Patient-specific Deformation Research Articles

CT respiratory motion synthesis using joint supervised and adversarial learning

Objective. Four-dimensional computed tomography (4DCT) imaging consists in reconstructing a CT acquisition into multiple phases to track internal organ and tumor motion. It is commonly used in radiotherapy treatment planning to establish planning target volumes. However, 4DCT increases protocol complexity, may not align with patient breathing during treatment, and lead to higher radiation delivery. Approach. In this study, we propose a deep synthesis method to generate pseudo respiratory CT phases from static images for motion-aware treatment planning. The model produces patient-specific deformation vector fields (DVFs) by conditioning synthesis on external patient surface-based estimation, mimicking respiratory monitoring devices. A key methodological contribution is to encourage DVF realism through supervised DVF training while using an adversarial term jointly not only on the warped image but also on the magnitude of the DVF itself. This way, we avoid excessive smoothness typically obtained through deep unsupervised learning, and encourage correlations with the respiratory amplitude. Main results. Performance is evaluated using real 4DCT acquisitions with smaller tumor volumes than previously reported. Results demonstrate for the first time that the generated pseudo-respiratory CT phases can capture organ and tumor motion with similar accuracy to repeated 4DCT scans of the same patient. Mean inter-scans tumor center-of-mass distances and Dice similarity coefficients were 1.97 mm and 0.63, respectively, for real 4DCT phases and 2.35 mm and 0.71 for synthetic phases, and compares favorably to a state-of-the-art technique (RMSim). Significance. This study presents a deep image synthesis method that addresses the limitations of conventional 4DCT by generating pseudo-respiratory CT phases from static images. Although further studies are needed to assess the dosimetric impact of the proposed method, this approach has the potential to reduce radiation exposure in radiotherapy treatment planning while maintaining accurate motion representation. Our training and testing code can be found at https://github.com/cyiheng/Dynagan.

Compliance boundary conditions for patient-specific deformation simulation using the finite element method

Tissue deformation simulations for pre-operative planning or intra-operative guidance of medical procedures require accurate patient-specific models and are commonly performed using the Finite Element Method (FEM). Since only a small part of the entire body is typically observed with medical imaging, the deformation models are often limited to a relatively small region-of-interest (ROI). Surrounding this ROI, one then needs to define suitable boundary conditions for an accurate simulation. Conventionally, boundary conditions are set arbitrarily or heuristically at chosen model locations; typically as either zero-displacement or -force constraint, which obviously are suboptimal where ROI borders are neither fixed (e.g. on bone) or free (e.g. skin facing the air). In this work, we present a novel boundary-condition formulation, called compliance boundary conditions (CBC), which approximate the effect of anatomy outside this ROI and augment this onto the ROI border nodes. CBC can be parametrized from observed tissue displacements, e.g. tracked in ultrasound (US) or magnetic-resonance imaging. It is inherently embedded in the FEM deformation model to be used for computing any interaction response. CBC is a generalization of conventional boundary constraints, where the typical zero-displacement and -force constraints are obtained at the two extremes of the given CBC parameter. We demonstrate CBC for linear- and quadratic-strain FEM models in 2D and 3D numerical phantoms, for which different element/integration formulations and the effect of noise are studied. CBC is shown to reduce displacement errors for both 2D and 3D numerical phantoms by more than 50% compared to conventional boundary conditions. We also present CBC on tissue-mimicking gelatin phantom experiments from displacements observed in US images. In an application scenario of simulating needle insertion for prostate brachytherapy, CBC is shown to reduce seed placement errors by more than 70% compared to conventional boundary conditions.

A machine learning approach for real-time modelling of tissue deformation in image-guided neurosurgery.

Accurate reconstruction and visualisation of soft tissue deformation in real time is crucial in image-guided surgery, particularly in augmented reality (AR) applications. Current deformation models are characterised by a trade-off between accuracy and computational speed. We propose an approach to derive a patient-specific deformation model for brain pathologies by combining the results of pre-computed finite element method (FEM) simulations with machine learning algorithms. The models can be computed instantaneously and offer an accuracy comparable to FEM models. A brain tumour is used as the subject of the deformation model. Load-driven FEM simulations are performed on a tetrahedral brain mesh afflicted by a tumour. Forces of varying magnitudes, positions, and inclination angles are applied onto the brain's surface. Two machine learning algorithms-artificial neural networks (ANNs) and support vector regression (SVR)-are employed to derive a model that can predict the resulting deformation for each node in the tumour's mesh. The tumour deformation can be predicted in real time given relevant information about the geometry of the anatomy and the load, all of which can be measured instantly during a surgical operation. The models can predict the position of the nodes with errors below 0.3mm, beyond the general threshold of surgical accuracy and suitable for high fidelity AR systems. The SVR models perform better than the ANN's, with positional errors for SVR models reaching under 0.2mm. The results represent an improvement over existing deformation models for real time applications, providing smaller errors and high patient-specificity. The proposed approach addresses the current needs of image-guided surgical systems and has the potential to be employed to model the deformation of any type of soft tissue.

