Conventional Reconstruction Methods Research Articles

Digital Implant-Supported Restoration Planning Placed in Autologous Graft Using Titanium Implants Produced by Additive Manufacturing

This clinical report presents a technique to reconstruct extensively resected mandibles using a combination of autologous bone grafts and additive manufacturing techniques. Mandibular defects, often arising from trauma, tumors, or congenital anomalies, can severely impact both function and aesthetics. Conventional reconstruction methods have their limitations, often resulting in suboptimal outcomes. In these reports, we detail clinical cases where patients with different mandibular defects underwent reconstructive surgery. In each instance, autologous grafts were harvested to ensure the restoration of native bone tissue, while advanced virtual planning techniques were employed for precise graft design and dental implant placement. The patients experienced substantial improvements in masticatory function, speech, and facial aesthetics. Utilizing autologous grafts minimized the risk of rejection and complications associated with foreign materials. The integration of virtual planning precision allowed customized solutions, reducing surgical duration and optimizing implant positioning. These 2 cases underscores the potential of combining autologous grafts with virtual planning precision and dental implants produced by additive manufacturing for mandible reconstruction.

A method for generating the digital twin models of plants

In the process of creating digital twin models of plants, the simulation accuracy of models generated by conventional 3D reconstruction methods is insufficient. Considering the increasing number of models over time, adopting deep learning-based models not only consume a significant amount of time but also require substantial computational resources consumption. Therefore, we proposed a method for generating digital twin models of plants, aiming to ensure the simulation credibility while effectively reducing resource costs. Experimental results show that this method is capable of generating plant models that closely resemble physical entities at different growth periods. Compared with other construction methods, this approach achieves higher simulation credibility (CD=0.089, EMD=0.034, KNN=0.005, RGB=4.97) with lower computational resources consumption (ROM=7.46MB, VRAM=2456MB, Time=0h5min48s) when generating a digital twin model.

Res-U2Net: untrained deep learning for phase retrieval and image reconstruction

Conventional deep learning-based image reconstruction methods require a large amount of training data, which can be hard to obtain in practice. Untrained deep learning methods overcome this limitation by training a network to invert a physical model of the image formation process. Here we present a novel, to our knowledge, untrained Res-U2Net model for phase retrieval. We use the extracted phase information to determine changes in an object’s surface and generate a mesh representation of its 3D structure. We compare the performance of Res-U2Net phase retrieval against UNet and U2Net using images from the GDXRAY dataset.

Single shot interferenceless coded aperture correlation holography via a learnable Wiener deconvolution network

One of the challenges in interferenceless coded aperture correlation holography (I-COACH) is high-quality single-shot reconstruction, as conventional cross-correlation reconstruction algorithm requires recording multiple holograms and complex reconstruction operations, resulting in serious background noise. Here, we present a deep learning based single-shot reconstruction by introducing a learnable Wiener deconvolution network in I-COACH. By feeding the point spread hologram (PSH) of I-COACH system and object hologram (OH) into Wiener deconvolution, a mapping relationship between Wiener deconvolution result and ground truth is generated. A Res-UNet is trained to learn optimal filter and noise regularization parameters, thereby achieving high quality reconstruction. Moreover, the training dataset used in our network are simulated using PSH that is obtained through only one experiment acquisition in advance, greatly simplifying experimental OH dataset acquisition. Compared to the conventional correlation reconstruction methods, the proposed Wiener deconvolution network can conveniently realize high-quality single-shot reconstruction in I-COACH, and the axial sensitivity in 3D reconstruction can be demonstrated by varying scattering degree of coded phase mask. Importantly, this Wiener deconvolution solution will provide an enlightening reference for dynamic imaging in I-COACH.

