Residual Registration Error Research Articles

An automated methodology for whole-body, multimodality tracking of individual cancer lesions.

Objective. Manual analysis of individual cancer lesions to assess disease response is clinically impractical and requires automated lesion tracking methodologies. However, no methodology has been developed for whole-body individual lesion tracking, across an arbitrary number of scans, and acquired with various imaging modalities.Approach. This study introduces a lesion tracking methodology and benchmarked it using 2368Ga-DOTATATE PET/CT and PET/MR images of eight neuroendocrine tumor patients. The methodology consists of six steps: (1) alignment of multiple scans via image registration, (2) body-part labeling, (3) automatic lesion-wise dilation, (4) clustering of lesions based on local lesion shape metrics, (5) assignment of lesion tracks, and (6) output of a lesion graph. Registration performance was evaluated via landmark distance, lesion matching accuracy was evaluated between each image pair, and lesion tracking accuracy was evaluated via identical track ratio. Sensitivity studies were performed to evaluate the impact of lesion dilation (fixed versus automatic dilation), anatomic location, image modalities (inter- versus intra-modality), registration mode (direct versus indirect registration), and track size (number of time-points and lesions) on lesion matching and tracking performance.Main results. Manual contouring yielded 956 lesions, 1570 lesion-matching decisions, and 493 lesion tracks. The median residual registration error was 2.5 mm. The automatic lesion dilation led to 0.90 overall lesion matching accuracy, and an 88% identical track ratio. The methodology is robust regarding anatomic locations, image modalities, and registration modes. The number of scans had a moderate negative impact on the identical track ratio (94% for 2 scans, 91% for 3 scans, and 81% for 4 scans). The number of lesions substantially impacted the identical track ratio (93% for 2 nodes versus 54% for ≥5 nodes).Significance. The developed methodology resulted in high lesion-matching accuracy and enables automated lesion tracking in PET/CT and PET/MR.

Optical Rules to Mitigate the Parallax-Related Registration Error in See-Through Head-Mounted Displays for the Guidance of Manual Tasks

Head-mounted displays (HMDs) are hands-free devices particularly useful for guiding near-field tasks such as manual surgical procedures. See-through HMDs do not significantly alter the user’s direct view of the world, but the optical merging of real and virtual information can hinder their coherent and simultaneous perception. In particular, the coherence between the real and virtual content is affected by a viewpoint parallax-related misalignment, which is due to the inaccessibility of the user-perceived reality through the semi-transparent optical combiner of the OST Optical See-Through (OST) display. Recent works demonstrated that a proper selection of the collimation optics of the HMD significantly mitigates the parallax-related registration error without the need for any eye-tracking cameras and/or for any error-prone alignment-based display calibration procedures. These solutions are either based on HMDs that projects the virtual imaging plane directly at arm’s distance, or they require the integration on the HMD of additional lenses to optically move the image of the observed scene to the virtual projection plane of the HMD. This paper describes and evaluates the pros and cons of both the suggested solutions by providing an analytical estimation of the residual registration error achieved with both solutions and discussing the perceptual issues generated by the simultaneous focalization of real and virtual information.

Residual image registration error by fiducial markers in accelerated partial breast irradiation using C-arm linac: a phantom study.

External beam accelerated partial breast irradiation (APBI) is an alternative treatment for patients with early-stage breast cancer. The efficacy of image-guided radiotherapy (IGRT) using fiducial markers, such as gold markers or surgical clips, has been demonstrated. However, the effects of respiratory motion during a single fraction have not been reported. This study aimed to evaluate the residual image registration error of fiducial marker-based IGRT by respiratory motion and propose a suitable treatment strategy. We developed an acrylic phantom embedded with surgical clips to verify the registration error under moving conditions. The frequency of the phase difference in the respiratory cycle due to sequential acquisition was verified in a preliminary study. Fiducial marker-based IGRT was then performed in ten scenarios. The residual registration error (RRE) was calculated on the basis of the differences in the coordinates of clips between the true position if not moved and the last position. The frequencies of the phase differences in 0.0-0.99, 1.0-1.99, 2.0-2.99, 3.0-3.99, and 4.0-5.0mm were 23%, 24%, 22%, 20%, and 11%, respectively. When assuming a clinical case, the mean RREs for all directions were within 1.0mm, even if respiratory motion of 5mm existed in two axes. For APBI with fiducial marker-based IGRT, the introduction of an image registration strategy that employs stepwise couch correction using at least three orthogonal images should be considered.

