Volumetric Image Registration Research Articles

Unsupervised Multi-parametric MRI Registration Using Neural Optimal Transport.

Precise deformable image registration of multi-parametric MRI sequences is necessary for radiologists in order to identify abnormalities and diagnose diseases, such as prostate cancer and lymphoma. Despite recent advances in unsupervised learning-based registration, volumetric medical image registration that requires considering the variety of data distributions is still challenging. To address the problem of multi-parametric MRI sequence data registration, we propose an unsupervised domain-transported registration method, called OTMorph by employing neural optimal transport that learns an optimal transport plan to map different data distributions. We have designed a novel framework composed of a transport module and a registration module: the former transports data distribution from the moving source domain to the fixed target domain, and the latter takes the transported data and provides the deformed moving volume that is aligned with the fixed volume. Through end-to-end learning, our proposed method can effectively learn deformable registration for the volumes in different distributions. Experimental results with abdominal multi-parametric MRI sequence data show that our method has superior performance over around 67-85% in deforming the MRI volumes compared to the existing learning-based methods. Our method is generic in nature and can be used to register inter-/intra-modality images by mapping the different data distributions in network training.

Intra-Cardiac Flow from Geometry Prescribed Computational Fluid Dynamics: Comparison with Ultrasound Vector Flow Imaging.

This paper investigates the accuracy of blood flow velocities simulated from a geometry prescribed computational fluid dynamics (CFD) pipeline by applying it to a dynamic heart phantom. The CFD flow patterns are compared to a direct flow measurement by ultrasound vector flow imaging (VFI). The hypothesis is that the simulated velocity magnitudes are within one standard deviation of the measured velocities. The CFD pipeline uses computed tomography angiography (CTA) images with 20 volumes per cardiac cycle as geometry input. Fluid domain movement is prescribed from volumetric image registration using the CTA image data. Inlet and outlet conditions are defined by the experimental setup. VFI is systematically measured in parallel planes, and compared to the corresponding planes in the simulated time dependent three dimensional fluid velocity field. The measured VFI and simulated CFD have similar flow patterns when compared qualitatively. A quantitative comparison of the velocity magnitude is also performed at specific regions of interest. These are evaluated at 11 non-overlapping time bins and compared by linear regression giving R2 = 0.809, SD = 0.060 m/s, intercept = -0.039 m/s, and slope = 1.09. Excluding an outlier at the inlet, the correspondence between CFD and VFI improves to: R2 = 0.823, SD = 0.048 m/s, intercept = -0.030 m/s, and slope = 1.01. The direct comparison of flow patterns shows that the proposed CFD pipeline provide realistic flow patterns in a well-controlled experimental setup. The demanded accuracy is obtained close to the inlet and outlet, but not in locations far from these.

TransMorph: Transformer for unsupervised medical image registration.

In the last decade, convolutional neural networks (ConvNets) have been a major focus of research in medical image analysis. However, the performances of ConvNets may be limited by a lack of explicit consideration of the long-range spatial relationships in an image. Recently, Vision Transformer architectures have been proposed to address the shortcomings of ConvNets and have produced state-of-the-art performances in many medical imaging applications. Transformers may be a strong candidate for image registration because their substantially larger receptive field enables a more precise comprehension of the spatial correspondence between moving and fixed images. Here, we present TransMorph, a hybrid Transformer-ConvNet model for volumetric medical image registration. This paper also presents diffeomorphic and Bayesian variants of TransMorph: the diffeomorphic variants ensure the topology-preserving deformations, and the Bayesian variant produces a well-calibrated registration uncertainty estimate. We extensively validated the proposed models using 3D medical images from three applications: inter-patient and atlas-to-patient brain MRI registration and phantom-to-CT registration. The proposed models are evaluated in comparison to a variety of existing registration methods and Transformer architectures. Qualitative and quantitative results demonstrate that the proposed Transformer-based model leads to a substantial performance improvement over the baseline methods, confirming the effectiveness of Transformers for medical image registration.

