Single View Research Articles

QLCM Method for Determining the Quality Level of Circular Movement in Floor Plans of Houses and Apartments

This article deals with the specific spatial organization of residential buildings and highlights the advantages of ‘floor plans with circular paths’. When renovating the existing floor plans of houses or apartments, different floor plan variants can be designed that enable circular movement. Choosing the best variant is sometimes a difficult process. For this purpose, the so-called QLCM method (quality level of circular movement) was developed, which makes it possible to calculate the quality level of the adapted floor plan variants in terms of circular movement. The method covers the aspects of movement and visibility. Different path types, such as the path from room to room, door to door, door to window, and the path along the outer walls are analyzed. Different types of views such as single, multiple, and wide views are examined. Using the QLCM method, a special quality-level coefficient is calculated for each modified floor plan. The variant with the highest calculated coefficient can be selected for implementation. Based on case studies of converting three typologically different apartments/houses from non-circular to circular floor plans, the benefits in terms of enabling circular movement, increasing visual distance, and the perception of the size of the housing unit are demonstrated.

Deep learning for the single-view echocardiographic assessment of biventricular ejection fractions: the dual-task EF2Net model

Abstract Background 2D echocardiography requires multiple views for the assessment of left ventricular (LV) systolic function and does not enable the calculation of right ventricular (RV) ejection fraction (EF). 3D echocardiography (3DE) has clear incremental value over 2D echocardiography; nevertheless, its availability and feasibility is limited. Purpose We aimed to develop a dual-task deep learning model – EF2Net – for predicting 3DE-derived LVEF and RVEF from 2D apical 4-chamber (A4C) view echocardiographic videos. Methods The EF2Net model comprises two video transformers, which were first pre-trained on 29,424 unlabeled A4C videos from 15,533 echocardiographic studies in a self-supervised fashion. In the subsequent supervised training phase, one of the transformers was trained for predicting LVEF on the publicly available EchoNet-Dynamic dataset and a dual-center international 3DE dataset comprising 5,341 labeled A4C videos from 1,408 echocardiographic studies, whereas the other transformer was trained for predicting RVEF only on the latter. Beyond testing the model internally on 20% of the dual-center dataset (i.e., internal test set), it was also validated in a labeled external validation set comprising (1) 244 A4C videos of 244 patients with different cardiac diseases and available outcome data and (2) 4,421 A4C videos of 853 healthy adults from the World Alliance of Societies of Echocardiography (WASE) study. Last, we evaluated the model on a low-risk, community-based cohort (1,166 unlabeled A4C videos of 1,166 individuals) with a 10-year follow-up to investigate the associations between the predictions and all-cause mortality. Results In the internal test set and the labeled external validation set, EF2Net predicted 3DE-derived LVEF with a mean absolute error (MAE) of 4.58 and 4.67 percentage points, respectively, whereas it achieved an MAE of 4.82 and 5.43 percentage points in predicting 3DE-derived RVEF. In the labeled external validation set, the model identified an LVEF <50% and an RVEF <45% with an area under the receiver operating characteristic curve of 0.95 and 0.86, respectively. In patients with cardiac diseases, the EF2Net-predicted LVEF and RVEF values were associated with the composite of all-cause death and heart failure hospitalization (LVEF – HR: 0.94 [0.91-0.98], p=0.001; RVEF – HR: 0.92 [0.88-0.98], p=0.004) independent of age and sex. Moreover, in the community-based cohort, the EF2Net-predicted EF values were also associated with 10-year all-cause mortality (LVEF – HR: 0.97 [0.95-0.99], p=0.026; RVEF – HR: 0.91 [0.88-0.95], p<0.001) independent of age, sex, and LV diastolic function. Conclusions EF2Net enabled the automated and accurate assessment of biventricular systolic function based on a single 2D echocardiographic view. It also exhibited robust performance when validated in a multi-ethnic dataset, and the prognostic value of its predictions was confirmed in patients with cardiac diseases and in the community.

Influence of Obesity, Race and Gender on Radiation Exposure for Epidural Procedures.