The impact of visualization paradigms on the detectability of spatial misalignment in mixed reality surgical guidance.

Mixed reality (MR) for image-guided surgery may enable unobtrusive solutions for precision surgery. To display preoperative treatment plans at the correct physical position, it is essential to spatially align it with the patient intra-operatively. Accurate alignment is safety critical because it will guide treatment, but cannot always be achieved for varied reasons. Effective visualization mechanisms that reveal misalignment are crucial to prevent adverse surgical outcomes to ensure safe execution. We test the effectiveness of three MR visualization paradigms in revealing spatial misalignment: wireframe, silhouette, and heatmap, which encodes residual registration error. We conduct a user study among 12 participants and use an anthropomorphic phantom mimicking total shoulder arthroplasty. Participants wearing Microsoft HoloLens 2 are presented with 36 randomly ordered spatial (mis)alignments of a virtual glenoid model overlaid on the phantom, each rendered using one of the three methods. Users choose whether to accept or reject the spatial alignment at every trial. Upon completion, participants report their perceived difficulty while using the visualization paradigms. Across all visualization paradigms, the ability of participants to reliably judge the accuracy of spatial alignment was moderate (58.33%).The three visualization paradigms showed comparable performance. However, the heatmap-based visualization resulted in significantly better detectability than random chance ([Formula: see text]). Despite heatmap enabling the most accurate decisions according to our measurements, wireframe was the most liked paradigm (50%), followed by silhouette (41.7%) and heatmap (8.3%). Our findings suggest that conventional mixed reality visualization paradigms are not sufficiently effective in enabling users to differentiate between accurate and inaccurate spatial alignment of virtual content to the environment.

Robust, fiducial-free drift correction for super-resolution imaging

We describe a robust, fiducial-free method of drift correction for use in single molecule localization-based super-resolution methods. The method combines periodic 3D registration of the sample using brightfield images with a fast post-processing algorithm that corrects residual registration errors and drift between registration events. The method is robust to low numbers of collected localizations, requires no specialized hardware, and provides stability and drift correction for an indefinite time period.

Residual setup errors in cranial stereotactic radiosurgery without six degree of freedom robotic couch: Frameless versus rigid immobilization systems.

