Search

Training medical professionals for hand hygiene is challenging, especially due to the invisibility of microorganisms to the human eye. As the use of virtual reality (VR) in medical training is still novel, this exploratory study investigated how preexisting technology acceptance and in-training engagement predict VR hand hygiene performance scores. The effect of training in the VR environment on the behavioral intention to further use this type of training device (a component of technology acceptance) was also investigated. Participants completed a VR hand hygiene training comprising three levels of the same task with increasing difficulty. We measured technology acceptance, composed of performance expectancy, effort expectancy, and behavioral intention, pre- and post-training, and in-training engagement using adaptations of existing questionnaires. We used linear regression models to determine predictors of performance in level-3 and of behavioral intention to further use VR training. Forty-three medical students participated in this exploratory study. In-training performance significantly increased between level-1 and level-3. Performance in level-3 was predicted by prior performance expectancy and engagement during the training session. Intention to further use VR to learn medical procedures was predicted by both prior effort expectancy and engagement. Our results provide clarification on the relationship between VR training, engagement, and technology acceptance. Future research should assess the long-term effectiveness of hand hygiene VR training and the transferability of VR training to actual patient care in natural settings. A more complete VR training could also be developed, with additional levels including more increased difficulty and additional medical tasks.

We push the boundaries of electronic structure-based ab-initio molecular dynamics (AIMD) beyond 100 million atoms. This scale is otherwise barely reachable with classical force-field methods or novel neural network and machine learning potentials. We achieve this breakthrough by combining innovations in linear-scaling AIMD, efficient and approximate sparse linear algebra, low and mixed-precision floating-point computation on GPUs, and a compensation scheme for the errors introduced by numerical approximations. The core of our work is the non-orthogonalized local submatrix method (NOLSM), which scales very favorably to massively parallel computing systems and translates large sparse matrix operations into highly parallel, dense matrix operations that are ideally suited to hardware accelerators. We demonstrate that the NOLSM method, which is at the center point of each AIMD step, is able to achieve a sustained performance of 324 PFLOP/s in mixed FP16/FP32 precision corresponding to an efficiency of 67.7% when running on 1536 NVIDIA A100 GPUs., Parallel Computing, 111, ISSN:0167-8191, ISSN:1872-7336

Soft-tissue lesions in the oral cavity, one of the most common sites for tumors and tumor-like lesions, can be challenging to diagnose and treat due to the wide spectrum from benign indolent to invasive malignant lesions. We report an abnormally large, rapidly growing hyperplastic lesion originating from the buccal mucosa in a 28-year-old male patient. Clinical examination revealed a well-circumscribed, smooth-surfaced, pinkish nodular lesion measuring 2.3 × 2 cm, which suggested the differential diagnosis of irritation fibroma, pyogenic granuloma, oral lipoma, and other benign or malignant neoplasms such as hemangioma, non-Hodgkin’s lymphoma, or metastases to the oral cavity. Dental MRI using a 15-channel mandibular coil was performed to improve perioperative radiological and surgical management, avoiding adverse intraoperative events and misdiagnosis of vascular malformations, especially hemangiomas. Black bone MRI protocols such as STIR (short-tau inversion recovery) and DESS (double-echo steady-state) were used for high-resolution radiation-free imaging. Radiologic findings supported the suspected diagnosis of an irritation fibroma and ruled out any further head and neck lesions, therefore complete surgical resection was performed. Histology confirmed the tentative diagnosis. This article evaluates the use of this novel technique for MR diagnosis in the perioperative management of soft-tissue tumors in oral and maxillofacial surgery.