High-fidelity intravoxel incoherent motion parameter mapping using locally low-rank and subspace modeling

PurposeIntravoxel incoherent motion (IVIM) is a quantitative magnetic resonance imaging (MRI) method used to quantify perfusion properties of tissue non-invasively without contrast. However, clinical applications are limited by unreliable parameter estimates, particularly for the perfusion fraction (f) and pseudodiffusion coefficient (D*). This study aims to develop a high-fidelity reconstruction for reliable estimation of IVIM parameters. The proposed method is versatile and amenable to various acquisition schemes and fitting methods. MethodsTo address current challenges with IVIM, we adapted several advanced reconstruction techniques. We used a low-rank approximation of IVIM images and temporal subspace modeling to constrain the magnetization dynamics of the bi-exponential diffusion signal decay. In addition, motion-induced phase variations were corrected between diffusion directions and b-values, facilitating the use of high SNR real-valued diffusion data. The proposed method was evaluated in simulations and in vivo brain acquisitions in six healthy subjects and six individuals with a history of SARS-CoV-2 infection and compared with the conventionally reconstructed magnitude data. Following reconstruction, IVIM parameters were estimated voxel-wise. ResultsOur proposed method reduced noise contamination in simulations, resulting in a 60%, 58.9%, and 83.9% reduction in the NRMSE for D, f, and D*, respectively, compared to the conventional reconstruction. In vivo, anisotropic properties of D, f, and D* were preserved with the proposed method, highlighting microvascular differences in gray matter between individuals with a history of COVID-19 and those without (p = 0.0210), which wasn't observed with the conventional reconstruction. ConclusionThe proposed method yielded a more reliable estimation of IVIM parameters with less noise than the conventional reconstruction. Further, the proposed method preserved anisotropic properties of IVIM parameter estimates and demonstrated differences in microvascular perfusion in COVID-affected subjects, which weren't observed with conventional reconstruction methods.

A Novel Fixation Method of the Graft to the Fibular Head in Knee Lateral Collateral Ligament Reconstruction: Technical Note.

The lateral collateral ligament (LCL) is the strongest lateral stabilizer of the knee. It provides support against varus stress and posterolateral rotation of the knee. Lateral collateral ligament injuries mostly occur together with anterior and/or posterior cruciate ligament injuries. While grades 1 and 2 injuries are treated conservatively since they are partial injuries, total ruptures, as in grade 3, require surgical treatment. In conventional LCL reconstruction methods, hamstring grafts are used, and bioscrews are used in bone-tendon fixation. Lateral collateral ligament reconstruction is usually performed as a component of multiple ligament surgery. Therefore, there is a need for a contralateral hamstring tendon or allograft. The present article aims to define a technique that does not require tendon grafts and bioscrews in fibular fixation.

Reconstruction refinement of hybrid background-oriented schlieren tomography

This paper introduces a hybrid method that leverages the advantages of both the window-based separate algorithm and the regularization-based unifying algorithm to refine the three-dimensional refractive index field measured by the background-oriented schlieren (BOS) tomography. Two image-warping approaches, under paraxial and non-paraxial ray-tracing assumption, are developed as a connecting link between coarse and refined reconstruction. An orthogonal test is conducted to examine the influence of various factors on the reconstruction fields in the hybrid approach, providing insights into the impact of factors such as the ray-tracing assumption, initial value for refractive index iteration, and image displacement algorithm. Various synthetic and real BOS cases demonstrate that our hybrid method reconstructs a more detailed and robust field, outperforming conventional BOS reconstruction methods in scenarios with small or large density gradients, as well as laminar or turbulent structures.

An attenuation field network for dedicated cone beam breast CT with short scan and offset detector geometry

The feasibility of full-scan, offset-detector geometry cone-beam CT has been demonstrated for several clinical applications. For full-scan acquisition with offset-detector geometry, data redundancy from complementary views can be exploited during image reconstruction. Envisioning an upright breast CT system, we propose to acquire short-scan data in conjunction with offset-detector geometry. To tackle the resulting incomplete data, we have developed a self-supervised attenuation field network (AFN). AFN leverages the inherent redundancy of cone-beam CT data through coordinate-based representation and known imaging physics. A trained AFN can query attenuation coefficients using their respective coordinates or synthesize projection data including the missing projections. The AFN was evaluated using clinical cone-beam breast CT datasets (n = 50). While conventional analytical and iterative reconstruction methods failed to reconstruct the incomplete data, AFN reconstruction was not statistically different from the reference reconstruction obtained using full-scan, full-detector data in terms of image noise, image contrast, and the full width at half maximum of calcifications. This study indicates the feasibility of a simultaneous short-scan and offset-detector geometry for dedicated breast CT imaging. The proposed AFN technique can potentially be expanded to other cone-beam CT applications.