This retrospective, case controlled, comparative evaluation review of radiation exposure during epidural procedures in interventional pain management assessed variations in radiation exposure based on obesity, race, and gender. Numerous publications have shown increasing radiation exposure based on body mass index (BMI).However, the influence of race and gender have not been studied. A recent study assessing radiation exposure for epidural procedures with mandated 2 views, compared to a single view or optional 2 views, showed an increase in exposure time of 21%, with radiation dose increases of 133%. The influence of obesity, has been studied, but potential influence of race, and gender on radiation exposure for epidural procedures has not been studied. The present review shows a prevalence of total obesity with BMI of 30 or above of 50.9%, with a higher prevalence in patients below the age of 45, female gender, and African Americans. Mean fluoroscopy time and dose increased significantly from 6.9s to 0.925 milligray (mGy) for all epidural procedures for those who are underweight to 10.0s and 6.7 mGy for those with Class 3 obesity above BMI of 40. However, after adjusting for body mass index, no significant differences were noted in race for both fluoroscopy time and radiation dose across multiple epidural procedures. Fluoroscopy time and radiation dose increased in males in cervical interlaminar and lumbar transforaminal procedures.

ComNet: A Multiview Deep Learning Model for Predicting Drug Combination Side Effects.

As combination therapy becomes more common in clinical applications, predicting adverse effects of combination medications is a challenging task. However, there are three limitations of the existing prediction models. First, they rely on a single view of the drug and cannot fully utilize multiview information, resulting in limited performance when capturing complex structures. Second, they ignore subgraph information at different scales, which limits the ability to model interactions between subgraphs. Third, there has been limited research on effectively integrating multiview features of molecules. Therefore, we propose ComNet, a deep learning model that improves the accuracy of side effect prediction by integrating multiview features of drugs. First, to capture diverse features of drugs, a multiview feature extraction module is proposed, which not only uses molecular fingerprints but also extracts semantic information on SMILES and spatial information on 3D conformations. Second, to enhance the modeling ability of complex structures, a multiscale subgraph fusion mechanism is proposed, which can fuse local and global graph structures of drugs. Finally, a multiview feature fusion mechanism is proposed, which uses an attention mechanism to adaptively adjust the weights of different views to achieve multiview data fusion. Experiments on several publicly available data sets show that ComNet performs better than existing methods in various complex scenarios, especially in cold-start scenarios. Ablation experiments show that each core structure in ComNet contributes to the overall performance. Further analysis shows that ComNet not only converges rapidly and has good generalization ability but also identifies different substructures in the molecule. Finally, a case study on a self-collected data set validates the superior performance of ComNet in practical applications.

RUL prediction method based on cross-view hybrid network model

Remaining Useful Life (RUL) prediction has become a core technology in the field of prognostics and health management (PHM). However, due to the non-stationarity, weak signal characteristics and concurrent multiple faults of original signals, the estimation of RUL in a single view tends to ignore the structural relationship of samples in different spaces. To this end, this paper designs a RUL prediction framework based on a cross-view hybrid network model (CVHNet). Firstly, a dual-channel feature extraction hybrid network (DCF-HybridNet) is constructed. The original features are decomposed into time-frequency features and implicit spatial features through short-time Fourier transform (STFT) and Gramian angular difference field (GADF) - CNN, and then fused into a comprehensive feature representation with cross information. Secondly, a RUL regression algorithm integrating Transformer encoder and nonlinear fitter is developed to automatically learn the correlation between features in different views and predict RUL. The XJTU-SY rolling bearing dataset was used as an example for experimental verification. The results show that the Average CRA of the CVHNet method proposed in this paper is 0.9195 under three different working conditions. Compared with the methods in the same field, it shows superior performance and can provide strong support for the predictive maintenance of equipment.

Memory-based predictions prime perceptual judgments across head turns in immersive, real-world scenes.

Memory-based predictions prime perceptual judgments across head turns in immersive, real-world scenes.

Osfit™, a Novel Stent Designed for the Treatment of Coronary Ostial Lesions: Initial Clinical Experience and Intravascular Ultrasound Evaluation.

Stent implantation for coronary ostial lesions is challenging. This study evaluated the feasibility, safety, and immediate procedural results of the Osfit™, innovatively designed sirolimus drug-eluting stent delivery system for the treatment of coronary ostial lesions. The Osfit™ has a 5 mm long extended balloon proximal to the stent (bare balloon without mounted stent). When inflated at 1.5 atm, only bare balloon inflated and acts like a stopper. We evaluated the technical feasibility in 49 patients with a coronary ostial lesion (defined as stenosis within 3 mm from the origin of the vessel) who had received the Osfit™. Intravascular ultrasound (IVUS) analysis was done in 11 patients who consented to IVUS examination, and the depth from the proximal edge of the stent to the ostial plane (DSO) was measured. In all 49 lesions, stents were successfully implanted in one single angiographic view without obvious stent protrusion or definite angiographic missing of the ostium. The proportions of aorto-ostial, and bifurcation lesions were 28.6% and 71.4%, respectively. The DSO was 0.2±0.69 mm, and the proximal stent edges were located within 1 mm of the ostial edge in all patients. No adverse events like death, myocardial infarction, target lesion revascularization, target vessel revascularization or stent thrombosis were reported during the in-hospital period or within 30 days. The implantation of the Osfit™ for coronary ostial lesions appears to be an accurate and safe procedure that may reduce multiple angiographic projections and advanced skills.