Spiking Neural Networks (SNNs) are developed as a promising alternative to Artificial Neural networks (ANNs) due to their more realistic brain-inspired computing models. SNNs have sparse neuron firing over time, i.e., spatio-temporal sparsity; thus, they are useful to enable energy-efficient hardware inference. However, exploiting spatio-temporal sparsity of SNNs in hardware leads to unpredictable and unbalanced workloads, degrading the energy efficiency. In this work, we propose an FPGA-based convolutional SNN accelerator called Skydiver that exploits spatio-temporal workload balance. We propose the Approximate Proportional Relation Construction (APRC) method that can predict the relative workload channel-wisely and a Channel-Balanced Workload Schedule (CBWS) method to increase the hardware workload balance ratio to over 90%. Skydiver was implemented on a Xilinx XC7Z045 FPGA and verified on image segmentation and MNIST classification tasks. Results show improved throughput by 1.4X and 1.2X for the two tasks. Skydiver achieved 22.6 KFPS throughput, and 42.4 uJ/Image prediction energy on the classification task with 98.5% accuracy., Accepted to be published in the IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems, 2022

We consider uniform random cographs (either labeled or unlabeled) of large size. Our first main result is the convergence towards a Brownian limiting object in the space of graphons. We then show that the degree of a uniform random vertex in a uniform cograph is of order $n$, and converges after normalization to the Lebesgue measure on $[0,1]$. We finally analyze the vertex connectivity (i.e. the minimal number of vertices whose removal disconnects the graph) of random connected cographs, and show that this statistics converges in distribution without renormalization. Unlike for the graphon limit and for the degree of a random vertex, the limiting distribution is different in the labeled and unlabeled settings. Our proofs rely on the classical encoding of cographs via cotrees. We then use mainly combinatorial arguments, including the symbolic method and singularity analysis., 36 pages, 6 figures; v3 includes a new reference to Diaconis-Holmes-Janson (arXiv:0908.2448) for the continuity of the degree distribution for graphons

Injury to the peripheral branches of the trigeminal nerve, particularly the lingual nerve (LN) and the inferior alveolar nerve (IAN), is a rare but serious complication that can occur during oral and maxillofacial surgery. Mandibular third molar surgery, one of the most common surgical procedures in dentistry, is most often associated with such a nerve injury. Proper preoperative radiologic assessment is hence key to avoiding neurosensory dysfunction. In addition to the well-established conventional X-ray-based imaging modalities, such as panoramic radiography and cone-beam computed tomography, radiation-free magnetic resonance imaging (MRI) with the recently introduced black-bone MRI sequences offers the possibility to simultaneously visualize osseous structures and neural tissue in the oral cavity with high spatial resolution and excellent soft-tissue contrast. Fortunately, most LN and IAN injuries recover spontaneously within six months. However, permanent damage may cause significant loss of quality of life for affected patients. Therefore, therapy should be initiated early in indicated cases, despite the inconsistency in the literature regarding the therapeutic time window. In this report, we present the visualization of two cases of nerve pathology using 3D double-echo steady-state MRI and evaluate evidence-based decision-making for iatrogenic nerve injury regarding a wait-and-see strategy, conservative drug treatment, or surgical re-intervention.

Agent based modelling is a simulation method in which autonomous agents interact with their environment and one another, given a predefined set of rules. It is an integral method for modelling and simulating complex systems, such as socio-economic problems. Since agent based models are not described by simple and concise mathematical equations, code that generates them is typically complicated, large, and slow. Here we present Agents.jl, a Julia-based software that provides an ABM analysis platform with minimal code complexity. We compare our software with some of the most popular ABM software in other programming languages. We find that Agents.jl is not only the most performant, but also the least complicated software, providing the same (and sometimes more) features as the competitors with less input required from the user. Agents.jl also integrates excellently with the entire Julia ecosystem, including interactive applications, differential equations, parameter optimization, and more. This removes any ``extensions library'' requirement from Agents.jl, which is paramount in many other tools., Comment: SIMULATION. January 2022