Evaluation of a deep learning magnetic resonance imaging reconstruction method for synthetic computed tomography generation in prostate radiotherapy

Background and PurposeIn magnetic resonance imaging (MRI) only radiotherapy computed tomography (CT) is excluded. The method relies entirely on synthetic CT images generated from MRI. This study evaluates the compatibility of a commercial synthetic CT (sCT) with an accelerated commercial deep learning reconstruction (DLR) in MRI-only prostate radiotherapy. Materials and MethodsFor a group of 24 patients (cohort 1) the effects of DLR were studied in isolation. MRI data were reconstructed conventionally and with DLR from identical k-space data, and sCTs were generated for both reconstructions. The sCT quality, Hounsfield Unit (HU) and dosimetric impact were investigated. In another group of 15 patients (cohort 2) effects on sCT generation using accelerated MRI acquisition (40 % time reduction) reconstructed with DLR were investigated. ResultssCT images from both cohorts, generated from DLR MRI data, were of clinically expected image quality. The mean dose differences for targets and organs at risks in cohort 1 were <0.06 Gy, corresponding to a 0.1 % prescribed dose difference. Similar dose differences were observed in cohort 2. Gamma pass rates for cohort 1 were 100 % for criteria 3 %/3mm, 2 %/2mm and 1 %/1mm for all dose levels. Mean error and mean absolute error inside the body, between sCTs, averaged over all cohort 1 subjects, were −1.1 ± 0.6 [−2.4 0.2] and 2.9 ± 0.4 [2.3 3.9] HU, respectively. ConclusionsDLR was suitable for sCT generation with clinically negligible differences in HU and calculated dose compared to the conventional MRI reconstruction method. For sCT generation DLR enables scan time reduction, without compromised sCT quality.

DOTnet 2.0: Deep learning network for diffuse optical tomography image reconstruction

Breast cancer is the most common cancer worldwide. The standard imaging modality for breast cancer screening is X-ray mammography, which suffers from low sensitivities in women with dense breasts and can potentially cause cancers despite a low radiation dosage. Diffuse Optical Tomography (DOT) is a noninvasive imaging technique that can potentially be employed to improve breast cancer early detection. However, conventional model-based algorithms for reconstructing DOT images usually produce low-quality images with limited resolution and low reconstruction accuracy. We propose to integrate deep neural networks (DNNs) with the conventional DOT reconstruction methods. This hybrid framework significantly enhances image quality. The DNNs have been trained and tested with sample data derived from clinically relevant breast models. The sample dataset contains blood vessel structures from breast structures and artificially created vessels using the Lindenmayer-system algorithm. By comparing the hybrid reconstruction with the ground truth image, we demonstrated a multi scale - structural similarity index measure (MS-SSIM) score of 0.80–0.90. Whereas using conventional reconstruction, MS-SSIM provided a much inferior score of 0.36–0.59. In terms of DOT image quality, both qualitative and quantitative assessments of the reconstructed images signify that the hybrid approach is superior to conventional methods. This improvement suggests that DOT can potentially become a viable alternative to breast cancer screening, providing a step towards the next-generation device for optical mammography.

MMV-Net: A Multiple Measurement Vector Network for Multifrequency Electrical Impedance Tomography.

Multifrequency electrical impedance tomography (mfEIT) is an emerging biomedical imaging modality to reveal frequency-dependent conductivity distributions in biomedical applications. Conventional model-based image reconstruction methods suffer from low spatial resolution, unconstrained frequency correlation, and high computational cost. Deep learning has been extensively applied in solving the EIT inverse problem in biomedical and industrial process imaging. However, most existing learning-based approaches deal with the single-frequency setup, which is inefficient and ineffective when extended to the multifrequency setup. This article presents a multiple measurement vector (MMV) model-based learning algorithm named MMV-Net to solve the mfEIT image reconstruction problem. MMV-Net considers the correlations between mfEIT images and unfolds the update steps of the Alternating Direction Method of Multipliers for the MMV problem (MMV-ADMM). The nonlinear shrinkage operator associated with the weighted l2,1 regularization term of MMV-ADMM is generalized in MMV-Net with a cascade of a Spatial Self-Attention module and a Convolutional Long Short-Term Memory (ConvLSTM) module to better capture intrafrequency and interfrequency dependencies. The proposed MMV-Net was validated on our Edinburgh mfEIT Dataset and a series of comprehensive experiments. The results show superior image quality, convergence performance, noise robustness, and computational efficiency against the conventional MMV-ADMM and the state-of-the-art deep learning methods.