Purpose and ObjectivesThis IRB‐approved study was to compare the residual inter‐fractional setup errors and intra‐fractional motion of patients treated with cranial stereotactic radiosurgery without a 6 degree of freedom (DoF) couch. We evaluated both frameless non‐invasive vacuum‐suction immobilization (Aktina PinPoint) and TALON rigid screw immobilization.Materials and MethodsTwenty consecutive patients treated by Varian TrueBeam STX or Tomotherapy were selected for data collection. The dose and number of fractions received by each patient ranged from 18 Gy in 1 fraction (SRS) to 25 Gy in 5 fractions (SRT). Twelve patients were immobilized using PinPoint, a frameless suction system (Aktina Medical, New York) and eight patients were immobilized using the TALON rigid screw system. Customized head cushions were used for all patients. Six Atkina patients received pre‐ and post‐treatment cone‐beam CT (CBCT) to evaluate the intra‐fractional motion of the Aktina system. The intra‐fractional motion with the TALON rigid screw system has been reported to be negligible and was not repeated in this study. All patients received pre‐treatment CBCT or megavoltage CT (MVCT) to assess inter‐fractional setup accuracy. Shifts to the final treatment position were determined based on matching bony anatomy in the pre‐treatment setup CT and the planning CT. Setup CT and planning CT were registered retrospectively based on bony anatomy using image registration software to quantify rotational and translational errors.ResultsFor the frameless Aktina system, mean and standard deviation of the intra‐fractional motion were −0.5 ± 0.7 mm (lateral), 0.1 ± 0.9 mm (vertical), −0.5 ± 0.6 mm (longitudinal), −0.04 ± 0.18°(pitch), −0.1 ± 0.23°(yaw), and −0.03 ± 0.17°(roll) indicating negligible intra‐fractional motion. Inter‐fractional rotation errors were −0.10 ± 0.25° (pitch), −0.08 ± 0.16° (yaw), and −0.20 ± 0.41° (roll) for TALON rigid screw immobilization versus 0.20 ± 0.69° (pitch), 0.34 ± 0.56° (yaw), 0.35 ± 0.82° (roll) for frameless vacuum‐suction immobilization showing that the rigid immobilization setup is more reproducible than the frameless immobilization. Without rotational correction by a 6 DoF couch, residual registration error exists and increases with distance from the image fusion center. In a 3D vector space, a tumor located 5 cm from the center of image fusion would require a 0.9 mm margin with the TALON system and a 2.1 mm margin with Aktina.ConclusionsWith image‐guided radiotherapy, translational setup errors can be corrected by image registration between pre‐treatment setup CT and planning CT. However, rotational errors cannot be accounted for without a 6 DoF couch. Our study showed that the frameless Aktina immobilization system provided negligible intra‐fractional motion. The inter‐fractional rotation setup error using Aktina was larger than rigid immobilization with the TALON system. To treat a single lesion far from the center of image registration or for multiple lesions in a single plan, additional margin may be needed to account for the uncorrectable rotational setup errors.

A Robust, Fiducial Free Drift Correction for Superresolution Imaging

We present a robust, fiducial-free method of drift correction for use in single molecule localization-based super-resolution methods. The method combines periodic 3D registration of the sample using brightfield images with a fast post-processing algorithm that corrects residual registration errors and drift between registration events. The method is robust to low numbers of collected localizations, requires no specialized hardware, and provides stability and drift correction for an indefinite time period. For brightfield registration, we establish a protocol in which super-resolution imaging is performed for a sequence of movie frames, which we term a dataset. A z-stack of brightfield reference images is obtained initially. After each dataset is collected, a new brightfield z-stack is obtained and compared with the previous reference z-stack, finding the best z and (x, y) fits via a scaled cross-correlation, allowing the reference z-stack and the current z-stack to be aligned, and subsequently the following dataset. The procedure then repeats with the next dataset. Once all the datasets are collected, a post-drift correction procedure is applied to the data. This procedure divides the problem into intra-dataset and inter-dataset drift correction, both minimizing a cost function that is simply the thresholded sum of the nearest neighbor distances for all the predicted emitter positions in a dataset, either with respect to itself for intra-dataset drift correction, or the first dataset for inter-dataset drift correction. Intra-dataset drift correction is performed first, with the results saved for the next phase. For inter-dataset drift correction, the datasets are drift corrected in sequence against the first dataset. We demonstrate our methodology on experimental 2D and simulated 3D data.