Additive manufacturing requires support structures to fabricate parts with overhangs. In this paper, we revisit a known support structure based on bridges-and-pillars (see Figure 1). The support structures are made of vertical pillars supporting horizontal bridges. Their scaffolding structure makes them stable and reliable to print. However, the algorithm heuristic search does not scale well and is prone to produce contacts with the parts, leaving scars after removal. We propose a novel algorithm for this type of supports, focusing on avoiding unnecessary contacts with the part as much as possible. Our approach builds upon example-based model synthesis to enable early detection of collision-free passages as well as non-reachable regions., Procedural Modelling, Marco Freire, Samuel Hornus, Salim Perchy, and Sylvain Lefebvre

Voluntary and involuntary patient motion is a major problem for data quality in clinical routine of Magnetic Resonance Imaging (MRI). It has been thoroughly investigated and, yet it still remains unresolved. In quantitative MRI, motion artifacts impair the entire temporal evolution of the magnetization and cause errors in parameter estimation. Here, we present a novel strategy based on residual learning for retrospective motion correction in fast 3D whole-brain multiparametric MRI. We propose a 3D multiscale convolutional neural network (CNN) that learns the non-linear relationship between the motion-affected quantitative parameter maps and the residual error to their motion-free reference. For supervised model training, despite limited data availability, we propose a physics-informed simulation to generate self-contained paired datasets from a priori motion-free data. We evaluate motion-correction performance of the proposed method for the example of 3D Quantitative Transient-state Imaging at 1.5T and 3T. We show the robustness of the motion correction for various motion regimes and demonstrate the generalization capabilities of the residual CNN in terms of real-motion in vivo data of healthy volunteers and clinical patient cases, including pediatric and adult patients with large brain lesions. Our study demonstrates that the proposed motion correction outperforms current state of the art, reliably providing a high, clinically relevant image quality for mild to pronounced patient movements. This has important implications in clinical setups where large amounts of motion affected data must be discarded as they are rendered diagnostically unusable.

Image classification based on neural networks has been widely explored in machine learning and most research have focused on developing more efficient and accurate network models for given image dataset mostly over natural scene. However, industrial image data have different features with natural scene images in shape of target objects, background patterns, and color. Additionally, data imbalance is one of the most challenging problems to degrade classification accuracy for industrial images. This paper proposes a novel GAN-based image generation method to improve classification accuracy for defect images of OLED panels. We validate our method can synthetically generate defect images of OLED panels and classification accuracy can be improved by training minor classes with the generated defect images., Industry Track Papers, Yongmoon Jeon, Haneol Kim, Hyeona Lee, Seonghoon Jo, and Jaewon Kim

In the context of Laser Metal Deposition (LMD), temporary support structures are needed to manufacture overhanging features. In order to limit the need for supports, multi-axis machines intervene in the deposition by sequentially repositioning the part. Under multi-axis rotations and translations, slicing and toolpath generation represent significant challenges. Slicing has been partially addressed by authors in multi-axis LMD. However, tool-path generation in multi-axis LMD is rarely touched. One of the reasons is that the required slices for LMD may be strongly non-developable. This fact produces a significant mismatch between the tool-path speeds and other parameters in Parametric space vs. actual Euclidean space. For the particular case of developable slices present in workpieces with cylindrical kernel and overhanging neighborhoods, this manuscript presents a methodology for LMD tool path generation. Our algorithm takes advantage of existing cylindrical iso-radial slicing by generating a path in the (?, z) parameter space and isometrically translating it into the R3 Euclidean space. The presented approach is advantageous because it allows the path-planning of complex structures by using the methods for conventional 2.5-axis AM. Our computer experiments show that the presented approach can be effectively used in manufacturing industrial/mechanical pieces (e.g., spur gears). Future work includes the generation of the machine g-code for actual LMD equipment., CAD, Simulation, and Grasping, Diego Montoya-Zapata, Aitor Moreno, Igor Ortiz, Oscar Ruiz-Salguero, and Jorge Posada

