Mean Target Registration Error Research Articles

Estimation of motion fields by non-linear registration for local lung motion analysis in 4D CT image data

Motivated by radiotherapy of lung cancer non- linear registration is applied to estimate 3D motion fields for local lung motion analysis in thoracic 4D CT images. Reliability of analysis results depends on the registration accuracy. Therefore, our study consists of two parts: optimization and evaluation of a non-linear registration scheme for motion field estimation, followed by a registration-based analysis of lung motion patterns. The study is based on 4D CT data of 17 patients. Different distance measures and force terms for thoracic CT registration are implemented and compared: sum of squared differences versus a force term related to Thirion's demons registration; masked versus unmasked force computation. The most accurate approach is applied to local lung motion analysis. Masked Thirion forces outperform the other force terms. The mean target registration error is 1.3 ± 0.2 mm, which is in the order of voxel size. Based on resulting motion fields and inter-patient normalization of inner lung coordinates and breathing depths a non-linear dependency between inner lung position and corresponding strength of motion is identified. The dependency is observed for all patients without or with only small tumors. Quantitative evaluation of the estimated motion fields indicates high spatial registration accuracy. It allows for reliable registration-based local lung motion analysis. The large amount of information encoded in the motion fields makes it possible to draw detailed conclusions, e.g., to identify the dependency of inner lung localization and motion. Our examinations illustrate the potential of registration-based motion analysis.

Electromagnetic Navigation for Thoracic Aortic Stent-graft Deployment: A Pilot Study in Swine

To determine the feasibility of electromagnetic tracking as a method to augment conventional imaging guidance for the safe delivery, precise positioning, and accurate deployment of thoracic aortic endografts. Custom guide wires were fabricated, and the delivery catheters for thoracic aortic endoprostheses were retrofitted with integrated electromagnetic coil sensors to enable real-time endovascular tracking. Preprocedure thoracic computed tomographic (CT) angiograms were obtained after the placement of fiducial skin patches on the chest wall of three anesthetized swine, enabling automatic registration. The stent-graft deployment location target near the subclavian artery was selected on the preprocedure CT angiogram. Two steps were analyzed: advancing a tracked glidewire to the aortic arch and positioning the tracked stent-graft assembly by using electromagnetic guidance alone. Multiple CT scans were obtained to evaluate the accuracy of the electromagnetic tracking system by measuring the target registration error, which compared the actual position of the tracked devices to the displayed "virtual" electromagnetic-tracked position. Postdeployment CT angiography and necropsy helped confirm stent-graft position and subclavian artery patency. A stent-graft was successfully delivered and deployed in each of the three animals by using real-time electromagnetic tracking alone. The mean fiducial registration error with autoregistration was 1.5 mm. Sixteen comparative scans were obtained to determine the target registration error, which was 4.3 mm +/- 0.97 (range, 3.0-6.0 mm) for the glidewire sensor coil. The mean target registration error for the stent-graft delivery catheter sensor coil was 2.6 mm +/- 0.7 (range, 1.9-3.8 mm). The mean deployment error for the stent-graft, defined as deployment deviation from the target, was 2.6 mm +/- 3.0. Delivery and deployment of customized thoracic stent-grafts with use of electromagnetic tracking alone is feasible and accurate in swine. Combining endovascular electromagnetic tracking with conventional fluoroscopy may further improve accuracy and be a more realistic multimodality approach.

Robust initialization of 2D-3D image registration using the projection-slice theorem and phase correlation

The image registration literature comprises many methods for 2D-3D registration for which accuracy has been established in a variety of applications. However, clinical application is limited by a small capture range. Initial offsets outside the capture range of a registration method will not converge to a successful registration. Previously reported capture ranges, defined as the 95% success range, are in the order of 4-11 mm mean target registration error. In this article, a relatively computationally inexpensive and robust estimation method is proposed with the objective to enlarge the capture range. The method uses the projection-slice theorem in combination with phase correlation in order to estimate the transform parameters, which provides an initialization of the subsequent registration procedure. The feasibility of the method was evaluated by experiments using digitally reconstructed radiographs generated from in vivo 3D-RX data. With these experiments it was shown that the projection-slice theorem provides successful estimates of the rotational transform parameters for perspective projections and in case of translational offsets. The method was further tested on ex vivo ovine x-ray data. In 95% of the cases, the method yielded successful estimates for initial mean target registration errors up to 19.5 mm. Finally, the method was evaluated as an initialization method for an intensity-based 2D-3D registration method. The uninitialized and initialized registration experiments had success rates of 28.8% and 68.6%, respectively. The authors have shown that the initialization method based on the projection-slice theorem and phase correlation yields adequate initializations for existing registration methods, thereby substantially enlarging the capture range of these methods.