MOSAICING FIDELITY GEOMETRICAL ASSESSMENT BASED ON SURF POINT CLASSIFICATION

Abstract. The recent development of automated UAV imaging applications for geomatics is leading to an unprecedented rapid growth in mosaic and DEM images. Newer, more advanced algorithms are being extensively studied to fulfill constantly increasing requirements. Sequences of georeferenced images that cannot be shot at once are merged with ideally no geometrical distortion to allow for 3D reconstruction and orthophotography generation. If newer robust mosaicing algorithms are being developed to withstand UAV tighter real-time constraints, very little attention has been given to objectively assess their capacity to coherently combine different image geometries. Changes in FOV and illumination, oblique scenes and dynamic sequences featuring movement and object occlusion are not yet fully handled by state-of-the-art algorithms. Despite the large panel of existing applications featuring different quality constraints, our work focuses on the geomatics context, where the main requirement is the respect of geometrical proportions among images. Geometrically coherent mosaics and 3D reconstructions can only result from fully static, high-altitude orthogonal views. GCPs or well known geographical landmarks are employed with RTK-GPS for scene georeferencing. The overall fidelity is computed as the global variance of the distances between SGPs. However, very few ways seem to exist to measure the content coherence of large geographical scenes, interiors or unreferenced scenes. Distances are mostly expressed in a local coordinate system and GCPs are often unavailable. The main contribution of this paper consists in the definition of a formal metric for measuring local coordinate geometrical fidelity of mosaics issued of UAV image sequences, in real operating conditions. A Mosaicing Fidelity Assessment (MSF) index is therefore computed for every couple of consecutive images of a sequence. The metric generates an index based on the distances of SURF feature points extracted in the images and compares them to estimate geometrical changes transferred to the mosaic. The solution can detect perspective inaccuracies caused by residual registration errors. It can fulfill the requirements of most of imaging and geomatics applications and can be executed as nearly real-time due to its low computational complexity.

Visualization of Brain Shift Corrected Functional Magnetic Resonance Imaging Data for Intraoperative Brain Mapping

BackgroundBrain tumor surgery requires careful balance between maximizing tumor excision and preserving eloquent cortex. In some cases, the surgeon may opt to perform an awake craniotomy including intraoperative mapping of brain function by direct cortical stimulation (DCS) to assist in surgical decision-making. Preoperatively, functional magnetic resonance imaging (fMRI) facilitates planning by identification of eloquent brain areas, helping to guide DCS and other aspects of the surgical plan. However, brain deformation (shift) limits the usefulness of preoperative fMRI during surgery. To address this, an integrated visualization method for fMRI and DCS results is developed that is intuitive for the surgeon.MethodsAn image registration pipeline was constructed to display preoperative fMRI data corrected for brain shift overlaid on images of the exposed cortical surface at the beginning and completion of DCS mapping. Preoperative fMRI and DCS data were registered for a range of misalignments, and the residual registration errors were calculated. The pipeline was validated on imaging data from five brain tumor patients who underwent awake craniotomy.ResultsRegistration errors were well under 5 mm (the approximate spatial resolution of DCS) for misalignments of up to 25 mm and approximately 10–15°. For rotational misalignments up to 20°, the success rate was 95% for an error tolerance of 5 mm. Failures were negligible for rotational misalignments up to 10°. Good quality registrations were observed for all five patients.ConclusionsA proof-of-concept image registration pipeline is presented with acceptable accuracy for intraoperative use, providing multimodality visualization with potential benefits for intraoperative brain mapping.

Biomechanical deformable image registration of longitudinal lung CT images using vessel information

Spatial correlation of lung tissue across longitudinal images, as the patient responds to treatment, is a critical step in adaptive radiotherapy. The goal of this work is to expand a biomechanical model-based deformable registration algorithm (Morfeus) to achieve accurate registration in the presence of significant anatomical changes. Six lung cancer patients previously treated with conventionally fractionated radiotherapy were retrospectively evaluated. Exhale CT scans were obtained at treatment planning and following three weeks of treatment. For each patient, the planning CT was registered to the follow-up CT using Morfeus, a biomechanical model-based deformable registration algorithm. To model the complex response of the lung, an extension to Morfeus has been developed: an initial deformation was estimated with Morfeus consisting of boundary conditions on the chest wall and incorporating a sliding interface with the lungs. It was hypothesized that the addition of boundary conditions based on vessel tree matching would provide a robust reduction of the residual registration error. To achieve this, the vessel trees were segmented on the two images by thresholding a vesselness image based on the Hessian matrix’s eigenvalues. For each point on the reference vessel tree centerline, the displacement vector was estimated by applying a variant of the Demons registration algorithm between the planning CT and the deformed follow-up CT. An expert independently identified corresponding landmarks well distributed in the lung to compute target registration errors (TRE). The TRE was: , and mm after rigid registration, Morfeus and Morfeus with boundary conditions on the vessel tree, respectively. In conclusion, the addition of boundary conditions on the vessels significantly improved the accuracy in modeling the response of the lung and tumor over the course of radiotherapy. Minimizing and modeling these geometrical uncertainties will enable future plan adaptation strategies.