The exploration of large-scale conflicts, as well as their causes and effects, is an important aspect of socio-political analysis. Since event data related to major conflicts are usually obtained from different sources, researchers developed a semi-automatic matching algorithm to integrate event data of different origins into one comprehensive dataset using hierarchical taxonomies. The validity of the corresponding integration results is not easy to assess since the results depend on user-defined input parameters and the relationships between the original data sources. However, only rudimentary visualization techniques have been used so far to analyze the results, allowing no trustworthy validation or exploration of how the final dataset is composed. To overcome this problem, we developed VEHICLE, a web-based tool to validate and explore the results of the hierarchical integration. For the design, we collaborated with a domain expert to identify the underlying domain problems and derive a task and workflow description. The tool combines both traditional and novel visual analysis techniques, employing statistical and map-based depictions as well as advanced interaction techniques. We showed the usefulness of VEHICLE in two case studies and by conducting an evaluation together with conflict researchers, confirming domain hypotheses and generating new insights.

When conveying information about routes to follow in complex environments, human route-givers adapt to route-receivers’ familiarity with the environments in their choice of landmarks. Meanwhile, as route-givers themselves have experienced the environments within a social role, the landmarks they select may also differ significantly. This research investigated how these two factors influence landmark selection when communicating routes in indoor environments. Two groups of participants were recruited to conduct indoor landmark selection experiments for familiar and unfamiliar route-receivers in a multi-functional university building. The results show an interaction effect between these factors in indoor landmark selection. These findings lay an empirical ground for developing human-centered mobile navigation systems that can adapt to users’ social roles and their familiarity with the environment.

Until recently, computer simulation of interior decoration was reserved to architects and other highly skilled professionals, because existing tools require technical knowledge on computer aided design. During the last years, several easy-to-use tools for home planning and interior design have been released. These tools intend to allow users, without specific skills, to create their virtual house, decorate it and navigate in the resulting scenery. In this paper we present three of these applications and describe task analysis using them. As a result of this experiment, we identify user requirements for this kind of tools, along with positive and negative aspects, providing good insights and directions for future work., Interac����o 2006 - 2�� Confer��ncia Nacional em Interac����o Pessoa-M��quina, Posters e Demos, 225, 228, Alfredo Ferreira, Roberto Medeiros, Rui Barreto, Manuel J. Fonseca, and Joaquim A. Jorge, Computer-Aided Design, Sketch-Based Interfaces, User Interfaces, Task Analysis, Evaluation

In this paper, we present a work-in-progress aimed at developing a pipeline for the fabrication of shapes reproducing digital models with a combination of standard LEGO bricks and 3D printed custom elements. The pipeline starts searching for the ideal alignment of the 3D model with the brick grid. It then employs a novel approach for shape "legolization" using a outside-in heuristic to limit critical configuration, and separates an external shell and an internal part. Finally, it exploits shape booleans to create the external custom parts to be 3D printed., Smart Tools and Apps for Graphics - Eurographics Italian Chapter Conference, Short Papers 2: Miscellanea, 175, 179, Filippo Andrea Fanni, Alberto Dal Bello, Simone Sbardellini, and Andrea Giachetti, CCS Concepts: Computing methodologies --> Shape modeling; Applied computing --> Computer-aided design

We present a parametric physics-informed neural network for the simulation of personalised left-ventricular biomechanics. The neural network is constrained to the biophysical problem in two ways: (i) the network output is restricted to a subspace built from radial basis functions capturing characteristic deformations of left ventricles and (ii) the cost function used for training is the energy potential functional specifically tailored for hyperelastic, anisotropic, nearly-incompressible active materials. The radial bases are generated from the results of a nonlinear Finite Element model coupled with an anatomical shape model derived from high-resolution cardiac images. We show that, by coupling the neural network with a simplified circulation model, we can efficiently generate computationally inexpensive estimations of cardiac mechanics. Our model is 30 times faster than the reference Finite Element model used, including training time, while yielding satisfactory average errors in the predictions of ejection fraction (-3%), peak systolic pressure (7%), stroke work (4%) and myocardial strains (14%). This physics-informed neural network is well suited to efficiently augment cardiac images with functional data and to generate large sets of synthetic cases for training deep network classifiers while it provides efficient personalization to the specific patient of interest with a high level of detail., Medical Image Analysis, 71, ISSN:1361-8415, ISSN:1361-8423