Automatic Parameter Selection for Multimodal Image Registration

Over the past ten years similarity measures based on intensity distributions have become state-of-the-art in automatic multimodal image registration. An implementation for clinical usage has to support a plurality of images. However, a generally applicable parameter configuration for the number and sizes of histogram bins, optimal Parzen-window kernel widths or background thresholds cannot be found. This explains why various research groups present partly contradictory empirical proposals for these parameters. This paper proposes a set of data-driven estimation schemes for a parameter-free implementation that eliminates major caveats of heuristic trial and error. We present the following novel approaches: a new coincidence weighting scheme to reduce the influence of background noise on the similarity measure in combination with Max-Lloyd requantization, and a tradeoff for the automatic estimation of the number of histogram bins. These methods have been integrated into a state-of-the-art rigid registration that is based on normalized mutual information and applied to CT-MR, PET-MR, and MR-MR image pairs of the RIRE 2.0 database. We compare combinations of the proposed techniques to a standard implementation using default parameters, which can be found in the literature, and to a manual registration by a medical expert. Additionally, we analyze the effects of various histogram sizes, sampling rates, and error thresholds for the number of histogram bins. The comparison of the parameter selection techniques yields 25 approaches in total, with 114 registrations each. The number of bins has no significant influence on the proposed implementation that performs better than both the manual and the standard method in terms of acceptance rates and target registration error (TRE). The overall mean TRE is 2.34 mm compared to 2.54 mm for the manual registration and 6.48 mm for a standard implementation. Our results show a significant TRE reduction for distortion-corrected magnetic resonance images.

Assessment of image registration accuracy in three‐dimensional transrectal ultrasound guided prostate biopsy

Prostate biopsy, performed using two-dimensional (2D) transrectal ultrasound (TRUS) guidance, is the clinical standard for a definitive diagnosis of prostate cancer. Histological analysis of the biopsies can reveal cancerous, noncancerous, or suspicious, possibly precancerous, tissue. During subsequent biopsy sessions, noncancerous regions should be avoided, and suspicious regions should be precisely rebiopsied, requiring accurate needle guidance. It is challenging to precisely guide a needle using 2D TRUS due to the limited anatomic information provided, and a three-dimensional (3D) record of biopsy locations for use in subsequent biopsy procedures cannot be collected. Our tracked, 3D TRUS-guided prostate biopsy system provides additional anatomic context and permits a 3D record of biopsies. However, targets determined based on a previous biopsy procedure must be transformed during the procedure to compensate for intraprocedure prostate shifting due to patient motion and prostate deformation due to transducer probe pressure. Thus, registration is a critically important step required to determine these transformations so that correspondence is maintained between the prebiopsied image and the real-time image. Registration must not only be performed accurately, but also quickly, since correction for prostate motion and deformation must be carried out during the biopsy procedure. The authors evaluated the accuracy, variability, and speed of several surface-based and image-based intrasession 3D-to-3D TRUS image registration techniques, for both rigid and nonrigid cases, to find the required transformations. Our surface-based rigid and nonrigid registrations of the prostate were performed using the iterative-closest-point algorithm and a thin-plate spline algorithm, respectively. For image-based rigid registration, the authors used a block matching approach, and for nonrigid registration, the authors define the moving image deformation using a regular, 3D grid of B-spline control points. The authors measured the target registration error (TRE) as the postregistration misalignment of 60 manually marked, corresponding intrinsic fiducials. The authors also measured the fiducial localization error (FLE), the effect of segmentation variability, and the effect of fiducial distance from the transducer probe tip. Lastly, the authors performed 3D principal component analysis (PCA) on the x, y, and z components of the TREs to examine the 95% confidence ellipsoids describing the errors for each registration method. Using surface-based registration, the authors found mean TREs of 2.13 +/- 0.80 and 2.09 +/- 0.77 mm for rigid and nonrigid techniques, respectively. Using image-based rigid and non-rigid registration, the authors found mean TREs of 1.74 +/- 0.84 and 1.50 +/- 0.83 mm, respectively. Our FLE was 0.21 mm and did not dominate the overall TRE. However, segmentation variability contributed substantially approximately50%) to the TRE of the surface-based techniques. PCA showed that the 95% confidence ellipsoid encompassing fiducial distances between the source and target registra- tion images was reduced from 3.05 to 0.14 cm3, and 0.05 cm3 for the surface-based and image-based techniques, respectively. The run times for both registration methods were comparable at less than 60 s. Our results compare favorably with a clinical need for a TRE of less than 2.5 mm, and suggest that image-based registration is superior to surface-based registration for 3D TRUS-guided prostate biopsies, since it does not require segmentation.