Automatic Segmentation of Wrist Bones in CT Using a Statistical Wrist Shape + Pose Model.

Segmentation of the wrist bones in CT images has been frequently used in different clinical applications including arthritis evaluation, bone age assessment and image-guided interventions. The major challenges include non-uniformity and spongy textures of the bone tissue as well as narrow inter-bone spaces. In this work, we propose an automatic wrist bone segmentation technique for CT images based on a statistical model that captures the shape and pose variations of the wrist joint across 60 example wrists at nine different wrist positions. To establish the correspondences across the training shapes at neutral positions, the wrist bone surfaces are jointly aligned using a group-wise registration framework based on a Gaussian Mixture Model. Principal component analysis is then used to determine the major modes of shape variations. The variations in poses not only across the population but also across different wrist positions are incorporated in two pose models. An intra-subject pose model is developed by utilizing the similarity transforms at all wrist positions across the population. Further, an inter-subject pose model is used to model the pose variations across different wrist positions. For segmentation of the wrist bones in CT images, the developed model is registered to the edge point cloud extracted from the CT volume through an expectation maximization based probabilistic approach. Residual registration errors are corrected by application of a non-rigid registration technique. We validate the proposed segmentation method by registering the wrist model to a total of 66 unseen CT volumes of average voxel size of 0.38 mm. We report a mean surface distance error of 0.33 mm and a mean Jaccard index of 0.86.

Landmark constellation models for medical image content identification and localization.

Many medical imaging tasks require the detection and localization of anatomical landmarks, for example for the initialization of model-based segmentation or to detect anatomical regions present in an image. A large number of landmark and object localization methods have been described in the literature. The detection of single landmarks may be insufficient to achieve robust localization across a variety of imaging settings and subjects. Furthermore, methods like the generalized Hough transform yield the most likely location of an object, but not an indication whether or not the landmark was actually present in the image. For these reasons, we developed a simple and computationally efficient method combining localization results from multiple landmarks to achieve robust localization and to compute a localization confidence measure. For each anatomical region, we train a constellation model indicating the mean relative locations and location variability of a set of landmarks. This model is registered to the landmarks detected in a test image via point-based registration, using closed-form solutions. Three different outlier suppression schemes are compared, two using iterative re-weighting based on the residual landmark registration errors and the third being a variant of RANSAC. The mean weighted residual registration error serves as a confidence measure to distinguish true from false localization results. The method is optimized and evaluated on synthetic data, evaluating both the localization accuracy and the ability to classify good from bad registration results based on the residual registration error. Two application examples are presented: the identification of the imaged anatomical region in trauma CT scans and the initialization of model-based segmentation for C-arm CT scans with different target regions. The identification of the target region with the presented method was in 96% of the cases correct. The presented method is a simple solution for combining multiple landmark localization results. With appropriate parameters, outlier suppression clearly improves the localization performance over model registration without outlier suppression. The optimum choice of method and parameters depends on the expected level of noise and outliers in the application at hand, as well as on the focus on localization, classification, or both. The method allows detecting and localizing anatomical fields of view in medical images and is well suited to support a wide range of applications comprising image content identification, anatomical navigation and visualization, or initializing the pose of organ shape models.

