurement of the RD was robust and easily reproducible; the optimal RD was 45%. The surgeons identified the elbow rotation axis with a mean rotational error decreasing from 7.6° ± 3.4° to 6.7° ± 3.3° after teaching how to use the RD. The mean translational error decreased from 4.2 ± 2.0 to 3.7 ± 2.0 mm after teaching. The humeral RD as additional landmark yielded small but relevant improvements. Although fluoroscopy-based external fixator alignment to the elbow remains prone to error, it is recommended to use the RD as additional landmark.

Introduction of a computer-based method for automated planning of reduction paths under consideration of simulated muscular forces.

Reduction is a crucial step in the surgical treatment of bone fractures. Finding an optimal path for restoring anatomical alignment is considered technically demanding because collisions as well as high forces caused by surrounding soft tissues can avoid desired reduction movements. The repetition of reduction movements leads to a trial-and-error process which causes a prolonged duration of surgery. By planning an appropriate reduction path-an optimal sequence of target-directed movements-these problems should be overcome. For this purpose, a computer-based method has been developed. Using the example of simple femoral shaft fractures, 3D models are generated out of CT images. A reposition algorithm aligns both fragments by reconstructing their broken edges. According to the criteria of a deduced planning strategy, a modified A*-algorithm searches collision-free route of minimal force from the dislocated into the computed target position. Muscular forces are considered using a musculoskeletal reduction model (OpenSim model), and bone collisions are detected by an appropriate method. Five femoral SYNBONE models were broken into different fracture classification types and were automatically reduced from ten randomly selected displaced positions. Highest mean translational and rotational error for achieving target alignment is [Formula: see text] and [Formula: see text]. Mean value and standard deviation of occurring forces are [Formula: see text] for M. tensor fasciae latae and [Formula: see text] for M. semitendinosus over all trials. These pathways are precise, collision-free, required forces are minimized, and thus regarded as optimal paths. A novel method for planning reduction paths under consideration of collisions and muscular forces is introduced. The results deliver additional knowledge for an appropriate tactical reduction procedure and can provide a basis for further navigated or robotic-assisted developments.

Geometric calibration algorithm of polarization camera using planar patterns

Polarization pattern provides additional information besides spectral signatures. It can be used in many applications, such as navigation, defect detection and object identification. A novel polarization camera composed of four synchronized cameras is proposed, and it can realize real-time polarization measurement. This study particularly concentrates on the geometric calibration of the system. The projection model is analyzed and the multi-camera calibration algorithm is proposed. Firstly, each camera is calibrated separately using planar patterns, and then the geometric calibration algorithms are performed. Due to the geometrical constraint, a global optimization method results in smaller estimation uncertainties. A mean rotation error of 0.025° and a mean translation error of 0.26mm are achieved after geometric calibration. The images are rectified to establish a correspondence among cameras and are combined to acquire the polarization measurement. The polarization pattern of the skylight is measured by the system and the results are consistent with the previous studies.

Positional Accuracy of Treating Multiple Versus Single Vertebral Metastases With Stereotactic Body Radiotherapy.

The aim of this study is to determine whether stereotactic body radiotherapy for multiple vertebral metastases treated with a single isocenter results in greater intrafraction errors than stereotactic body radiotherapy for single vertebral metastases and to determine whether the currently used spinal cord planning organ at risk volume and planning target volume margins are appropriate. Intrafraction errors were assessed for 65 stereotactic body radiotherapy treatments for vertebral metastases. Cone beam computed tomography images were acquired before, during, and after treatment for each fraction. Residual translational and rotational errors in patient positioning were recorded and planning organ at risk volume and planning target volume margins were calculated in each direction using this information. The mean translational residual errors were smaller for single (0.4 (0.4) mm) than for multiple vertebral metastases (0.5 (0.7) mm; P = .0019). The mean rotational residual errors were similar for single (0.3° (0.3°) and multiple vertebral metastases (0.3° (0.3°); P = .862). The maximum calculated planning organ at risk volume margin in any direction was 0.83 mm for single and 1.22 for multiple vertebral metastases. The maximum calculated planning target volume margin in any direction was 1.4 mm for single and 1.9 mm for multiple vertebral metastases. Intrafraction errors were small for both single and multiple vertebral metastases, indicating that our strategy for patient immobilization and repositioning is robust. Calculated planning organ at risk volume and planning target volume margins were smaller than our clinically employed margins, indicating that our clinical margins are appropriate.

Accuracy and Precision of Head Motion Information in Multi-Channel Free Induction Decay Navigators for Magnetic Resonance Imaging.