Image Registration Accuracy of a 3-Dimensional Transrectal Ultrasound-Guided Prostate Biopsy System

For a follow-up prostate biopsy procedure, it is useful to know the previous biopsy locations in anatomic relation to the current transrectal ultrasound (TRUS) scan. The goal of this study was to validate the performance of a 3-dimensional TRUS-guided prostate biopsy system that can accurately relocate previous biopsy sites. To correlate biopsy locations from a sequence of visits by a patient, the prostate surface data obtained from a previous visit needs to be registered to the follow-up visits. Two interpolation methods, thin-plate spline (TPS) and elastic warping (EW), were tested for registration of the TRUS prostate image to follow-up scans. We validated our biopsy system using a custom-built phantom. Beads were embedded inside the phantom and were located in each TRUS scan. We recorded the locations of the beads before and after pressures were applied to the phantom and then compared them with computer-estimated positions to measure performance. In our experiments, before system processing, the mean target registration error (TRE) +/- SD was 6.4 +/- 4.5 mm (range, 3-13 mm). After registration and TPS interpolation, the TRE was 5.0 +/- 1.03 mm (range, 2-8 mm). After registration and EW interpolation, the TRE was 2.7 +/- 0.99 mm (range, 1-4 mm). Elastic warping was significantly better than the TPS in most cases (P < .0011). For clinical applications, EW can be implemented on a graphics processing unit with an execution time of less than 2.5 seconds. Elastic warping interpolation yields more accurate results than the TPS for registration of TRUS prostate images. Experimental results indicate potential for clinical application of this method.

Demons deformable registration for CBCT‐guided procedures in the head and neck: Convergence and accuracy

The accuracy and convergence behavior of a variant of the Demons deformable registration algorithm were investigated for use in cone-beam CT (CBCT)-guided procedures of the head and neck. Online use of deformable registration for guidance of therapeutic procedures such as image-guided surgery or radiation therapy places trade-offs on accuracy and computational expense. This work describes a convergence criterion for Demons registration developed to balance these demands; the accuracy of a multiscale Demons implementation using this convergence criterion is quantified in CBCT images of the head and neck. Using an open-source "symmetric" Demons registration algorithm, a convergence criterion based on the change in the deformation field between iterations was developed to advance among multiple levels of a multiscale image pyramid in a manner that optimized accuracy and computation time. The convergence criterion was optimized in cadaver studies involving CBCT images acquired using a surgical C-arm prototype modified for 3D intraoperative imaging. CBCT-to-CBCT registration was performed and accuracy was quantified in terms of the normalized cross-correlation (NCC) and target registration error (TRE). The accuracy and robustness of the algorithm were then tested in clinical CBCT images of ten patients undergoing radiation therapy of the head and neck. The cadaver model allowed optimization of the convergence factor and initial measurements of registration accuracy: Demons registration exhibited TRE=(0.8+/-0.3) mm and NCC =0.99 in the cadaveric head compared to TRE=(2.6+/-1.0) mm and NCC=0.93 with rigid registration. Similarly for the patient data, Demons registration gave mean TRE=(1.6+/-0.9) mm compared to rigid registration TRE=(3.6+/-1.9) mm, suggesting registration accuracy at or near the voxel size of the patient images (1 x 1 x 2 mm3). The multiscale implementation based on optimal convergence criteria completed registration in 52 s for the cadaveric head and in an average time of 270 s for the larger FOV patient images. Appropriate selection of convergence and multiscale parameters in Demons registration was shown to reduce computational expense without sacrificing registration performance. For intraoperative CBCT imaging with deformable registration, the ability to perform accurate registration within the stringent time requirements of the operating environment could offer a useful clinical tool allowing integration of preoperative information while accurately reflecting changes in the patient anatomy. Similarly for CBCT-guided radiation therapy, fast accurate deformable registration could further augment high-precision treatment strategies.