An efficient total variation algorithm for super-resolution in fetal brain MRI with adaptive regularization

Although fetal anatomy can be adequately viewed in new multi-slice MR images, many critical limitations remain for quantitative data analysis. To this end, several research groups have recently developed advanced image processing methods, often denoted by super-resolution (SR) techniques, to reconstruct from a set of clinical low-resolution (LR) images, a high-resolution (HR) motion-free volume. It is usually modeled as an inverse problem where the regularization term plays a central role in the reconstruction quality. Literature has been quite attracted by Total Variation energies because of their ability in edge preserving but only standard explicit steepest gradient techniques have been applied for optimization. In a preliminary work, it has been shown that novel fast convex optimization techniques could be successfully applied to design an efficient Total Variation optimization algorithm for the super-resolution problem. In this work, two major contributions are presented. Firstly, we will briefly review the Bayesian and Variational dual formulations of current state-of-the-art methods dedicated to fetal MRI reconstruction. Secondly, we present an extensive quantitative evaluation of our SR algorithm previously introduced on both simulated fetal and real clinical data (with both normal and pathological subjects). Specifically, we study the robustness of regularization terms in front of residual registration errors and we also present a novel strategy for automatically select the weight of the regularization as regards the data fidelity term. Our results show that our TV implementation is highly robust in front of motion artifacts and that it offers the best trade-off between speed and accuracy for fetal MRI recovery as in comparison with state-of-the art methods.

SU‐E‐J‐29: Automatic Image Registration Performance of Three IGRT Systems for Prostate Radiotherapy

Purpose:To compare the performance of an automatic image registration algorithm on image sets collected on three commercial image guidance systems, and explore its relationship with imaging parameters such as dose and sharpness.Methods:Images of a CIRS Virtually Human Male Pelvis phantom (VHMP) were collected on the CBCT systems of Varian TrueBeam/OBI and Elekta Synergy/XVI linear accelerators, across a range of mAs settings; and MVCT on a Tomotherapy Hi‐ART accelerator with a range of pitch. Using the 6D correlation ratio algorithm of XVI, each image was registered to a mask of the prostate volume with a 5 mm expansion. Registrations were repeated 100 times, with random initial offsets introduced to simulate daily matching. Residual registration errors were calculated by correcting for the initial phantom set‐up error. Automatic registration was also repeated after reconstructing images with different sharpness filters.Results:All three systems showed good registration performance, with residual translations <0.5mm (1σ) for typical clinical dose and reconstruction settings. Residual rotational error had larger range, with 0.8°, 1.2° and 1.9° for 1σ in XVI, OBI and Tomotherapy respectively. The registration accuracy of XVI images showed a strong dependence on imaging dose, particularly below 4mGy. No evidence of reduced performance was observed at the lowest dose settings for OBI and Tomotherapy, but these were above 4mGy. Registration failures (maximum target registration error > 3.6 mm on the surface of a 30mm sphere) occurred in 5% to 10% of registrations. Changing the sharpness of image reconstruction had no significant effect on registration performance.Conclusions:Using the present automatic image registration algorithm, all IGRT systems tested provided satisfactory registrations for clinical use, within a normal range of acquisition settings.

A system to use electromagnetic tracking for the quality assurance of brachytherapy catheter digitization.