Early experience of direct-to-implant breast reconstruction using acellular dermal matrix after robot-assisted nipple-sparing mastectomy.

This study aimed to analyze the novel operative outcomes of breast mound reconstruction followed by robot-assisted mastectomy in retrospective review. Patients who underwent nipple-sparing mastectomy with a robotic device (Da Vinci Xi) and immediate prosthetic reconstruction prepectorally via lateral incision from June 2018 to July 2019 were enrolled. Patient characteristics, complications, and satisfaction via BREAST-Q were analyzed. The surgical technique was described in detail. Thirty-nine cases, including 7 bilateral cases (total 46 breasts), underwent robot-assisted nipple-sparing mastectomy followed by immediate prosthetic implant reconstruction prepectorally. The median patient age was 46.63 years (range: 21-63 years). The mean operation time for each prepectoral breast mound reconstruction using the direct-to-implant technique was 126.55 min. Overall satisfaction of robotic use was evaluated as superior to the conventional reconstruction method using BREAST-Q. Major infection was found in seven cases (15.2%), and complete nipple loss was found in three cases (6.6%). Severe complications requiring breast implant removal in the surgical technique occurred in four breasts (8.7%). Two cases were due to the coexistence of infection and skin necrosis; in one case, the skin flap had undergone the congestive phase on postoperative day (POD) 3 and required additional surgery to change the expander. Other complications were resolved by conservative care or minor revision. This report is the first concerning robot-assisted NSM followed by prepectoral ADM-wrapped prosthetic reconstruction. In our experience, this procedure seems to be not inferior to other methods. Further prospective research to evaluate oncologic outcomes is warranted.

Sparse-view reconstruction for photoacoustic tomography combining diffusion model with model-based iteration

As a non-invasive hybrid biomedical imaging technology, photoacoustic tomography combines high contrast of optical imaging and high penetration of acoustic imaging. However, the conventional standard reconstruction under sparse view could result in low-quality image in photoacoustic tomography. Here, a novel model-based sparse reconstruction method for photoacoustic tomography via diffusion model was proposed. A score-based diffusion model is designed for learning the prior information of the data distribution. The learned prior information is utilized as a constraint for the data consistency term of an optimization problem based on the least-square method in the model-based iterative reconstruction, aiming to achieve the optimal solution. Blood vessels simulation data and the animal in vivo experimental data were used to evaluate the performance of the proposed method. The results demonstrate that the proposed method achieves higher-quality sparse reconstruction compared with conventional reconstruction methods and U-Net. In particular, under the extreme sparse projection (e.g., 32 projections), the proposed method achieves an improvement of ∼ 260 % in structural similarity and ∼ 30 % in peak signal-to-noise ratio for in vivo data, compared with the conventional delay-and-sum method. This method has the potential to reduce the acquisition time and cost of photoacoustic tomography, which will further expand the application range.

Quantitative 3D Characterization of Functionally Relevant Parameters in Heavy-Oxide-Supported 4d Metal Nanocatalysts.