Rapid Dynamic Image Registration of the Beating Heart for Diagnosis and Surgical Navigation

Dynamic cardiac magnetic resonance imaging (MR) and computed tomography (CT) provide cardiologists and cardiac surgeons with high-quality 4-D images for diagnosis and therapy, yet the effective use of these high-quality anatomical models remains a challenge. Ultrasound (US) is a flexible imaging tool, but the US images produced are often difficult to interpret unless they are placed within their proper 3-D anatomical context. The ability to correlate real-time 3-D US volumes (RT3D US) with dynamic MR/CT images would offer a significant contribution to improve the quality of cardiac procedures. In this paper, we present a rapid two-step method for registering RT3D US to high-quality dynamic 3-D MR/CT images of the beating heart. This technique overcomes some major limitations of image registration (such as the correct registration result not necessarily occurring at the maximum of the mutual information (MI) metric) using the MI metric. We demonstrate the effectiveness of our method in a dynamic heart phantom (DHP) study and a human subject study. The achieved mean target registration error of CT+US images in the phantom study is 2.59 mm. Validation using human MR/US volumes shows a target registration error of 1.76 mm. We anticipate that this technique will substantially improve the quality of cardiac diagnosis and therapies.

TH‐C‐304A‐06: Assessment of Registration Accuracy in 3D Transrectal Ultrasound Images of Prostates

Purpose: To evaluate the accuracy of our rigid and non‐rigid registration methods to align three‐dimensional (3D) transrectal ultrasound (TRUS) images of the prostate in order to improve repeat biopsy planning and guidance for prostate cancer diagnosis. Methods and Materials: Three 3D prostate ultrasound images were obtained from each of the 16 patients during 3D prostate ultrasound‐guided biopsy procedures. These 3D images had identifiable calcifications and cysts, which were used as fiducials for the registration validation. For each prostate, images at two different time points (to and t1) during the procedure were selected and the 3D coordinates of the identified fiducials were recorded. A total of 62 fiducials were identified. All images were then segmented manually and semi‐automatically. Images at to were registered to the corresponding images at t1 using rigid and non‐rigid registration. Rigid and non‐rigid registration was performed using boundary‐based iterative closest point and thin‐plate spline algorithms, respectively. We evaluated the registration methods by determining the target registration error (TRE), effect of segmentation variability on registration variability, the TRE dependence on fiducial distance from the transducer tip, and the fiducial localization error (FLE). Results: Tests showed that the FLE was 1.3mm. The mean TRE for manual and semi‐automatic segmentation was 2.2mm and 2.3mm respectively in both rigid and non‐rigid cases. There was little correlation between the TRE and the distance of the fiducials from the ultrasound transducer tip, and segmentation variability resulted in a 0.57mm standard deviation of the TRE. Conclusions: Our results suggest that both rigid and non‐rigid registration techniques for 3D TRUS images during a biopsy procedure are accurate to within 2.3mm, which is less than the 5mm radius of the smallest tumors considered clinically significant. Future tests will evaluate registration accuracy for multimodal cases, specifically magnetic resonance to ultrasound images.