Patient-specific Deformation Modelling via Elastography: Application to Image-guided Prostate Interventions

Image-guided prostate interventions often require the registration of preoperative magnetic resonance (MR) images to real-time transrectal ultrasound (TRUS) images to provide high-quality guidance. One of the main challenges for registering MR images to TRUS images is how to estimate the TRUS-probe-induced prostate deformation that occurs during TRUS imaging. The combined statistical and biomechanical modeling approach shows promise for the adequate estimation of prostate deformation. However, the right setting of the biomechanical parameters is very crucial for realistic deformation modeling. We propose a patient-specific deformation model equipped with personalized biomechanical parameters obtained from shear wave elastography to reliably predict the prostate deformation during image-guided interventions. Using data acquired from a prostate phantom and twelve patients with suspected prostate cancer, we compared the prostate deformation model with and without patient-specific biomechanical parameters in terms of deformation estimation accuracy. The results show that the patient-specific deformation model possesses favorable model ability, and outperforms the model without patient-specific biomechanical parameters. The employment of the patient-specific biomechanical parameters obtained from elastography for deformation modeling shows promise for providing more precise deformation estimation in applications that use computer-assisted image-guided intervention systems.

Cross Population Motion Modeling Applied to Attenuation Correction of Respiratory Gated F18-FDG PET

Accurate attenuation correction in PET/CT relies upon anatomically matched PET and CT scans, however, attenuation mismatch artifacts are often present due to continuous respiration during both the PET and CT scans. Acquiring respiratory gated CT and PET scans in which each frame of the 4DCT is matched with a frame in a respiratory gated PET scan is one strategy for obtaining temporally correlated PET and CT images. In this paper we propose a cross population model as an alternative to 4DCT for estimating the deformation of attenuating objects within the thorax. Principal components analysis was used to extract components of motion from the 4DCT scans of a cohort of 10 patients and the components were then fitted to the frames from a respiratory gated PET scan. The cross population model produced an artificial, yet realistic patient specific deformation field for each frame even though no 4DCT was acquired in conjunction with the gated PET scan. The deformation field was then applied to the static spiral CT image to produce CT frames matching each PET frame. The first 5 components of motion accounted for 95% of the movement in regions where there was sufficient PET camera sensitivity to detect motion. We propose this approach as the basis for a robust method of transforming a static spiral CT into spatially matched attenuation maps. We investigate improvements in image contrast as a result of using our method for attenuation correction.

Upper Airway Elasticity Estimation in Pediatric Down Syndrome Sleep Apnea Patients Using Collapsible Tube Theory.

Elasticity of the soft tissues surrounding the upper airway lumen is one of the important factors contributing to upper airway disorders such as snoring and obstructive sleep apnea. The objective of this study is to calculate patient specific elasticity of the pharynx from magnetic resonance (MR) images using a 'tube law', i.e., the relationship between airway cross-sectional area and transmural pressure difference. MR imaging was performed under anesthesia in children with Down syndrome (DS) and obstructive sleep apnea (OSA). An airway segmentation algorithm was employed to evaluate changes in airway cross-sectional area dilated by continuous positive airway pressure (CPAP). A pressure-area relation was used to make localized estimates of airway wall stiffness for each patient. Optimized values of patient specific Young's modulus for tissue in the velopharynx and oropharynx, were estimated from finite element simulations of airway collapse. Patient specific deformation of the airway wall under CPAP was found to exhibit either a non-linear 'hardening' or 'softening' behavior. The localized airway and tissue elasticity were found to increase with increasing severity of OSA. Elasticity based patient phenotyping can potentially assist clinicians in decision making on CPAP and airway or tissue elasticity can supplement well-known clinical measures of OSA severity.

A method for optimal fit of patient-specific fracture fixation plates

Anatomically pre-contoured fracture fixation plates are a treatment option for bone fractures. A well-fitting plate can be used as a tool for anatomical reduction of the fractured bone. However, recent studies showed that some plates fit poorly for many patients due to considerable shape variations between bones of the same anatomical site. Therefore, the plates have to be manually fitted and deformed by surgeons to fit each patient optimally. The process is time-intensive and labor-intensive, and could lead to adverse clinical implications such as wound infection or plate failure. This paper proposes a new iterative method to simulate the patient-specific deformation of an optimally fitting plate for pre-operative planning purposes. We further demonstrate the validation of the method through a case study. The proposed method involves the integration of four commercially available software tools, Matlab, Rapidform2006, SolidWorks, and ANSYS, each performing specific tasks to obtain a plate shape that fits optimally for an individual tibia and is mechanically safe. A typical challenge when crossing multiple platforms is to ensure correct data transfer. We present an example of the implementation of the proposed method to demonstrate successful data transfer between the four platforms and the feasibility of the method.

Simulating liver deformation during respiration using sparse local features.