To investigate the use of a system using electromagnetic tracking (EMT), post-processing and an error-detection algorithm for detecting errors and resolving uncertainties in high-dose-rate brachytherapy catheter digitization for treatment planning. EMT was used to localize 15 catheters inserted into a phantom using a stepwise acquisition technique. Five distinct acquisition experiments were performed. Noise associated with the acquisition was calculated. The dwell location configuration was extracted from the EMT data. A CT scan of the phantom was performed, and five distinct catheter digitization sessions were performed. No a priori registration of the CT scan coordinate system with the EMT coordinate system was performed. CT-based digitization was automatically extracted from the brachytherapy plan DICOM files (CT), and rigid registration was performed between EMT and CT dwell positions. EMT registration error was characterized in terms of the mean and maximum distance between corresponding EMT and CT dwell positions per catheter. An algorithm for error detection and identification was presented. Three types of errors were systematically simulated: swap of two catheter numbers, partial swap of catheter number identification for parts of the catheters (mix), and catheter-tip shift. Error-detection sensitivity (number of simulated scenarios correctly identified as containing an error/number of simulated scenarios containing an error) and specificity (number of scenarios correctly identified as not containing errors/number of correct scenarios) were calculated. Catheter identification sensitivity (number of catheters correctly identified as erroneous across all scenarios/number of erroneous catheters across all scenarios) and specificity (number of catheters correctly identified as correct across all scenarios/number of correct catheters across all scenarios) were calculated. The mean detected and identified shift was calculated. The maximum noise ±1 standard deviation associated with the EMT acquisitions was 1.0 ± 0.1 mm, and the mean noise was 0.6 ± 0.1 mm. Registration of all the EMT and CT dwell positions was associated with a mean catheter error of 0.6 ± 0.2 mm, a maximum catheter error of 0.9 ± 0.4 mm, a mean dwell error of 1.0 ± 0.3 mm, and a maximum dwell error of 1.3 ± 0.7 mm. Error detection and catheter identification sensitivity and specificity of 100% were observed for swap, mix and shift (≥2.6 mm for error detection; ≥2.7 mm for catheter identification) errors. A mean detected shift of 1.8 ± 0.4 mm and a mean identified shift of 1.9 ± 0.4 mm were observed. Registration of the EMT dwell positions to the CT dwell positions was possible with a residual mean error per catheter of 0.6 ± 0.2 mm and a maximum error for any dwell of 1.3 ± 0.7 mm. These low residual registration errors show that quality assurance of the general characteristics of the catheters and of possible errors affecting one specific dwell position is possible. The sensitivity and specificity of the catheter digitization verification algorithm was 100% for swap and mix errors and for shifts ≥2.6 mm. On average, shifts ≥1.8 mm were detected, and shifts ≥1.9 mm were detected and identified.

An efficient approach for differentiating Alzheimer's disease from normal elderly based on multicenter MRI using gray-level invariant features.

Machine learning techniques, along with imaging markers extracted from structural magnetic resonance images, have been shown to increase the accuracy to differentiate patients with Alzheimer's disease (AD) from normal elderly controls. Several forms of anatomical features, such as cortical volume, shape, and thickness, have demonstrated discriminative capability. These approaches rely on accurate non-linear image transformation, which could invite several nuisance factors, such as dependency on transformation parameters and the degree of anatomical abnormality, and an unpredictable influence of residual registration errors. In this study, we tested a simple method to extract disease-related anatomical features, which is suitable for initial stratification of the heterogeneous patient populations often encountered in clinical data. The method employed gray-level invariant features, which were extracted from linearly transformed images, to characterize AD-specific anatomical features. The intensity information from a disease-specific spatial masking, which was linearly registered to each patient, was used to capture the anatomical features. We implemented a two-step feature selection for anatomic recognition. First, a statistic-based feature selection was implemented to extract AD-related anatomical features while excluding non-significant features. Then, seven knowledge-based ROIs were used to capture the local discriminative powers of selected voxels within areas that were sensitive to AD or mild cognitive impairment (MCI). The discriminative capability of the proposed feature was measured by its performance in differentiating AD or MCI from normal elderly controls (NC) using a support vector machine. The statistic-based feature selection, together with the knowledge-based masks, provided a promising solution for capturing anatomical features of the brain efficiently. For the analysis of clinical populations, which are inherently heterogeneous, this approach could stratify the large amount of data rapidly and could be combined with more detailed subsequent analyses based on non-linear transformation.