Accurate 3D nanometrology of catalysts with small nanometer-sized particles of light 3d or 4d metals supported on high-atomic-number oxides is crucial for understanding their functionality. However, performing quantitative 3D electron tomography analysis on systems involving metals like Pd, Ru, or Rh supported on heavy oxides (e.g., CeO2) poses significant challenges. The low atomic number (Z) of the metal complicates discrimination, especially for very small nanoparticles (1-3 nm). Conventional reconstruction methods successful for catalysts with 5d metals (e.g., Au, Pt, or Ir) fail to detect 4d metal particles in electron tomography reconstructions, as their contrasts cannot be effectively separated from those of the underlying support crystallites. To address this complex 3D characterization challenge, we have developed a full deep learning (DL) pipeline that combines multiple neural networks, each one optimized for a specific image-processing task. In particular, single-image super-resolution (SR) techniques are used to intelligently denoise and enhance the quality of the tomographic tilt series. U-net generative adversarial network algorithms are employed for image restoration and correcting alignment-related artifacts in the tilt series. Finally, semantic segmentation, utilizing a U-net-based convolutional neural network, splits the 3D volumes into their components (metal and support). This approach enables the visualization of subnanometer-sized 4d metal particles and allows for the quantitative extraction of catalytically relevant structural information, such as particle size, sphericity, and truncation, from compressed sensing electron tomography volume reconstructions. We demonstrate the potential of this approach by characterizing nanoparticles of a metal widely used in catalysis, Pd (Z = 46), supported on CeO2, a very high density (7.22 g/cm3) oxide involving a quite high-atomic-number element, Ce (Z = 58).

Deep Gaussian Scale Mixture Prior for Image Reconstruction.

Image reconstruction from partial observations has attracted increasing attention. Conventional image reconstruction methods with hand-crafted priors often fail to recover fine image details due to the poor representation capability of the hand-crafted priors. Deep learning methods attack this problem by directly learning mapping functions between the observations and the targeted images can achieve much better results. However, most powerful deep networks lack transparency and are nontrivial to design heuristically. This paper proposes a novel image reconstruction method based on the Maximum a Posterior (MAP) estimation framework using learned Gaussian Scale Mixture (GSM) prior. Unlike existing unfolding methods that only estimate the image means (i.e., the denoising prior) but neglected the variances, we propose characterizing images by the GSM models with learned means and variances through a deep network. Furthermore, to learn the long-range dependencies of images, we develop an enhanced variant based on the Swin Transformer for learning GSM models. All parameters of the MAP estimator and the deep network are jointly optimized through end-to-end training. Extensive simulation and real data experimental results on spectral compressive imaging and image super-resolution demonstrate that the proposed method outperforms existing state-of-the-art methods.

DIRECTIONAL SPECTRA OF INFRAGRAVITY WAVES DURING STORMY CONDITIONS

Both magnitude and directions of infragravity (IG) waves are essential information for accurate hindcast/prediction of coastal hazard due to stormy waves (e.g., coastal inundation and beach erosion). However, conventional reconstruction methods of directional spectra relying on linear wave theory are not applicable to IG waves which are composed of free (linear) and bound (nonlinear) components. This study proposes a novel method to reconstruct directional spectra of IG waves and presents diverse wave directions of free IG waves during stormy conditions.

FDU-Net: Deep Learning-Based Three-Dimensional Diffuse Optical Image Reconstruction.

Near-infrared diffuse optical tomography (DOT) is a promising functional modality for breast cancer imaging; however, the clinical translation of DOT is hampered by technical limitations. Specifically, conventional finite element method (FEM)-based optical image reconstruction approaches are time-consuming and ineffective in recovering full lesion contrast. To address this, we developed a deep learning-based reconstruction model (FDU-Net) comprised of a Fully connected subnet, followed by a convolutional encoder-Decoder subnet, and a U-Net for fast, end-to-end 3D DOT image reconstruction. The FDU-Net was trained on digital phantoms that include randomly located singular spherical inclusions of various sizes and contrasts. Reconstruction performance was evaluated in 400 simulated cases with realistic noise profiles for the FDU-Net and conventional FEM approaches. Our results show that the overall quality of images reconstructed by FDU-Net is significantly improved compared to FEM-based methods and a previously proposed deep-learning network. Importantly, once trained, FDU-Net demonstrates substantially better capability to recover true inclusion contrast and location without using any inclusion information during reconstruction. The model was also generalizable to multi-focal and irregularly shaped inclusions unseen during training. Finally, FDU-Net, trained on simulated data, could successfully reconstruct a breast tumor from a real patient measurement. Overall, our deep learning-based approach demonstrates marked superiority over the conventional DOT image reconstruction methods while also offering over four orders of magnitude acceleration in computational time. Once adapted to the clinical breast imaging workflow, FDU-Net has the potential to provide real-time accurate lesion characterization by DOT to assist the clinical diagnosis and management of breast cancer.