Treatment effect analysis of curettage for jaw cysts based on volumetric assessment and image registration

The study was conducted to develop a visual and intuitive quantitative evaluation method for maxillary cystic lesions after curettage. Mimics 16.0 and Geomagic Studio 2013 were used to form a precise reconstruction of the cystic lesion morphology of 60 cases; the average reduction rates and 95% confidence interval were calculated. Computed tomography (CT) registration was performed before and after surgery to observe morphology features of the bone regeneration of the cystic area. The average reduction rates (RR) of the cysts after curettage were (43.56 ± 16.79)%, (54.33 ± 17.15)% and (68.53 ± 15.99)% at 3 months, 6 months and 12 months after surgery, respectively. The average monthly reduction rates (MRR) were (12.07 ± 4.35)%, (8.16 ± 2.84)% and (5.35 ± 1.52)% at 3 months, 6 months and 12 months after surgery, respectively. Correlation analysis by comparing with each group showed that the effect of sex and age in the 3-month group and the initial size in the 12-month group on RR and MRR were statistically significant. Within the limitations of the study it seems that the chosen approach for quantitative evaluation of the therapeutic effect of curettage for jaw cystic lesions might facilitate visual and quantitative follow-up of cyst curettage and timely detection of recurrence.

FDRN: A fast deformable registration network for medical images.

Deformable image registration is a fundamental task in medical imaging. Due to the large computational complexity of deformable registration of volumetric images, conventional iterative methods usually face the tradeoff between the registration accuracy and the computation time in practice. In order to boost the performance of deformable registration in both accuracy and runtime, we propose a fast unsupervised convolutional neural network for deformable image registration. The proposed registration model FDRN possesses a compact encoder-decoder network architecture which employs a pair of fixed and moving images as input and outputs a three-dimensional displacement vector field (DVF) describing the offsets between the corresponding voxels in the fixed and moving images. In order to efficiently utilize the memory resources and enlarge the model capacity, we adopt additive forwarding instead of channel concatenation and deepen the network in each encoder and decoder stage. To facilitate the learning efficiency, we leverage skip connection within the encoder and decoder stages to enable residual learning and employ an auxiliary loss at the bottom layer with lowest resolution to involve deep supervision. Particularly, the low-resolution auxiliary loss is weighted by an exponentially decayed parameter during the training phase. In conjunction with the main loss in high-resolution grid, a coarse-to-fine learning strategy is achieved. Last but not least, we involve a proposed multi-label segmentation loss (SL) to improve the network performance in Dice score in case the segmentation prior is available. Comparing to the SL using average Dice score, the proposed SL does not require additional memory in the training phase and improves the registration accuracy efficiently. We evaluated FDRN on multiple brain MRI datasets from different aspects including registration accuracy, model generalizability, and model analysis. Experimental results demonstrate that FDRN performs better than the state-of-the-art registration method VoxelMorph by 1.46% in Dice score in LPBA40. In addition to LPBA40, FDRN obtains the best Dice and NCC among all the investigated methods in the unseen MRI datasets including CUMC12, MGH10, ABIDE, and ADNI by a large margin. The proposed FDRN provides better performance than the existing state-of-the-art registration methods for brain MR images by resorting to the compact autoencoder structure and efficient learning. Additionally, FDRN is a generalized framework for image registration which is not confined to a particular type of medical images or anatomy.

Joint affine and deformable three-dimensional networks for brain MRI registration.

Volumetric medical image registration has important clinical significance. Traditional registration methods may be time-consuming when processing large volumetric data due to their iterative optimizations. In contrast, existing deep learning-based networks can obtain the registration quickly. However, most of them require independent rigid alignment before deformable registration; these two steps are often performed separately and cannot be end-to-end. We propose an end-to-end joint affine and deformable network for three-dimensional (3D) medical image registration. The proposed network combines two deformation methods; the first one is for obtaining affine alignment and the second one is a deformable subnetwork for achieving the nonrigid registration. The parameters of the two subnetworks are shared. The global and local similarity measures are used as loss functions for the two subnetworks, respectively. Moreover, an anatomical similarity loss is devised to weakly supervise the training of the whole registration network. Finally, the trained network can perform deformable registration in one forward pass. The efficacy of our network was extensively evaluated on three public brain MRI datasets including Mindboggle101, LPBA40, and IXI. Experimental results demonstrate our network consistently outperformed several state-of-the-art methods with respect to the metrics of Dice index (DSC), Hausdorff distance (HD), and average symmetric surface distance (ASSD). The proposed network provides accurate and robust volumetric registration without any pre-alignment requirement, which facilitates the end-to-end deformable registration.

Unsupervised 3D End-to-end Deformable Network for Brain MRI Registration.