A new method deforms a 3D liver mesh in an arbitrary phase of respiration. During preprocessing, the method step defines a patient-specific deformation space using two polar shapes of the liver during respiration. 3D magnetic resonance imaging captures patient livers during exhaling and inhaling. Next, using a fully automated nonrigid mesh registration, this method creates the two phases' corresponding surface meshes. Then, it defines the respiration's deformation space by extracting deformation gradients between the exhalation and inhalation meshes. At runtime, the method uses sparse local features suitably obtained from 2D ultrasound imaging to solve the constraint optimization problem that minimizes dissimilarity of deformation gradients between the target deformation and the patient-specific deformation space. Researchers used real patient data to evaluate this method, which could be applicable to image-guided tumor ablations.

SU-E-J-64: Towards a Patient Specific Deformation Model in the Male Pelvis for IGRT via Limited Angle Imaging.

To evaluate the feasibility of patient specific deformation models (PSDM) in the male pelvis for IGRT by limited angular imaging. In IGRT via limited angular imaging, insufficient angular projections are acquired to uniquely determine a 3D attenuation distribution. For highly limited geometries, image quality may be too poor for successful non-rigid registration. This can be overcome by restricting the transformation space to one containing only feasible transformations learned from prior 3D images. This has been successfully applied in the lung region where a majority of deformation is due to respiratory motion which can be adequately observed at planning time with RCCT. Typically, the phases of the RCCT are registered together to form an group-wise mean image and transformations to each training image. PCA is then performed on the transformation displacement vector fields. The transformation is found at treatment time by registration of digitally reconstructed radiographs of the transformed image to the measured projections, optimizing over the parameters of the PCA subspace. In the male pelvis, deformation is much more complicated than respiratory deformation and is largely inter-fractional due to changes in bladder and rectal contents, articulation, and motion of the bowels. A similar model is developed for the male pelvis which takes into account pelvic anatomical information and handles the more complicated deformation space. Using the leave-one-out method, dice similarity coefficients in the prostate compared with manual segmentations are increased over the those obtained by rigid registration and are comparable with those obtained by 3D non-rigid registration methods. This method produces better results than rigid registration and is comparable with results obtained by 3D/3D registration even though it uses limited angle projections. However, its relies on daily training CTs, so it is not yet a viable clinical method. Funding provided in part by Siemens Medical.

Navigator channel adaptation to reconstruct three dimensional heart volumes from two dimensional radiotherapy planning data

BackgroundBiologically-based models that utilize 3D radiation dosimetry data to estimate the risk of late cardiac effects could have significant utility for planning radiotherapy in young patients. A major challenge arises from having only 2D treatment planning data for patients with long-term follow-up. In this study, we evaluate the accuracy of an advanced deformable image registration (DIR) and navigator channels (NC) adaptation technique to reconstruct 3D heart volumes from 2D radiotherapy planning images for Hodgkin's Lymphoma (HL) patients.MethodsPlanning CT images were obtained for 50 HL patients who underwent mediastinal radiotherapy. Twelve image sets (6 male, 6 female) were used to construct a male and a female population heart model, which was registered to 23 HL "Reference" patients' CT images using a DIR algorithm, MORFEUS. This generated a series of population-to-Reference patient specific 3D deformation maps. The technique was independently tested on 15 additional "Test" patients by reconstructing their 3D heart volumes using 2D digitally reconstructed radiographs (DRR). The technique involved: 1) identifying a matching Reference patient for each Test patient using thorax measurements, 2) placement of six NCs on matching Reference and Test patients' DRRs to capture differences in significant heart curvatures, 3) adapting the population-to-Reference patient-specific deformation maps to generate population-to-Test patient-specific deformation maps using linear and bilinear interpolation methods, 4) applying population-to-Test patient specific deformation to the population model to reconstruct Test-patient specific 3D heart models. The percentage volume overlap between the NC-adapted reconstruction and actual Test patient's true heart volume was calculated using the Dice coefficient.ResultsThe average Dice coefficient expressed as a percentage between the NC-adapted and actual Test model was 89.4 ± 2.8%. The modified NC adaptation technique made significant improvements to the population deformation heart models (p = 0.01). As standard evaluation, the residual Dice error after adaptation was comparable to the volumetric differences observed in free-breathing heart volumes (p = 0.62).ConclusionsThe reconstruction technique described generates accurate 3D heart models from limited 2D planning data. This development could potentially be used to retrospectively calculate delivered dose to the heart for historically treated patients and thereby provide a better understanding of late radiation-related cardiac effects.

Predictive cardiac motion modelling and correction

The quest for higher image quality and resolution for assessing cardiovascular anatomy and function requires an effective way of handing non-periodic, patient specific tissue deformation. This paper presents a new technique for predictive cardiac motion modelling and correction. It uses partial least squares regression to extract intrinsic relationships between 3D cardiac deformation due to respiration and 1D real time measurable intensity traces taken from tissue interfaces. This abstraction allows for the construction of a patient specific model for dense motion field prediction. The effectiveness of the proposed method is validated for a range of respiratory positions with data acquired from 10 asymptomatic subjects, using reconstructed image data sets.