In vivo accuracy assessment of a needle‐based navigation system for CT‐guided radiofrequency ablation of the liver

Computed tomography (CT)-guided percutaneous radiofrequency ablation (RFA) has become a commonly used procedure in the treatment of liver tumors. One of the main challenges related to the method is the exact placement of the instrument within the lesion. To address this issue, a system was developed for computer-assisted needle placement which uses a set of fiducial needles to compensate for organ motion in real time. The purpose of this study was to assess the accuracy of the system in vivo. Two medical experts with experience in CT-guided interventions and two nonexperts used the navigation system to perform 32 needle insertions into contrasted agar nodules injected into the livers of two ventilated swine. Skin-to-target path planning and real-time needle guidance were based on preinterventional 1 mm CT data slices. The lesions were hit in 97% of all trials with a mean user error of 2.4 +/- 2.1 mm, a mean target registration error (TRE) of 2.1 +/- 1.1 mm, and a mean overall targeting error of 3.7 +/- 2.3 mm. The nonexperts achieved significantly better results than the experts with an overall error of 2.8 +/- 1.4 mm (n=16) compared to 4.5 +/- 2.7 mm (n=16). The mean time for performing four needle insertions based on one preinterventional planning CT was 57 +/- 19 min with a mean setup time of 27 min, which includes the steps fiducial insertion (24 +/- 15 min), planning CT acquisition (1 +/- 0 min), and registration (2 +/- 1 min). The mean time for path planning and targeting was 5 +/- 4 and 2 +/- 1 min, respectively. Apart from the fiducial insertion step, experts and nonexperts performed comparably fast. It is concluded that the system allows for accurate needle placement into hepatic tumors based on one planning CT and could thus enable considerable improvement to the clinical treatment standard for RFA procedures and other CT-guided interventions in the liver. To support clinical application of the method, optimization of individual system modules to reduce intervention time is proposed.

Sensor-based neuronavigation: Evaluation of a large continuous patient population

Objective Navigation systems enable neurosurgeons to guide operations with imaging data. Sensor-based neuronavigation uses an electromagnetic field and sensors to measure the positions of the patient's brain anatomy and the surgical instruments. The aim of this investigation was to determine the accuracy level of sensor-based tracking in a large patient collection. Methods This study covers 250 patients operated upon during a continuous 5.5-year period. The patients had a wide range of indications and surgical procedures. The operations were performed with a direct current (DC) pulsed sensor-based electromagnetic navigation system. Four kinds of errors were measured: the fiducial registration error (FRE), the target registration error (TRE), brain shift, and the position error (PE). These errors were calculated for five subgroups of indications: target determination and trajectory guidance, functional navigation, skull base and neurocranium, determination of resection volume, and transnasal and transsphenoidal access. Results The overall mean FRE was 1.66 mm (±0.61 mm). The overall mean TREs were 1.33 mm (±0.51 mm) centroid and 1.59 mm (±0.57 mm) lesional. The overall mean brain shift for applicable cases was 1.61 mm (±1.14 mm). The overall mean PE was 0.92 mm (±0.54 mm). Conclusions By and large, modern sensor-based neuronavigation operates within an acceptable and commonplace degree of error. However, the neurosurgeon must remain critical in cases of small lesions, and must exert caution not to introduce further interference from metal objects or electromagnetic devices.

A computationally efficient method for automatic registration of orthogonal x-ray images with volumetric CT data

The paper presents a computationally efficient 3D–2D image registration algorithm for automatic pre-treatment validation in radiotherapy. The novel aspects of the algorithm include (a) a hybrid cost function based on partial digitally reconstructed radiographs (DRRs) generated along projected anatomical contours and a level set term for similarity measurement; and (b) a fast search method based on parabola fitting and sensitivity-based search order. Using CT and orthogonal x-ray images from a skull and a pelvis phantom, the proposed algorithm is compared with the conventional ray-casting full DRR based registration method. Not only is the algorithm shown to be computationally more efficient with registration time being reduced by a factor of 8, but also the algorithm is shown to offer 50% higher capture range allowing the initial patient displacement up to 15 mm (measured by mean target registration error). For the simulated data, high registration accuracy with average errors of 0.53 mm ± 0.12 mm for translation and 0.61° ± 0.29° for rotation within the capture range has been achieved. For the tested phantom data, the algorithm has also shown to be robust without being affected by artificial markers in the image.