Volumetric medical image registration has important clinical significance. Traditional registration methods may be time-consuming when processing large volumetric data due to their iterative optimizations. In contrast, existing deep learning-based networks can obtain the registration quickly. However, most of them require independent rigid alignment before deformable registration; these two steps are often performed separately and cannot be end-to-end. Moreover, registration ground-truth is difficult to obtain for supervised learning methods. To tackle the above issues, we propose an unsupervised 3D end-to-end deformable registration network. The proposed network cascades two subnetworks; the first one is for obtaining affine alignment, and the second one is a deformable subnetwork for achieving the non-rigid registration. The parameters of the two subnetworks are shared. The global and local similarity measures are used as loss functions for the two subnetworks, respectively. The trained network can perform end-to-end deformable registration. We conducted experiments on brain MRI datasets (LPBA40, Mindboggle101, and IXI) and experimental results demonstrate the efficacy of the proposed registration network.

The Genetics of Encephalization in Eight Inbred Mouse Strains and their F1 Crosses

Increased relative brain size (encephalization) is an important feature of primate and human evolution that has been associated with increased intelligence. Although the heritability of brain size is known to be high, the extent to which different genetic factor types (e.g., additive genotype, maternal genotype, sex, nonadditive genetic interactions) contribute to measured brain size and encephalization are not as well understood. Endocast measures from a large normative sample of genetically diverse mouse lines were estimated to quantify 1) how increased encephalization impacts neurocranial shape and 2) the contribution of various genetic factors towards determining encephalization. Digital endocasts were estimated for 1182 mice representing the eight Collaborative Cross mouse inbred founder lines and 62 reciprocal F1 crosses. Endocasts were estimated by applying a single atlas image endocast segmentation to micro‐computed tomography (μCT) images of each specimen’s skull, based on the nonlinear volumetric image registration of each μCT to the representative atlas image. This method was validated by comparison of automatically extracted endocasts to endocasts that were manually segmented (using Endex software) for a large subset of the specimens. Measures of endocast volume, length, width, and height represent different aspects of endocast size. Ratios of volume over cranial base length (IRE) and volume over body weight (V2W) represent encephalization relative to cranial size and body size, respectively. Diallel analysis (BayesDiallel in R) was used to estimate the genetic influence of multiple additive and nonadditive genetic factors on each measure. The total heritability of each size and encephalization measure is between 0.82 and 0.87, with additive founder strain genotype effects accounting for the vast majority of this heritability. There are also significant inbreeding, sex, maternal genotype, and founder strain cross‐specific effects. The additive effects of founder genotypes were more pronounced for V2W than for IRE, with some of the strains showing opposite directions of effect between these encephalization measures. Care must be taken in the measure of overall size used to estimate encephalization. While inbred mice tend to have smaller overall size, they tend to have a higher encephalization, likely because the impact of inbreeding on brain volume is not as severe as the impact on skull size or body weight. In addition to the strength of genotypic effects, we quantified the influence of increased encephalization on overall cranial vault shape. As encephalization increases, endocasts display more substantial increases in cranial vault height than either length or width. Variation in the mouse cranial base size appears limited, while the shape of later ossifying vault bones strongly covaries with relative brain size. These results support the conclusion that encephalization is highly heritable and indicate which types of genetic factors contribute most strongly. Additionally, relatively large mouse brains are accommodated primarily through variation in cranial vault height, leading to a round cranial vault shape that is analogous to globular vault of modern humans.Support or Funding InformationFinancial support provided by Stony Brook University Startup Funds to CJP

Volumetric Image Registration From Invariant Keypoints

We present a method for image registration based on 3D scale- and rotation-invariant keypoints. The method extends the scale invariant feature transform (SIFT) to arbitrary dimensions by making key modifications to orientation assignment and gradient histograms. Rotation invariance is proven mathematically. Additional modifications are made to extrema detection and keypoint matching based on the demands of image registration. Our experiments suggest that the choice of neighborhood in discrete extrema detection has a strong impact on image registration accuracy. In head MR images, the brain is registered to a labeled atlas with an average Dice coefficient of 92%, outperforming registration from mutual information as well as an existing 3D SIFT implementation. In abdominal CT images, the spine is registered with an average error of 4.82 mm. Furthermore, keypoints are matched with high precision in simulated head MR images exhibiting lesions from multiple sclerosis. These results were achieved using only affine transforms, and with no change in parameters across a wide variety of medical images. This paper is freely available as a cross-platform software library.

Constrained distance transforms for spatial atlas registration.