Free induction decay (FID) navigators were found to qualitatively detect rigid-body head movements, yet it is unknown to what extent they can provide quantitative motion estimates. Here, we acquired FID navigators at different sampling rates and simultaneously measured head movements using a highly accurate optical motion tracking system. This strategy allowed us to estimate the accuracy and precision of FID navigators for quantification of rigid-body head movements. Five subjects were scanned with a 32-channel head coil array on a clinical 3T MR scanner during several resting and guided head movement periods. For each subject we trained a linear regression model based on FID navigator and optical motion tracking signals. FID-based motion model accuracy and precision was evaluated using cross-validation. FID-based prediction of rigid-body head motion was found to be with a mean translational and rotational error of 0.14±0.21 mm and 0.08±0.13°, respectively. Robust model training with sub-millimeter and sub-degree accuracy could be achieved using 100 data points with motion magnitudes of ±2 mm and ±1° for translation and rotation. The obtained linear models appeared to be subject-specific as inter-subject application of a "universal" FID-based motion model resulted in poor prediction accuracy. The results show that substantial rigid-body motion information is encoded in FID navigator signal time courses. Although, the applied method currently requires the simultaneous acquisition of FID signals and optical tracking data, the findings suggest that multi-channel FID navigators have a potential to complement existing tracking technologies for accurate rigid-body motion detection and correction in MRI.

High degree of accuracy of a novel image-free handheld robot for unicondylar knee arthroplasty in a cadaveric study.

Surgical robotics has been shown to improve the accuracy of bone preparation and soft tissue balance in unicondylar knee arthroplasty (UKA). However, although extensive data have emerged with regard to a CT scan-based haptically constrained robotic arm, little is known about the accuracy of a newer alternative, an imageless robotic system. We assessed the accuracy of a novel imageless semiautonomous freehand robotic sculpting system in performing bone resection and preparation in UKA using cadaveric specimens. In this controlled study, we compared the planned and final implant placement in 25 cadaveric specimens undergoing UKA using the new tool. A quantitative analysis was performed to determine the translational, angular, and rotational differences between the planned and achieved positions of the implants. The femoral implant rotational mean error was 1.04° to 1.88° and mean translational error was 0.72 to 1.29mm across the three planes. The tibial implant rotational mean error was 1.48° to 1.98° and the mean translational error was 0.79 to 1.27mm across the three planes. The image-free robotic sculpting tool achieved accurate implementation of the surgical plan with small errors in implant placement. The next step will be to determine whether accurate implant placement translates into a clinical and functional benefit for the patient.

SU-E-T-132: Dosimetric Impact of Positioning Errors in Hypo-Fractionated Cranial Radiation Therapy Using Frameless Stereotactic BrainLAB System

Purpose: To determine dosimetric impact of positioning errors in the stereotactic hypo-fractionated treatment of intracranial lesions using 3Dtransaltional and 3D-rotational corrections (6D) frameless BrainLAB ExacTrac X-Ray system. Methods: 20 cranial lesions, treated in 3 or 5 fractions, were selected. An infrared (IR) optical positioning system was employed for initial patient setup followed by stereoscopic kV X-ray radiographs for position verification. 6D-translational and rotational shifts were determined to correct patient position. If these shifts were above tolerance (0.7 mm translational and 1° rotational), corrections were applied and another set of X-rays was taken to verify patient position. Dosimetric impact (D95, Dmin, Dmax, and Dmean of planning target volume (PTV) compared to original plans) of positioning errors for initial IR setup (XC: Xray Correction) and post-correction (XV: X-ray Verification) was determined in a treatment planning system using a method proposed by Yue et al. (Med. Phys. 33, 21-31 (2006)) with 3D-translational errors only and 6D-translational and rotational errors. Results: Absolute mean translational errors (±standard deviation) for total 92 fractions (XC/XV) were 0.79±0.88/0.19±0.15 mm (lateral), 1.66±1.71/0.18 ±0.16 mm (longitudinal), 1.95±1.18/0.15±0.14 mm (vertical) and rotational errors were 0.61±0.47/0.17±0.15° (pitch), 0.55±0.49/0.16±0.24° (roll), and 0.68±0.73/0.16±0.15° (yaw). The average changes (loss of coverage) in D95, Dmin, Dmax, and Dmean were 4.5±7.3/0.1±0.2%, 17.8±22.5/1.1±2.5%, 0.4±1.4/0.1±0.3%, and 0.9±1.7/0.0±0.1% using 6Dshifts and 3.1±5.5/0.0±0.1%, 14.2±20.3/0.8±1.7%, 0.0±1.2/0.1±0.3%, and 0.7±1.4/0.0±0.1% using 3D-translational shifts only. The setup corrections (XC-XV) improved the PTV coverage by 4.4±7.3% (D95) and 16.7±23.5% (Dmin) using 6D adjustment. Strong correlations were observed between translation errors and deviations in dose coverage for XC. Conclusion: The initial BrainLAB IR system based on rigidity of the mask-frame setup is not sufficient for accurate stereotactic positioning; however, with X-ray imageguidance sub-millimeter accuracy is achieved with negligible deviations in dose coverage. The angular corrections (mean angle summation=1.84°) are important and cause considerable deviations in dose coverage.