Image-guided craniotomy for frontal sinus preservation during meningioma surgery

Objective Preservation of the frontal sinus (FS) during the frontolateral approach to the skull base reduces morbidity, enhances patient comfort, and speeds up the surgical procedure. Due to its irregular outline, mental reconstruction of the borders of FS from two-dimensional images is challenging during surgery. This study was designed to evaluate the impact of neuronavigation on identification and preservation of the FS during frontolateral craniotomies. Methods Forty-five patients with pathologies located in the anterior skull base and in the parasellar region were included. A standard computed tomography (CT) sequence was obtained from each patient and uploaded onto an image-guidance system for volumetric rendering of 3D images. The outline of the FS was visualized and the distance between its lateral border and the mid-pupillary line (MPL) was measured. The results were used for navigated craniotomies and compared to the intra-operative findings. Results The FS was located medial, on and lateral to the MPL in 32, 4 and 9 cases, respectively. The individual outline of the FS could be identified with a mean target registration error of 1.4 mm (±0.7 mm). The craniotomy could be custom-tailored for each patient according to the individualized landmarks while visualizing the lesion and the surgical landmarks simultaneously. Unintended opening of the frontal sinus or orbit did not occur in any of these cases. Conclusion Image-guided craniotomies based on 3D volumetric image rendering allow for fast and reliable demarcation of complex anatomical structures hidden from direct view in frontolateral approaches. The outline of the frontal sinus and the orbit can be appraised at a glance providing additional safety and precision during craniotomy.

Comparison of 4 registration strategies for computer-aided maxillofacial surgery

Surgeons have recently started to use computer-aided surgery (CAS) to assist with maxillofacial reconstructive surgery. This study evaluates four different CAS registration strategies in the maxillofacial skeleton. Fifteen fiducial markers were placed on each of four cadaveric heads. Four registration protocols were used: 1) group 1-invasive markers, 2) group 2-skin surface, 3) group 3-bony landmark, 4) group 4-intraoral splint. Two observers registered each head twice with each of the four protocols and measured the target registration error (TRE). The process was repeated on two different navigation systems for confirmation. The mean TRE values were: invasive, 1.13 +/- 0.05 mm (P < 0.05); skin, 2.03 +/- 0.07 mm (P < 0.05); bone, 3.17 +/- 0.10 mm (P < 0.05); and splint, 3.79 +/- 0.13 mm (P < 0.05). The TRE values were consistent across CAS systems. Of the techniques tested for CAS registration, invasive fiducial markers are the most accurate. Skin surface landmarks, bony landmarks, and an intraoral splint are incrementally less accurate.

Integrating Diagnostic B-Mode Ultrasonography Into CT-Based Radiation Treatment Planning

This paper presents methods and a clinical procedure for integrating B-mode ultrasound images tagged with position information with a planning computed tomography (CT) scan for radiotherapy. A workflow is described that allows the integration of these modalities into the clinic. A surface mapping approach provides a preregistration of the ultrasound image borders onto the patient's skin. Successively, a set of individual ultrasound images from a freehand sweep is chosen by the physician. These images are automatically registered with the planning CT scan using novel intensity-based methods. We put a particular focus on deriving an appropriate similarity measure based on the physical properties and artifacts of ultrasound. A combination of a weighted mutual information term, edge correlation, clamping to the skin surface, and occlusion detection is able to assess the alignment of structures in ultrasound images and information reconstructed from the CT data. We demonstrate the practicality of our methods on five patients with head and neck tumors and cervical lymph node metastases and provide a detailed report on the conducted experiments, including the setup, calibration, acquisition, and verification of our algorithms. The mean target registration error on nine data sets is 3.9 mm. Thus, the additional information about intranodal architecture and fulfillment of malignancy criteria derived from a high-resolution ultrasonography of lymph nodes can be localized and visualized in the CT scan coordinate space and is made available for further radiation treatment planning.