BackgroundSpatial frameworks are used to capture organ or whole organism image data in biomedical research. The registration of large biomedical volumetric images is a complex and challenging task, but one that is required for spatially mapped biomedical atlas systems. In most biomedical applications the transforms required are non-rigid and may involve significant deformation relating to variation in pose, natural variation and mutation. Here we develop a new technique to establish such transformations for mapping data that cannot be achieved by existing approaches and that can be used interactively for expert editorial review.ResultsThis paper presents the Constrained Distance Transform (CDT), a novel method for interactive image registration. The CDT uses radial basis function transforms with distances constrained to geodesics within the domains of the objects being registered. A geodesic distance algorithm is discussed and evaluated. Examples of registration using the CDT are presented.ConclusionThe CDT method is shown to be capable of simultaneous registration and foreground segmentation even when very large deformations are required.

Paired Inspiratory/Expiratory Volumetric CT and Deformable Image Registration for Quantitative and Qualitative Evaluation of Airflow Limitation in Smokers with or without COPD

To evaluate paired inspiratory/expiratory computed tomography (CT; iCT/eCT) and deformable image registration for quantitative and qualitative assessment of airflow limitation in smokers. Paired iCT/eCT images acquired from 35 smokers (30 men and 5 women) were coregistered and subtraction images (air trapping CT images [aCT]) generated. To evaluate emphysema quantitatively, the percentage of low-attenuation volume (LAV%) on iCT was calculated at -950 HU, as were mean and kurtosis on aCT for quantitative assessment of air trapping. Parametric response maps of emphysema (PRMe) and of functional small airways disease (PRMs) were also obtained. For qualitative evaluation of emphysema, low-attenuation areas on iCT were scored by consensus of two radiologists using Goddard classification. To assess air trapping qualitatively, the degree of air trapping on aCT was scored. For each quantitative and qualitative index, the Spearman rank correlation coefficient for forced expiratory flow in 1 second was calculated, and differences in correlation coefficients were statistically tested. The correlation coefficients for the indices were as follows: mean on aCT, 0.800; kurtosis on aCT, -0.726; LAV%, -0.472; PRMe, -0.570; PRMs, -0.565; addition of PRMe and PRMs, -0.653; emphysema score, -0.502; air trapping score, -0.793. The indices showing significant differences were as follows: mean on aCT and addition of PRMe and PRMs (P=1.43×10(-8)); air trapping score and emphysema score (P=.0169). Air trapping images yielded more accurate quantitative and qualitative evaluation of airflow limitation than did LAV%, PRMe, PRMs, and Goddard classification.

Effective Volumetric Feature Modeling and Coarse Correspondence via Improved 3DSIFT and Spectral Matching

This paper presents a nonrigid coarse correspondence computation algorithm for volumetric images. Our matching algorithm first extracts then correlates image features based on a revised and improved 3DSIFT (I3DSIFT) algorithm. With a scale-related keypoint reorientation and descriptor construction, this feature correlation is less sensitive to image rotation and scaling. Then, we present an improved spectral matching (ISM) algorithm on correlated features to obtain a one-to-one mapping between corresponded features. One can effectively extend this feature correspondence to dense correspondence between volume images. Our algorithm can benefit nonrigid volumetric image registration in many tasks such as motion modeling in medical image analysis and processing.

Nonrigid Registration of Volumetric Images Using Ranked Order Statistics

Nonrigid image registration techniques using intensity based similarity measures are widely used in medical imaging applications. Due to high computational complexities of these techniques, particularly for volumetric images, finding appropriate registration methods to both reduce the computation burden and increase the registration accuracy has become an intensive area of research. In this paper, we propose a fast and accurate nonrigid registration method for intra-modality volumetric images. Our approach exploits the information provided by an order statistics based segmentation method, to find the important regions for registration and use an appropriate sampling scheme to target those areas and reduce the registration computation time. A unique advantage of the proposed method is its ability to identify the point of diminishing returns and stop the registration process. Our experiments on registration of end-inhale to end-exhale lung CT scan pairs, with expert annotated landmarks, show that the new method is both faster and more accurate than the state of the art sampling based techniques, particularly for registration of images with large deformations.