SU‐E‐J‐116: Validation of Deformable Image Registration Within the 4D Lung Using Two Commercial Algorithms and the User Guided Tool Reg Refine

Purpose:To quantify the registration error between different phases of breathing using two commercially available deformable image registration algorithms. A secondary goal was to better understand the Reg Refine tool developed by MIM software and provide practical recommendations for its use.Methods:The 4D patient data sets of the POPI model were used as phantoms for this work. DIR was performed using MIM Maestro and the SmartAdapt module available within Eclipse. For each of six patients, 100 landmarks were deformed between inhalation and exhalation with their final location compared to the ground truth previously established by expert identification. The base rigid registration was varied to assess the effect on the DIR. For MIM, the Reg Refine tool was also used to try and improve the registration. The number of locked points was varied in addition to varying their location and the size of the window used for local box based alignment.Results:The average displacement for 5 of 6 patients was 6 – 8 mm. For these patients the mean error was low (1 – 1.5 mm) no matter which algorithm was used. SmartAdapt was less sensitive to the initial rigid registration, although both algorithms showed improvement when rigidly aligned to the base of the lung versus the apex. Reg Refine further improved accuracy albeit marginally. For the remaining patient the average displacement was larger (14 mm) and the Reg Refine tool showed marked improvement in reducing the mean and maximum errors from 5 and 35.5 mm to 1.1 and 4.8 mm using as little as ten points.Conclusion:The residual registration error using both SmartAdapt and MIM Maestro for 4D lung patients has been quantified on the order of 1 – 1.5 mm using the POPI model. Reg Refine was also shown effective at improving registrations within MIM Maestro.

Aligning digital holography images of tympanic membrane motion

Modern studies of sound-induced tympanic membrane (TM) motion include stroboscopic Opto-Electronic Holography (OEH). OEH generates high-definition maps of the motion of human and animal TMs that define and quantify spatial patterns of motion. In order to compare such patterns quantitatively between repeated measurements, a common coordinate system is necessary. Differences in the relative position of the sensor and the measured sample induce differences in magnification, shape, and location of the resulting TM image. We present a registration algorithm based on spatial features of the TM motion phase maps in two regimes of motion at distinct stimulus frequencies (200 Hz and 15 kHz). A log-polar representation of the motion phase maps at 15 kHz stimulation enabled the correction of the in-plane rotation. The projective transformation was deduced and corrected on the basis of a set of landmarks defined in the phase maps at 200 kHz stimulation. The technique presented here enables the comparison of repeated TM motion maps with a spatial resolution close to the pixel size and is thus an efficient strategy to align TM motion images. Some residual registration error still needs to be accounted for.

Automatic image registration performance for two different CBCT systems; variation with imaging dose

The performance of an automatic image registration algorithm was compared on image sets collected with two commercial CBCT systems, and the relationship with imaging dose was explored. CBCT images of a CIRS Virtually Human Male Pelvis phantom (VHMP) were collected on Varian TrueBeam/OBI and Elekta Synergy/XVI linear accelerators, across a range of mAs settings. Each CBCT image was registered 100 times, with random initial offsets introduced. Image registration was performed using the grey value correlation ratio algorithm in the Elekta XVI software, to a mask of the prostate volume with 5 mm expansion. Residual registration errors were calculated after correcting for the initial introduced phantom set-up error. Registration performance with the OBI images was similar to that of XVI. There was a clear dependence on imaging dose for the XVI images with residual errors increasing below 4mGy. It was not possible to acquire images with doses lower than ~5mGy with the OBI system and no evidence of reduced performance was observed at this dose. Registration failures (maximum target registration error > 3.6 mm on the surface of a 30mm sphere) occurred in 5% to 9% of registrations except for the lowest dose XVI scan (31%). The uncertainty in automatic image registration with both OBI and XVI images was found to be adequate for clinical use within a normal range of acquisition settings.

A Novel Framework for User-Intervened Correction of Deformable Registration

Purpose Many studies have been done to characterize the performance of deformable registration algorithms (DIR) in different image sets (1,2,3). Residual registration errors can range from insignificant to even worse than rigid registration. However, these algorithms remain essentially black boxes in the clinic, with little or no ability to improve the registration results on patients where the results are unacceptable. Here we evaluate a novel framework to allow the user to influence the registration algorithm to achieve a more accurate result.