The first part of the course is theoretical, and introduces the finite element method trough interactive Jupyter notebooks. It also covers recent advancements toward an integrated pipeline, considering meshing and element design as a single challenge, leading to a black box pipeline that can solve simulations on ten thousand in the wild meshes, without any parameter tuning. In the second part we will move to practice, introducing a set of easy-to-use Python packages for applications in geometric computing. The presentation will have the form of live coding in a Jupyter notebook. We have designed the presented libraries to have a shallow learning curve, while also enabling programmers to easily accomplish a wide variety of complex tasks. Furthermore, these libraries utilize NumPy arrays as a common interface, making them highly composable with each-other as well as existing scientific computing packages. Finally, our libraries are blazing fast, doing most of the heavy computations in C++ with a minimal constant-overhead interface to Python. In the course, we will present a set of real-world examples from geometry processing, physical simulation, and geometric deep learning. Each example is prototypical of a common task in research or industry and is implemented in a few lines of code. By the end of the course, attendees will have exposure to a swiss-army-knife of simple, composable, and high-performance tools for geometric computing., Eurographics 2020 - Tutorials, Tutorials, 1, 4, Sebastian Koch, Teseo Schneider, Chengchen Li, and Daniele Panozzo, CCS Concepts: Mathematics of computing --> Numerical analysis; Theory of computation --> Computational geometry; Computing methodologies --> Machine learning; Scientific visualization; Mesh models; Shape analysis; Applied computing --> Computer-aided design

The intravoxel incoherent motion (IVIM) model allows to map diffusion (D) and perfusion-related parameters (F and D*). Parameter estimation is, however, error-prone due to the non-linearity of the signal model, the limited signal-to-noise ratio (SNR) and the small volume fraction of perfusion in the in-vivo brain. In the present work, the performance of Bayesian inference was examined in the presence of brain pathologies characterized by hypo- and hyperperfusion. In particular, a hierarchical and a spatial prior were combined. Performance was compared relative to conventional segmented least squares regression, hierarchical prior only (non-segmented and segmented data likelihoods) and a deep learning approach. Realistic numerical brain IVIM simulations were conducted to assess errors relative to ground truth. In-vivo, data of 11 central nervous system cancer patients and 9 patients with acute stroke were acquired. The proposed method yielded reduced error in simulations for both the cancer and acute stroke scenarios compared to other methods across the whole investigated SNR range. The contrast-to-noise ratio of the proposed method was better or on par compared to the other techniques in-vivo. The proposed Bayesian approach hence improves IVIM parameter estimation in brain cancer and acute stroke., Medical Image Analysis, 73, ISSN:1361-8415, ISSN:1361-8423

In the context of Computer-Aided Design and Manufacturing, the problem of feature recognition plays a key role in the integration of systems. Until now, compromises have been reached by only using FACE-based geometric information of prismatic CAD models to prune the search domain. This manuscripts presents a feature recognition method which more aggressively prunes the search space with reconfigurable geometric tests. This reconfigurable approach allows to enforce arbitrary confluent tests which are topologic and geometric, with enlarged domain. The test sequence is itself a graph (i.e. not a linear or total-order sequence). Unlike the existing methods which are FACE-based, the present one permits combinations of topologies whose dimensions are 2, 1 or 0. This system has been implemented in an industrial environment. The industrial incarnation allows industry-based customization and is faster when compared to topology-based feature recognition. Future work is required in improving robustness of search conditions and improving the graphic input interface., Spanish Computer Graphics Conference (CEIG), Short Papers, 69, 72, Juan Camilo Pareja Corcho, Oscar Mauricio Betancur Acosta, Oscar E. Ruiz, and Carlos Cadavid, CCS Concepts: Applied computing--> Computer-aided design; Computer-aided manufacturing