Application accuracy in frameless image-guided neurosurgery: a comparison study of three patient-to-image registration methods

The aim of this study was to compare three patient-to-image registration methods in frameless stereotaxy in terms of their application accuracy (the accuracy with which the position of a target can be determined intraoperatively). In frameless stereotaxy, imaging information is transposed to the surgical field to show the spatial position of a localizer or surgical instrument. The mathematical relationship between the image volume and the surgical working space is calculated using a rigid body transformation algorithm, based on point-pair matching or surface matching. Fifty patients who were scheduled to undergo a frameless image-guided neurosurgical procedure were included in the study. Prior to surgery, the patients underwent either computerized tomography (CT) scanning or magnetic resonance (MR) imaging with widely distributed adhesive fiducial markers on the scalp. An extra fiducial marker was placed on the head as a target, as near as possible to the intracranial lesion. Prior to each surgical procedure, an optical tracking system was used to perform three separate patient-to-image registration procedures, using anatomical landmarks, adhesive markers, or surface matching. Subsequent to each registration, the target registration error (TRE), defined as the Euclidean distance between the image space coordinates and world space coordinates of the target marker, was determined. Independent of target location or imaging modality, mean application accuracy (+/- standard deviation) was 2.49 +/- 1.07 mm when using adhesive markers. Using the other two registration strategies, mean TREs were significantly larger (surface matching, 5.03 +/- 2.30 mm; anatomical landmarks, 4.97 +/- 2.29 mm; p < 0.001 for both). The results of this study show that skin adhesive fiducial marker registration is the most accurate noninvasive registration method. When images from an earlier study are to be used and accuracy may be slightly compromised, anatomical landmarks and surface matching are equally accurate alternatives.

MO‐D‐ValB‐03: Genetic Evolutionary Taboo Search: A Novel Approach for Optimal Marker Placement in Infrared Patient Positioning

Purpose: To develop the methods of a novel approach for optimal marker placement in infrared patient positioning. Method and Materials: A nondeterministic optimization technique (Genetic Evolutionary Taboo Search, GETS), combining genetic algorithms and taboo search, was implemented. A population‐based evolution is generated, where adaptive memory features guide the evolutionary process to thoroughly explore the solution space. Preset taboo solutions are introduced to reject marker configurations resulting collinear from the point of view of infrared cameras. The GETS algorithm was tested on 10 prostate patients: treatment planning CT scans were segmented to provide 3‐D representation of PTV (prostate + seminal vesicles), OARs (bladder and rectum) and skin surface model. Segmented data were fed to the GETS algorithm to obtain optimized configurations of markers, minimizing the target registration error (TRE), to be compared to a random configuration. The changes in the optimal marker configuration when OARs are included within the target were also investigated. Results: The GETS algorithm yielded a significant improvement in TRE values: optimal configurations ensured a 26.5% mean TRE decrease. Common features in the optimal marker configurations were found for the 10 patients group, being optimized solutions symmetrically distributed, with markers mostly placed on lateral sides. Optimal marker configurations when OARs were included within the target resulted in a similar spatial distribution, if compared to the PTV‐only condition. The implemented memory‐based design resulted in improved gene expression over the evolution process, with respect to memoryless genetic algorithms. Conclusion: The GETS algorithm revealed high performance in solving the optimal marker placement problem, leading to improved marker configuration for stereophotogrammetric patient positioning in radiotherapy. Memory features ensured enhanced capabilities in exploring the solution space, if compared to conventional genetic optimization. The application of the new algorithm to a 10 patient group provided practical indications toward better marker placement for prostate cases.

3-D/2-D registration by integrating 2-D information in 3-D

In image-guided therapy, high-quality preoperative images serve for planning and simulation, and intraoperatively as "background", onto which models of surgical instruments or radiation beams are projected. The link between a preoperative image and intraoperative physical space of the patient is established by image-to-patient registration. In this paper, we present a novel 3-D/2-D registration method. First, a 3-D image is reconstructed from a few 2-D X-ray images and next, the preoperative 3-D image is brought into the best possible spatial correspondence with the reconstructed image by optimizing a similarity measure (SM). Because the quality of the reconstructed image is generally low, we introduce a novel SM, which is able to cope with low image quality as well as with different imaging modalities. The novel 3-D/2-D registration method has been evaluated and compared to the gradient-based method (GBM) using standardized evaluation methodology and publicly available 3-D computed tomography (CT), 3-D rotational X-ray (3DRX), and magnetic resonance (MR) and 2-D X-ray images of two spine phantoms, for which gold standard registrations were known. For each of the 3DRX, CT, or MR images and each set of X-ray images, 1600 registrations were performed from starting positions, defined as the mean target registration error (mTRE), randomly generated and uniformly distributed in the interval of 0-20 mm around the gold standard. The capture range was defined as the distance from gold standard for which the final TRE was less than 2 mm in at least 95% of all cases. In terms of success rate, as the function of initial misalignment and capture range the proposed method outperformed the GBM. TREs of the novel method and the GBM were approximately the same. For the registration of 3DRX and CT images to X-ray images as few as 2-3 X-ray views were sufficient to obtain approximately 0.4 mm TREs, 7-9 mm capture range, and 80%-90% of successful registrations. To obtain similar results for MR to X-ray registrations, an image, reconstructed from at least 11 X-ray images was required. Reconstructions from more than 11 images had no effect on the registration results.

