Conference proceedings. International Conference on Image Formation in X-Ray Computed Tomography最新文献

There has been a great deal of work seeking to improve image quality in CT reconstruction through deep-learning-based denoising; however, there are many applications where it is spatial resolution that limits application and diagnostics. In this work, we week to improve spatial resolution in CT reconstructions through a combination of deep learning and physical modeling of detector blur. To achieve this goal, we leverage diffusion models as deep image priors to help regularize a joint deblurring and reconstruction problem. Specifically, we adopt Diffusion Posterior Sampling (DPS) as a way to combine a deep prior with a likelihood-based forward model for the measurements. The model we adopt is nonlinear since detector blur is applied after the nonlinear attenuation given by the Beer-Lambert lab. We trained a score estimator for a CT score-based prior, and then apply Bayes rule to combine this prior with a measurement likelihood score for CT reconstruction with detector blur. We demonstrate the approach in simulated data, and compare image outputs with traditional filtered-backprojection (FBP) and model-based iterative reconstruction (MBIR) across a range of exposures. We find a particular advantage of the DPS approach for low exposure data and report on major differences in the errors between DPS and classical reconstruction methods.

In this work, we introduce a new deep learning approach based on diffusion posterior sampling (DPS) to perform material decomposition from spectral CT measurements. This approach combines sophisticated prior knowledge from unsupervised training with a rigorous physical model of the measurements. A faster and more stable variant is proposed that uses a "jumpstarted" process to reduce the number of time steps required in the reverse process and a gradient approximation to reduce the computational cost. Performance is investigated for two spectral CT systems: dual-kVp and dual-layer detector CT. On both systems, DPS achieves high Structure Similarity Index Metric Measure(SSIM) with only 10% of iterations as used in the model-based material decomposition(MBMD). Jumpstarted DPS (JSDPS) further reduces computational time by over 85% and achieves the highest accuracy, the lowest uncertainty, and the lowest computational costs compared to classic DPS and MBMD. The results demonstrate the potential of JSDPS for providing relatively fast and accurate material decomposition based on spectral CT data.

Accurate scatter correction is essential to obtain highquality reconstructions in computed tomography. While many correction strategies for this longstanding issue have been developed, additional efforts may be required for spectral CT imaging - which is particularly sensitive to unmodeled biases. In this work we explore a joint estimation approach within a one-step model-based material decomposition framework to simultaneously estimate material densities and scatter profiles in spectral CT. The method is applied to simulated phantom data obtained using a parametric additive scatter mode, and compared to the unmodeled scatter scenario. In these preliminary experiments, We find that this joint estimation approach has the potential to significantly reduce artifacts associated with unmodeled scatter and to improve material density estimates.

Respiratory motion phantoms can be used for evaluation of CT imaging technologies such as motion artifact reduction algorithms and deformable image registration. However, current respiratory motion phantoms do not exhibit detailed lung tissue structures and thus do not provide a realistic testing environment. This paper presents PixelPrint^4D, a method for 3D-printing deformable lung phantoms featuring highly realistic internal structures, suitable for a broad range of CT evaluations, optimizations, and research. The phantom in this study was designed with a patient 4DCT as a reference and 3D-printed using an extended version of the PixelPrint method for developing patient-specific CT phantoms. A flexible thermoplastic polyurethane (TPU) 3D-printing material was used, which produced regions with attenuation between -840 and -48 Hounsfield units (HU). A linear compression device was then designed and used to compress the phantom in the superior-inferior (SI) direction, and the phantom was scanned at different compression levels matched to the diaphragm displacements measured on the reference patient 4DCT. Deformable image registration (DIR) was performed, and motion vector fields were obtained for both patient and phantom images. SI displacements of selected features in the lung had mean errors of 0.5 mm difference from the patient, or less than the reconstructed slice thickness. In conclusion, the deformable lung phantom developed in this study exhibits realistic lung structures and deformation characteristics under compression, indicating potential for advancing more lifelike respiratory motion phantoms.

Despite the evident benefits of spectral computed tomography (CT) in delivering qualitative imaging superior to that of conventional CT in adults, its application in pediatric diagnostic imaging is still relatively limited due to various reasons, including design limitations and radiation dose considerations. The use of specialized K-edge filters, in conjunction with other spectral technologies, has been demonstrated to improve spectral quantification accuracy. X-ray flux limitations generally pose challenges in these concepts when applied to adults. However, such limitations are not present in pediatric imaging, allowing the full exploitation of K-edge filters to improve performance. To facilitate the adoption of spectral CT's benefits, as seen in the adult population, into pediatric settings, we introduce an innovative double bowtie filter design. This design incorporates a K-edge material coupled with Teflon and is integrated with rapid kVp-switching technology. A Python simulation was built to model a rapid kVp-switching x-ray tube and to estimate Cramer-Rao lower bound (CRLB) noise in photoelectric and Compton scatter basis domains. We estimate a conventional bowtie filter and corresponding reference patient dose before optimizing double bowtie configurations to contain the highest obtainable spectral signal-to-noise content for the specified phantom. Our findings indicate that an optimal combination of holmium and Teflon in the filter geometry can increase spectral SNR up to twofold the conventional estimates, while still maintaining low radiation dose exposure. This study broadens the scope for pediatric patients to fully benefit from the capabilities of spectral CT.