Following commonly used reverse engineering techniques, it is very difficult to reconstruct editable CAD parts or assemblies that can later be used and modified in the Product Development Process (PDP). Traditional methods follow a sequential time-consuming patch-by-patch reconstruction strategy with cumbersome procedures in which designers usually have to face many issues (e.g. decomposition in patches, trimming and connection of the patches), and generally producing ''dead'' models that cannot be later modified as needed. This paper describes a new reverse engineering technique that allows fitting of parametric CAD parts or an assembly to a reference point cloud to be reconstructed. The proposed fitting method can also be applied to 2D configurations to adapt a parametric 2D sketch to a 2D point cloud section. The idea is to use a parametric CAD model or a parametric 2D sketch in an optimization algorithm allowing their perfect fitting into the point cloud of a scanned mechanical assembly for efficient reconstruction of good quality CAD models. Some well-known algorithms like ICP are also used to derive the orientation and position of pre-arranged CAD model or 2D sketch throughout the fitting process. Both global and local fittings are possible. The consistency of the CAD models is ensured by a modeler which updates the CAD models or 2D sketch according to the iterative dimensional modifications. The evaluation of the proposed approach is performed using as-scanned virtually generated point clouds which incorporate several artifacts that would appear using a real scanner. This technique allows for the comparison between the dimensions of the fitted parts and the ones of the parts used to generate the point clouds., Eurographics 2019 - Posters, Posters, 11, 12, Ghazanfar Ali Shah, Franca Giannini, Marina Monti, Arnaud Polette, and Jean-Philippe Pernot, Keywords: Shape modeling; Engineering Computer-aided design

Point-cloud-to-mesh registration estimates a rigid transformation that minimizes the distance between a point sample of a surface and a reference mesh of such a surface, both lying in different coordinate systems. Point-cloud-to-mesh-registration is an ubiquitous problem in medical imaging, CAD CAM CAE, reverse engineering, virtual reality and many other disciplines. Common registration methods include Iterative Closest Point (ICP), RANdom SAmple Consensus (RANSAC) and Normal Distribution Transform (NDT). These methods require to repeatedly estimate the distance between a point cloud and a mesh, which becomes computationally expensive as the point set sizes increase. To overcome this problem, this article presents the implementation of a Perfect Spatial Hashing for point-cloud-to-mesh registration. The complexity of the registration algorithm using Perfect Spatial Hashing is O(NYxn) (NY : point cloud size, n: number of max. ICP iterations), compared to standard octrees and kd-trees (time complexity O(NY log(NT)xn), NT : reference mesh size). The cost of pre-processing is O(NT +(N3H )2) (N3H : Hash table size). The test results show convergence of the algorithm (error below 7e-05) for massive point clouds / reference meshes (NY = 50k and NT = 28055k, respectively). Future work includes GPU implementation of the algorithm for fast registration of massive point clouds., Spanish Computer Graphics Conference (CEIG), Full Papers, 41, 50, Daniel Mejia-Parra, Juan Lalinde-Pulido, Jairo R. Sánchez, Oscar Ruiz-Salguero, and Jorge Posada, CCS Concepts: Theory of computation--> Convex optimization; Computational geometry; Computing methodologies--> Mesh models; Point-based models; Applied computing--> Computer-aided design

We propose a dictionary-matching-free pipeline for multi-parametric quantitative MRI image computing. Our approach has two stages based on compressed sensing reconstruction and deep learned quantitative inference. The reconstruction phase is convex and incorporates efficient spatiotemporal regularisations within an accelerated iterative shrinkage algorithm. This minimises the under-sampling (aliasing) artefacts from aggressively short scan times. The learned quantitative inference phase is purely trained on physical simulations (Bloch equations) that are flexible for producing rich training samples. We propose a deep and compact encoder-decoder network with residual blocks in order to embed Bloch manifold projections through multi-scale piecewise affine approximations, and to replace the non-scalable dictionary-matching baseline. Tested on a number of datasets we demonstrate effectiveness of the proposed scheme for recovering accurate and consistent quantitative information from novel and aggressively subsampled 2D/3D quantitative MRI acquisition protocols.