Feasibility and validation of registration of three-dimensional left atrial models derived from computed tomography with a noncontact cardiac mapping system

The purpose of this study was to determine the feasibility and assess the validity of registering three-dimensional (3D) models from computed tomographic (CT) images using a cardiac mapping system. Registration of 3D anatomic models with an interventional system could help identify and navigate mapping and ablation catheters over a complex structure such as the left atrium (LA). ECG-gated, contrast-enhanced cardiac CT imaging was performed in 14 patients with atrial fibrillation. Segmentation was used to create 3D models of the LA. The 3D models were registered with the mapping system using a series of fiducial points. Registration was accomplished retrospectively in the first 10 patients, and catheter navigation was visualized from recorded data. In the final four patients, registration was accomplished in real time during electrophysiologic study. The mapping catheter position, as it was navigated inside the LA, was applied to the registered model in real time. For the validation study, temporary pacing leads were implanted in the LA of 10 dogs. Following this, CT scanning, segmentation, LA model importation, and registration was described previously. After registration, a mapping catheter was positioned at the site of each buried lead according to the registered model with no fluoroscopic guidance. A radiofrequency lesion was created at this location, and the dog was sacrificed, the heart removed and stained, and the distance between the buried lead and the lesion measured. During the feasibility study, the location of the catheter in the registered model correlated with fluoroscopy, angiography, and intracardiac electrograms. LA endocardial potentials during sinus rhythm and any premature atrial contractions also were successfully delineated over the registered models. In the validation study, the mean target registration error was 2.0 +/- 3.6 mm. Registration of CT-derived 3D models of the LA using a cardiac mapping system is feasible and accurate.

Dynamic, three‐dimensional optical tracking of an ablative laser beam

Surgical resection remains the treatment of choice for brain tumors with infiltrating margins but is currently limited by visual discrimination between normal and neoplastic marginal tissues during surgery. Imaging modalities such as computed tomography, magnetic resonance, positron emission tomography, and optical techniques can accurately localize tumor margins. We believe coupling the fine resolution of current imaging techniques with the precise cutting of midinfrared lasers through image-guided neurosurgery can greatly enhance tumor margin resection. This paper describes a feasibility study designed to optically track in three-dimensional space the articulated arm delivery of a noncontact ablative laser beam. To enable optical tracking of the laser beam focus, infrared-emitting diodes (IREDs) were attached to a handpiece machined for the distal end of the articulated arm of a surgical carbon dioxide laser. Crosstalk between the ablative laser beam and the tracking diodes was measured. The geometry of the adapted laser handpiece was characterized to track an externally attached passive tip and the laser beam focus. Target localization accuracies were assessed for both instrument points-of-interest and the sources of tracking errors were investigated. Stray infrared laser light did not affect optical tracking accuracy. The mean target registration errors while optically tracking the laser handpiece with a passive tip and the laser beam focus were 1.31+/-0.50 mm and 2.31+/-0.92 mm, respectively, and were equivalent to the errors tracking a 24-IRED pen probe from Northern Digital in a side-by-side comparison. The majority of error during ablation tracking derived from registration accuracy between physical space and the defined space of the ablation phantom and from an inability to freehand align the laser focus with the target in a consistent manner. While their magnitudes depend on spatial details of the tracking setup (e.g., number and distribution of fiducial points, working distance from the camera, etc.), these errors are inherent to any freehand laser surgery.