In vivo imaging of immune cell dynamics in skin in response to zinc-oxide nanoparticle exposure

Zinc oxide (ZnO) nanoparticles (NPs) are widely used in cosmetic and sunscreen products which are applied topically to the skin. Despite their widespread use, the safety and biological response of these particles remains an active area of investigation. In this paper we present methods based on in vivo multiphoton microscopy (MPM) in skin to address relevant questions about the potential toxicity and immunological response of ZnO NPs. Registration of time-lapse volumetric MPM images allows the same skin site to be tracked across multiple days for visualizing and quantifying cellular and structural changes in response to NP exposure. Making use of the unique optical properties of ZnO enables high contrast detection of the NPs in the presence of strong autofluorescence and second harmonic generation (SHG) background from the skin. A green fluorescent protein (GFP) bone marrow (BM) transplanted mouse model is used to visualize and assess the dynamic response of BM-derived immune cells. These cells are visualized to assess the potential for ZnO NPs to interact with immune cells and elicit an immune reaction in skin. We investigate both topical and dermal exposure of the ZnO NPs. The methods and findings presented in this paper demonstrate a novel approach for tracking ZnO NPs in vivo and for visualizing the cellular response of the exposed tissue to assess the immunological response and potential toxicity of these particles.

μFEA successfully exhibits higher stresses and strains in microdamaged regions of whole vertebrae

Micro-finite element (μFE) modeling has shown promise in evaluating the structural integrity of trabecular bone. Histologic microcrack analyses have been compared to μFE models of trabecular bone cores to demonstrate the potential of this technique. To date this has not been achieved in whole bone structures, and comparisons of histologic microcrack and μFE results have been limited due to challenges in alignment of 2D sections with 3D data sets. The goal of this study was to ascertain if image registration can facilitate determination of a relationship between stresses and strains generated from μFE models of whole vertebrae and histologically identified microdamage. μFE models of three whole vertebrae, stained sequentially with calcein and fuchsin, were generated with accurate integration of element sets representing the histologic sections based on volumetric image registration. Displacement boundary conditions were applied to the μFE models based on registration of loaded and unloaded μCT images. Histologically labeled damaged regions were found to have significantly higher von Mises stresses and principle strains in the μFE models, as compared to undamaged regions. This work provides a new robust method for generating and histologically validating μFE models of whole bones that can represent trabecular damage resulting from complex physiologic loading.

Effect of Data Normalization on the Creation of Neuro-Probabilistic Atlases

In the past 15 years, rapid improvements in imaging technology and methodology have had a tremendous impact on how we study the human brain. During deep brain stimulation surgeries, detailed anatomical images can be combined with physiological data obtained by microelectrode recordings and microstimulations to address questions relating to the location of specific motor or sensorial functions. The main advantage of techniques such as microelectrode recordings and microstimulations over brain imaging is their ability to localize patient physiological activity with a high degree of spatial resolution. Aggregating data acquired from large populations permits to build what are commonly referred to as statistical atlases. Data points from statistical atlases can be combined to produce probabilistic maps. A crucial step in this process is the intersubject spatial normalization that is required to relate a position in one subject's brain to a position in another subject's brain. In this paper, we study the impact of spatial normalization techniques on building statistical atlases. We find that the Talairach or anterior-posterior commissure coordinate system commonly used in the medical literature produces atlases that are more dispersed than those obtained with normalization methods that rely on nonlinear volumetric image registration. We also find that the maps produced using nonlinear techniques correlate with their expected anatomic positions.

Volume sweeping and bodyline matching for automated prealignment in volumetric medical image registration

For an automated image registration to converge to a good registration, it is crucial that the initial registration is within the capture range of the true registration, as local optimization methods are frequently employed. The ways to set an initial registration in current practice are not ideal and it is highly desirable to automate this initial registration (prealignment). Two automatic prealignment methods are reported here. In the volume sweeping approach, one volume is swept through the other, the overlapping volumes are coarsely aligned in the x and y directions, and a similarity measure is calculated at each sweeping position. Once sweeping is done, the position that gives the best similarity measure is chosen as the prealignment. In the second bodyline matching approach, patient bodyline profiles (the furthest anterior or posterior body boundary points) are extracted from two volumes and objectively matched. A prealignment is then derived from the matched bodylines. Both methods are tested on 19 PET/CT alignments of five patients with known ground truths acquired on hybrid PET/CT scanners. The absolute differences in the three translational parameters between the volume sweeping prealignment and the ground truth are 6.1±3.9, 2.2±2.7, and 4.2±6.1mm, and between the bodyline matching and the ground truth are 5.2±3.0, 3.3±3.0, and 3.3±4.1mm, which are within the capture range of automatic registration algorithms. The volume sweeping and bodyline matching can thus be used as a preprocessing step for automatic registration, making it possible to run an automatic registration algorithm without user intervention.

