Search for: All records

3D image reconstruction from a limited number of 2D images has been a long-standing challenge in computer vision and image analysis. While deep learning-based approaches have achieved impressive performance in this area, existing deep networks often fail to effectively utilize the shape structures of objects presented in images. As a result, the topology of reconstructed objects may not be well preserved, leading to the presence of artifacts such as discontinuities, holes, or mismatched connections between different parts. In this paper, we propose a shape-aware network based on diffusion models for 3D image reconstruction, named SADIR, to address these issues. In contrast to previous methods that primarily rely on spatial correlations of image intensities for 3D reconstruction, our model leverages shape priors learned from the training data to guide the reconstruction process. To achieve this, we develop a joint learning network that simultaneously learns a mean shape under deformation models. Each reconstructed image is then considered as a deformed variant of the mean shape. We validate our model, SADIR, on both brain and cardiac magnetic resonance images (MRIs). Experimental results show that our method outperforms the baselines with lower reconstruction error and better preservation of the shape structure of objects within the images. 

Identifying regions of late mechanical activation (LMA) of the left ventricular (LV) myocardium is critical in determining the optimal pacing site for cardiac resynchronization therapy in patients with heart failure. Several deep learning-based approaches have been developed to predict 3D LMA maps of LV myocardium from a stack of sparse 2D cardiac magnetic resonance imaging (MRIs). However, these models often loosely consider the geometric shape structure of the myocardium. This makes the reconstructed activation maps suboptimal; hence leading to a reduced accuracy of predicting the late activating regions of hearts. In this paper, we propose to use shape-constrained diffusion models to better reconstruct a 3D LMA map, given a limited number of 2D cardiac MRI slices. In contrast to previous methods that primarily rely on spatial correlations of image intensities for 3D reconstruction, our model leverages object shape as priors learned from the training data to guide the reconstruction process. To achieve this, we develop a joint learning network that simultaneously learns a mean shape under deformation models. Each reconstructed image is then considered as a deformed variant of the mean shape. To validate the performance of our model, we train and test the proposed framework on a publicly available mesh dataset of 3D myocardium and compare it with state-of-the-art deep learning-based reconstruction models. Experimental results show that our model achieves superior performance in reconstructing the 3D LMA maps as compared to the state-of-the-art models. 

Ni/SBA-15 meso-structured catalysts modified with chromium and CeO2 (Ni–Cr-CeO2/SBA-15) were utilized to produce hydrogen from glycerol steam reforming (GSR). The catalysts were synthesized by a one-pot hydrothermal process and extensively characterized by analytical techniques such as N2 adsorption–desorption (BET), H2-temperature programmed reduction (H2-TPR), powder X-ray diffraction (PXRD), inductively coupled plasma-optical emission spectrometry (ICP-OES), and transmission electron microscopy (TEM). The low-angle XRD reflections affirmed that the catalysts were crystalline and possessed a 2D-ordered porosity. The BET results depicted that all the catalysts exhibited a good surface area ranging from 633 to 792m2/g, and the pore sizes were consistently in the mesoporous range (between 3 and 5 nm). TEM analysis of both calcined and spent catalysts revealed that the metal active sites were embedded in the hybrid CeO2-SiO2 support. Overall, the Ni-based catalysts exhibited higher glycerol conversion -12Ni-SBA-15–99.9%, 12Ni3CeO2-SBA-15–89.4%, and 8Ni4Cr3CeO2-SBA-15–99.7%. Monometallic 12Ni/SBA-15 performed exceptionally well, while 12Cr/SBA-15 performed poorly with the highest 71.48% CO selectivity. For short-term GSR reactions, CeO2 addition to 12Ni/SBA-15 did not have any effect, whereas Cr addition resulted in a 32% decrease in H2 selectivity. The long-term stability studies of 12Ni-SBA-15 showed H2 selectivity of ~ 64% and ~ 98% glycerol conversion. However, its activity was short-lived. After 20–30 h, the H2 selectivity and conversion dropped precipitously to 40%. The doping of mesoporous Ni/SBA-15 with Cr and CeO2 remarkably enhanced the long-term stability of the catalyst for 12Ni3CeO2-SBA-15, and 8Ni4Cr3CeO2-SBA-15 catalyst which showed ~ 58% H2 selectivity and ~ 100% conversion for the entire 60 h. Interestingly, Cr and CeO2 seem to improve the shelf-life of Ni-SBA-15 via different mechanistic pathways. CeO2 mitigated Ni poisoning through coke oxidation whereas Cr bolstered the catalyst stability via maintaining a well-defined pore size, structural rigidity, and integrity of the heterogeneous framework, thereby restricting structural collapse, and hence retard sintering of the Ni active sites during the long-term 60 h of continuous reaction. Hydrogen generation from renewable biomass like glycerol could potentially serve as a sustainable energy source and could substantially help reduce the carbon footprint of the environment 

Electronic devices have become indispensable in everyone’s life and so the computer hardware industry is demanding skilled professionals to design and physically implement devices to satisfy the market. However, misconceptions surrounding manufacturing jobs and the increasing initiatives to motivate students with engineering majors to focus on software-related topics such as artificial intelligence and blockchain are hindering students’ interest in hardware computing. Our project, funded by the NSF’s Improving Undergraduate STEM Education (IUSE) program, addresses the need to engage more students in explorations (and, eventually, design) of computer hardware by developing a set of games played on an easy-to-use hardware platform to understand and implement the fundamental concepts that are essential to modern computing systems (Figure 1). To encourage flexible and broad adoption, the games are conceived as standalone units within a curriculum design that leverages equitable pedagogical practices, experiential learning, and inquiry-based learning to cultivate engineering identity and persistence using situational interest and self-efficacy theories. We aim to offer the curriculum as an elective undergraduate course for all engineering majors at two US institutions and also research and evaluate the feasibility of implementing it as a summer program with high school students. Each module in the curriculum is divided into 5 phases: activation of prior knowledge, mini-lesson, gameplay, student-led work time, and debriefing. The games support collaboration rather than competition, and each lesson is tagged with equity spotlights, including Universal Design for Learning (UDL) and Culturally Sustaining Pedagogies (CSP) principles. Finally, informed by the Technological Pedagogical Content Knowledge (TPACK) framework, each lesson includes a teacher implementation guide and teacher educative materials to facilitate implementation (Figure 2). We have tested the first two games in the curriculum for usability and feasibility with a group of high school students. The topics of these games include binary arithmetic and Boolean logic gates. Participants were challenged to solve tasks using the hardware tools at their disposal. This usability and feasibility testing study provided us with important design and implementation implications.