Cone-beam CT imaging using non-circular orbits has been demonstrated to be effective in reducing artifacts around metal. With the increasing interest in spectral imaging in the interventional suite, there are potential advantages to combine both technologies to yield further image quality benefits. We simulated a neuro-interventional application where imaging around the embolization is challenged by metal artifacts and the differentiation of bleeds and contrast extravasation is difficult with single-energy imaging. The imaging system was simulated with a dual-layer detector and different sinusoidal orbits. Material decomposition used a projection-domain approach followed by a model-based reconstruction of the density line integrals of each basis. The spectral non-circular orbits acquisitions were compared with single-energy circular, single-energy non-circular, and spectral circular orbits. Results using spectral non-circular orbit contain minimal metal artifacts and allow the differentiation of bleeds and contrast extravasation, demonstrating the potential of the combined technologies.

Spectral CT has great potential for a variety of clinical applications due to the improved material discrimination with respect to conventional CT. Many clinical and preclinical spectral CT systems have two spectral channels for dual-energy CT using strategies such as split-filtration, dual-layer detectors, or kVp-switching. However, there are emerging clinical imaging applications which would require three or more spectral sensitivity channels, for example, multiple exogenous contrast agents in a single scan. Spatial-spectral filters are a new spectral CT technology which use x-ray beam modulation to offer greater spectral diversity. The device consists of an array of k-edge filters which divide the x-ray beam into spectrally varied beamlets. This design allows for an arbitrary number of spectral channels; however, traditional two-step reconstruction-decomposition schemes are typically not effective because the measured data for any individual spectral channel is sparse in the projection domain. Instead, we use a one-step model-based material decomposition algorithm to iteratively estimate material density images directly from spectral CT data. In this work, we present a prototype spatial-spectral filter integrated with an x-ray CT test-bench. The filter is composed of an array of tin, erbium, tantalum, and lead filter tiles which spatially modulate the system spectral sensitivity pattern. After the system was characterized and modeled, we conducted a spectral CT scan of a multi-contrast-enhanced phantom containing water, iodine, and gadolinium solutions. We present the resulting spectral CT data as well as the material density images estimated by model-based material decomposition. The calibrated system model is in close agreement with the measured data, and the reconstructed material density images match the ground truth concentrations for the multi-contrast phantom. These preliminary results demonstrate the potential of spatial-spectral filters to enable multi-contrast imaging and other new clinical applications of spectral CT.

Metal artifacts are a major confounding factor for image quality in CT, especially in image-guided surgery scenarios where surgical tools and implants frequently occur in the field-of-view. Traditional metal artifact correction methods typically use algorithmic solutions to interpolate over the highly attenuated projection measurements where metal is present but cannot recover the missing information obstructed by the metal. In this work, we treat metal artifacts as a missing data problem and employ noncircular orbits to maximize data completeness in the presence of metal. We first implement a local data completeness metric based on Tuy's condition as the percentage of great circles sampled by a particular orbit and accounted for the presence of metal by discounting any rays that pass through metal. We then compute the metric over many locations and many possible metal locations to reflect data completeness for arbitrary metal placements within a volume of interest. We used this metric to evaluate the effectiveness of sinusoidal orbits of different magnitudes and frequencies in metal artifact reduction. We also evaluated noncircular orbits in two imaging systems for phantoms with different metal objects and metal arrangements. Among a circular, tilted circular, and a sinusoidal orbit of two cycles per rotation, the latter is shown to most effectively remove metal artifacts. The noncircular orbit not only reduce the extent of streaks, but allows better visualization of spatial frequencies that cannot be recovered by metal artifact correction algorithms. These results illustrate the potential of relatively simple noncircular orbits to be robust against metal implants which ordinarily present significant challenges in interventional imaging.

In this work we compare a novel model-based material decomposition (MBMD) approach against a standard approach in high-resolution spectral CT using multi-layer flat-panel detectors. Physical experiments were conducted using a prototype dual-layer detector and a custom high-resolution iodine-enhanced line-pair phantom. Reconstructions were performed using three methods: traditional filtered back-projection (FBP) followed by image-domain decomposition, idealized MBMD with no blur modeling (iMBMD), and MBMD with system blur modeling (bMBMD). We find that both MBMD methods yielded higher resolution decompositions with lower noise than the FBP method, and that bMBMD further improves spatial resolution over iMBMD due to the additional blur modeling. These results demonstrate the advantages of MBMD in resolution performance and noise control over traditional methods for spectral CT. Model-based material decomposition hence has great potential in high-resolution spectral CT applications.

Spectral CT permits material discrimination beyond the structural information in conventional single-energy CT. Model-based material decomposition facilitates direct estimation of material density from spectral measurements, incorporating a general forward model for arbitrary spectral CT system, a statistical model of spectral CT measurements, and flexible regularization schemes. Such one-step approaches are promising for superior image quality, but the relationship between regularization parameters, imaging conditions, and reconstructed image properties is complicated. More specifically, the estimator is inherently nonlinear and may include additional nonlinearities like edge-preserving regularization, making image quality metrics intended for linear system evaluation difficult to apply. In this work, we seek approaches to quantify the image properties of this inherently nonlinear process through an investigation of perturbation response - the generalized system response to a local perturbation of arbitrary shape, location, and contrast. Such responses include cross-talk between material density channels, and we investigate the application of this metric in a sample spectral CT system. Inspired by the prior work under assumptions of local linearity and shift-invariant we also propose a prediction framework for perturbation response using a perceptron neural network. The proposed prediction framework offers an alternative to exhaustive evaluation and is a potential tool that can be used to prospectively choose optimal regularization parameters based on imaging conditions and diagnostic task.