Bilateral extended FDK: An improved weighting method for static CT imaging.

Improving imaging speed has always been the focus of research in CT technology, which is related to the radiation dose and imaging quality of moving organs, including heart and blood vessels. However, it is difficult to achieve further improvement by increasing the rotation speed of the gantry due to its structural strength limitation. Differing from the conventional CTs, the static CT employs dozens of ray sources to acquire projection data from different angular ranges, and each source only needs to be rotated in a small range to finish a full 360° scan, thus greatly increasing the scanningspeed. As sources of static CT need to be evenly distributed over 360°, the sources and detectors have to be arranged on two parallel rings independently. Such a geometry can be considered as a special case of CT systems with a significantly large cone angle, that is, a part of the detector is missing in the vicinity of the mid-plane. Due to restriction of upper and lower bounds of the cone angle of the static CT, there are uneven projection data varying in each portion of the reconstruction volume, the conventional analytical or iterative reconstruction methods may introduce artifacts in the reconstructedoutcomes. Following the weighting approach extended FDK (xFDK) by Grimmer etal., we propose an improved bilateral xFDK algorithm (bixFDK), which focuses on the reconstruction of the expanded volume. With the same philosophy as xFDK in terms of weighting function design, bixFDK takes the longitudinal offset of the detector with respect to the source into consideration, making our method applicable to a wide range of CT geometries, especially for the static CT. Based on the proposed bixFDK, a new iterative scheme bixFDK-IR is also constructed to extend the applications to a wide range of scan protocols such as sparse-viewscan. The proposed method has been validated with the simulated phantom data and the actual clinical data of the static CT, and demonstrates that it can ensure good image quality and enlarge the reconstruction volume in z-direction of the staticCT. The bixFDK algorithm is an ideal reconstruction approach for static CT geometry, and the iterative scheme of bixFDK-IR is applicable to a wide range of CT geometries and scan protocols, thus providing a wide range of applicationscenarios.

Structural Gaussian priors for Bayesian CT reconstruction of subsea pipes

A non-destructive testing (NDT) application of X-ray computed tomography (CT) is inspection of subsea pipes in operation via 2D cross-sectional scans. Data acquisition is time-consuming and costly due to the challenging subsea environment. While reducing the number of projections in a CT scan yields time and cost savings, this compromises the reconstruction quality when conventional reconstruction methods are used. To address this issue we take a Bayesian approach to CT reconstruction and focus on designing an effective prior that introduces additional information to the problem. We propose a new class of structural Gaussian priors to enforce expected material properties in different regions of the reconstructed image based on available a-priori information about the pipe's layered structure. The prior is composed from Gaussian distributions, and we utilize this to propose an efficient computational strategy to sample the resulting posterior distribution. This is essential in practical NDT applications for fast processing of the large-scale data we see in CT. Numerical experiments with synthetic and real data show that the proposed structural Gaussian prior reduces artifacts and enhances contrast in the reconstruction, compared to using only a conventional Gaussian Markov random field prior or no prior information at all. Furthermore, we obtain uncertainty estimates that indicate the certainty of the reconstructions increase when more prior information is added.

NERF FOR HERITAGE 3D RECONSTRUCTION

Abstract. Conventional or learning-based 3D reconstruction methods from images have clearly shown their potential for 3D heritage documentation. Nevertheless, Neural Radiance Field (NeRF) approaches are recently revolutionising the way a scene can be rendered or reconstructed in 3D from a set of oriented images. Therefore the paper wants to review some of the last NeRF methods applied to various cultural heritage datasets collected with smartphone videos, touristic approaches or reflex cameras. Firstly several NeRF methods are evaluated. It turned out that Instant-NGP and Nerfacto methods achieved the best outcomes, outperforming all other methods significantly. Successively qualitative and quantitative analyses are performed on various datasets, revealing the good performances of NeRF methods, in particular for areas with uniform texture or shining surfaces, as well as for small datasets of lost artefacts. This is for sure opening new frontiers for 3D documentation, visualization and communication purposes of digital heritage.