Computer-assisted fracture reduction: a new approach for repositioning femoral fractures and planning reduction paths.

Reduction is a crucial step in the surgical treatment of bone fractures to achieve anatomical alignment and facilitate healing. Surgical planning for treatment of simple femoral fractures requires suitable gentle reduction paths. A plan with optimal movement of fracture fragments from the initial to the desired target position should improve the reduction procedure. A virtual environment which repositions the fracture fragments automatically and provides the ability to plan reduction paths was developed and tested. Virtual 3D osseous fragments are created from CT scans. Based on the computed surface curvatures, strongly curved edges are selected and fracture lines are generated. After assignment of matching points, the lines are compared and the desired target position is calculated. Planning of reduction paths was achieved using a reference-coordinate-system for the computation of reduction parameters. The fracture is reduced by changing the reduction parameters step by step until the target position is reached. To test this system, nine different fractured SYNBONE models and one human fracture were reduced, based on CT scans with varying resolution. The highest mean translational error is 1.2 ± 0.9 (mm), and the rotational error is 2.6 ± 2.8 (°), both of which are considered as clinically acceptable. The reduction paths can be planned manually or semi-automatically for each fracture. Automated fracture reduction was achieved using a system based on preoperative CT scans. The automated system provides a clinically feasible basis for planning optimal reduction paths that may be augmented by further computer- or robot-assisted applications.

Surgical Accuracy in Identifying the Elbow Rotation Axis on Fluoroscopic Images

To assess the accuracy of surgeons in identifying elbow rotation axis (RA) on fluoroscopic images and to measure the interobserver variability. Five healthy subjects underwent 3-dimensional computed tomography (CT) analysis of their nondominant elbow. Real-time rotation software enabled surgeons to approximate the elbow RA on CT-reconstructed fluoroscopy, which was repeated twice with different starting positions to increase the number of observations. The surgeons used anatomical landmarks of choice. Analysis of variance (ANOVA) was used to determine structural error differences between surgeons, and intraclass correlation coefficients (ICCs) were used to determine the corresponding interobserver variability. Eight subspecialty-trained trauma surgeons (P.K., N.W.L.S., V.M.d.J., P.J., G.M.K., R.W.P., T.S., B.A.v.D.) participated and attempted to identify the RA on reconstructed fluoroscopy. A total of 15 RA definitions on 5 elbows were recorded per surgeon. The surgeons had a mean rotational error of 5° (range, < 1°-13°) and mean translational error of 1 mm (range, < 1-8 mm), compared with the true elbow RA as measured by the 3-dimensional CT analysis. The ANOVA showed structural differences between surgeons in rotational and translational errors, indicating that some surgeons consistently had more accurately identified the elbow RA than others. The ICC was 0.12 for rotational error and 0.10 for translational error, indicating a large interobserver variability. We show in this in vivo study that identification of the elbow RA on fluoroscopy is associated with substantial rotational errors and large inconsistencies among surgeons. Implementation of standardized anatomical landmarks is required to improve surgeons' accuracy. These landmarks should preferably take into account both the coronal and the sagittal planes, using the orientation of the capitellum and trochlea as well as the posterior distal humeral cortex. Diagnostic II.

MO‐D‐144‐02: Ultrasound Transducer Localization Using the CyberKnife's X‐Ray System

Purpose: 4D ultrasound has become an alternative for image guidance and motion compensation in radiosurgery. Nevertheless, a two‐step localization has to be performed when using ultrasound. In addition to target localization inside the ultrasound volume, the transducer itself has to be localized and the target position has to be transformed into treatment coordinates. Methods: The CyberKnife (Accuray Inc.) features a stereo X‐ray system which is used for patient and fiducial marker localization. Accessing only the raw X‐ray images, we designed a software package for additional marker detection. To measure the physical camera parameters, X‐ray phantoms were designed and a non‐orthogonal stereo camera calibration was performed on 50 calibration X‐ray pairs. Algorithms were developed for three‐dimensional single marker localization and six‐dimensional localization of marker geometries. Different X‐ray marker geometries were designed and attached to the ultrasound transducer. The final system was evaluated and compared to the on‐board localization system using a 6‐axis industrial robot to position different marker geometries at 150 randomly distributed positions over the acquisition volume. Results: Two marker geometries with 20 and 50mm base lengths were localized using the on‐board and newly developed software. The mean translational error for the on‐board localization is 0.202mm, for the new system 0.218mm. The rotational error of the on‐board system could be reduced from 0.212 to 0.053 degrees for the 50mm marker and from 0.272 to 0.076 degrees for the 20mm marker geometry. Conclusion: We have shown that ultrasound transducers can be localized using the CyberKnife x‐ray system. The rotational accuracy of the localization could be increased by a factor of four, which is important for high marker‐to‐ultrasound‐target distances. Furthermore, using the full area of the flat‐panel detectors without pre‐processing steps the tracking volume was increased by 70 percent, which helps detecting patient/fiducials and transducer at the same time.