SU-E-J-19: How Should CTV to PTV Margin Be Created - Analysis of Set-Up Uncertainties of Different Body Parts Using Daily Image Guidance.

The objective for this study is to systematically evaluate how the CTV to PTV margin should be created to compensate for setup error as afunction of treatment site. A total of 1757 setup MVCT images were analyzed using the registration between the MVCT and planning KVCT images. Among them, 526 images were from patients with head and neck cancers; 342 were from thoracic site; 447 from upper abdomen and 442 from patients treated at pelvic region. The correction was shifts in 3 directions from setup with skin marks to final treatment position due to volumetric image registration. Registration was based on soft tissue anatomy when tumor was visible in both KVCT and MVCT. Otherwise, bony landmarks were used when tumors could not be identified. The anterior to posterior (A/P) direction requires the smallest correction for all body sites. The percentage of treatment with < 3mm set-up correction in A/P direction was 82.4%-96.0% for all disease sites. In medial to lateral direction (M/L), that number was 51.5%-74.7%. The largest set-up correction occurred in the superior to inferior (S/I) direction. As a function of the body site, 68.6% of head and neck patient had <3mm shift in the S/I direction. However, that number dropped to 41.8%〠42.5% and 31.9% for thoracic, upper abdomen and pelvic sites respectively. In addition, 48.0% of pelvic tumor required >5mm shift in the S/I direction. For thoracic and abdominal tumors, there were still 34.9% and 30.7% of the treatments where >5mm shift in the S/I direction was needed. Uniform expansion margins in all directions are often used in radiation oncology practice. Our study showed setup error was directionally dependent. Therefore, non-uniform margin should be applied. Smaller margin can be used in A/P while larger expansion needs to be given in the S/I direction.

Implementation of Remote 3-Dimensional Image Guided Radiation Therapy Quality Assurance for Radiation Therapy Oncology Group Clinical Trials

To report the process and initial experience of remote credentialing of three-dimensional (3D) image guided radiation therapy (IGRT) as part of the quality assurance (QA) of submitted data for Radiation Therapy Oncology Group (RTOG) clinical trials; and to identify major issues resulting from this process and analyze the review results on patient positioning shifts. Image guided radiation therapy datasets including in-room positioning CT scans and daily shifts applied were submitted through the Image Guided Therapy QA Center from institutions for the IGRT credentialing process, as required by various RTOG trials. A centralized virtual environment is established at the RTOG Core Laboratory, containing analysis tools and database infrastructure for remote review by the Physics Principal Investigators of each protocol. The appropriateness of IGRT technique and volumetric image registration accuracy were evaluated. Registration accuracy was verified by repeat registration with a third-party registration software system. With the accumulated review results, registration differences between those obtained by the Physics Principal Investigators and from the institutions were analyzed for different imaging sites, shift directions, and imaging modalities. The remote review process was successfully carried out for 87 3D cases (out of 137 total cases, including 2-dimensional and 3D) during 2010. Frequent errors in submitted IGRT data and challenges in the review of image registration for some special cases were identified. Workarounds for these issues were developed. The average differences of registration results between reviewers and institutions ranged between 2 mm and 3 mm. Large discrepancies in thesuperior-inferior direction were found for megavoltage CT cases, owing to low spatial resolution in this direction for most megavoltage CT cases. This first experience indicated that remote review for 3D IGRT as part of QA for RTOG clinical trials is feasible and effective. The magnitude of registration discrepancy between institution and reviewer was presented, and the major issues were investigated to further improve this remote evaluation process.

A feasibility study of image registration using volumetrically classified, motion-free bony landmarks in thoracic 4DCT images for image-guided patient setup

As rigid image registration becomes unreliable in the presence of involuntary organ motion, we present a novel approach to register CT images using stable bony landmarks for image-guided patient setup. Using 3D Volumetric Image Registration (3DVIR) technique, bony anatomy is volumetrically-classified as registration landmark, while soft tissues are ignored. Based on 4DCT, it was found that the spine, posterior ribs and clavicles do not move with respiration and remain registered throughout the breathing cycle. However, mutual information based registration produces an error of 1–2 mm due to moving soft tissues. It is suggested that the 3DVIR can improve image-guided setup.