The quantitative analysis of images acquired in the diagnosis and treatment of patients with brain tumors has seen a significant rise in the clinical use of computational tools. The underlying technology to the vast majority of these tools are machine learning methods and, in particular, deep learning algorithms. This review offers clinical background information of key diagnostic biomarkers in the diagnosis of glioma, the most common primary brain tumor. It offers an overview of publicly available resources and datasets for developing new computational tools and image biomarkers, with emphasis on those related to the Multimodal Brain Tumor Segmentation (BraTS) Challenge. We further offer an overview of the state-of-the-art methods in glioma image segmentation, again with an emphasis on publicly available tools and deep learning algorithms that emerged in the context of the BraTS challenge.

International audience; In this paper, we propose a novel continuum finite strain formulation of the equilibrium gap regularization for image registration. The equilibrium gap regularization essentially penalizes any deviation from the solution of a hyperelastic body in equilibrium with arbitrary loads prescribed at the boundary. It thus represents a regularization with strong mechanical basis, especially suited for cardiac image analysis. We describe the consistent linearization and discretization of the regularized image registration problem, in the framework of the finite elements method. The method is implemented using FEniCS & VTK, and distributed as a freely available python library. We show that the equilibrated warping method is effective and robust: regularization strength and image noise have minimal impact on motion tracking, especially when compared to strain-based regularization methods such as hyperelastic warping. We also show that equilibrated warping is able to extract main deformation features on both tagged and untagged cardiac magnetic resonance images.

Virtual Reality (VR) simulations have long been proposed to allow users to explore both yet-to-built buildings in architectural design, and ancient, remote or disappeared buildings in cultural heritage. In this paper we describe an on-going VR project on an UNESCO World Heritage Site that simultaneously addresses both scenarios: supporting architects in the task of designing the remaining parts of a large unfinished building, and simulating existing parts that define the environment that new designs must conform to. The main challenge for the team of architects is to advance towards the project completion being faithful to the original Gaudí's project, since many plans, drawings and plaster models were lost. We analyze the main requirements for collaborative architectural design in such a unique scenario, describe the main technical challenges, and discuss the lessons learned after one year of use of the system., Eurographics Workshop on Graphics and Cultural Heritage, VR in CH Applications, 47, 55, Carlos Andujar, Pere Brunet, Jerónimo Buxareu, Joan Fons, Narcís Laguarda, Jordi Pascual, and Nuria Pelechano, CCS Concepts: Computing methodologies --> Virtual reality; Human-centered computing --> Collaborative interaction; Applied computing --> Computer-aided design

3D physical puzzles are typically used to engage audiences in the interpretation of archaeological artefacts in a museum exhibition. The reason for this is that a puzzle can be seen as a game but also as a complex activity that archaeologists undertake to re-assemble fragments. The contribution of this paper is a novel digital worfklow for the design and fabrication of 3D heritage puzzles. The input to the workflow is an authentic artefact from a heritage collection, which is then digitised using technologies such as 3D scanning and 3D modelling. Thereafter, a puzzle generator produces the 3D puzzle pieces using a cell fracture algorithm and generates a set of puzzle pieces (female) and a single core piece (male) for fabrication. Finally, the pieces are fabricated using 3D printing technology and post-processed to facilitate the puzzle assembly. To demonstrate the workflow, we deploy the proposed method to create a 3D puzzle of an artefact, the Saltdean urn, for the Archaeological Gallery of the Brighton Museum and Art Gallery. The significance of this research is that it eases the task of creating puzzle-like activities and maintaining them within a busy museum gallery., Eurographics Workshop on Graphics and Cultural Heritage, 3D Reconstruction for CH, 77, 85, Karina Rodriguez Echavarria and Myrsini Samaroudi, CCS Concepts: Computing methodologies --> Shape modeling; Mesh geometry models; Applied computing --> Computer-aided design; Fine arts