Integrating One-Shot View Planning With a Single Next-Best View via Long-Tail Multiview Sampling

Integrating One-Shot View Planning With a Single Next-Best View via Long-Tail Multiview Sampling

Advancing Single-Plane Wave Ultrasound Imaging with Implicit Multi-Angle Acoustic Synthesis via Deep Learning.

Plane Wave Imaging (PWI) is pivotal in medical ultrasound, prized for its ultrafast capabilities essential for real-time physiological monitoring. Traditionally, enhancing image quality in PWI has necessitated an increase in the number of plane waves, unfortunately compromising its hallmark high frame rates. To fully leverage the frame rate advantage of PWI, existing deep-learning-based methods often employ single-plane wave (PW) as the sole input for training strategies to replicate multi-PWs compounding results. However, these typically fail to capture the intricate information provided by steered waves. In response, we have developed a sophisticated architecture that implicitly integrates multi-angle information by generating and dynamically combining virtual steered plane waves within the network. Employing deep learning techniques, this system creates virtual steered waves from the single primary input view, simulating a limited number of steering angles. These virtual PWs are then expertly merged with actual single PW data through an advanced attention mechanism. Through implicit multi-angle acoustic synthesis, our approach achieves the high-quality output typically associated with extensive multi-angle compounding. Rigorously evaluated on datasets acquired from simulations, experimental phantoms, and in vivo targets, our method has demonstrated superior performance over traditional single-plane wave strategies by providing more stable, reliable, and robust imaging outcomes. It excels in restoring detailed speckle patterns and diagnostic characteristics crucial for in vivo imaging, thereby offering a promising advancement in PWI technology without sacrificing speed. The code of the network is publicly available at https://github.com/yijiaLiu12/Implicit-Plane-Wave-Synthesis.

Preliminary validation of Amped FIVE software for subject height estimation.

Single view metrology poses a persistent challenge in extracting accurate quantitative information from individual images or video frames within the realm of forensic video analysis. Methods such as reverse projection, projective geometry, and photogrammetry have been used in the past with success but require validation and understanding of the limitations of each method. This study aims to conduct a preliminary validation of the subject height estimation feature in Amped FIVE software, which relies on the principles of single view metrology. A group of 14 individuals assumed an upright posture at distances of 2.4 m, 5.4 m, and 10 m away from two security cameras with different resolutions 4k (3840 × 2160) and HD (1920 × 1080). Prior to recording, participants' heights were measured but were not provided to the researcher in this study until after the analysis was completed. A height scale with clearly marked black and white graduations was used as a control. Height estimations were subsequently obtained using the Measure 3D tool in Amped FIVE software. On average, the overall error was found to be approximately ± 1.3 cm with a standard deviation of 0.9 cm. This study shows that Amped FIVE can provide accurate height estimates in a controlled environment. Future work should be done to test more difficult scenarios in less-than-ideal conditions.

X-ray-induced acoustic computed tomography: 3D X-ray absorption imaging from a single view.

Computed tomography (CT) scanners are essential for modern imaging but require around 600 projections from various angles. We present x-ray-induced acoustic computed tomography (XACT), a method that uses radiation-induced acoustic waves for three-dimensional (3D) x-ray imaging. These spherical acoustic waves travel through tissue at 1.5 × 103 meters per second, much slower than x-rays, allowing ultrasound detectors to capture them and generate 3D images without mechanical scanning. We validate this theory by performing 3D numerical reconstructions of a human breast from a single x-ray projection and experimentally determining 3D structures of objects at different depths. Achieving resolutions of 0.4 millimeters in the XZ plane and 3.5 millimeters in the XY plane at a depth of 16 millimeters, XACT demonstrates the ability to produce 3D images from one x-ray projection, reducing radiation exposure and enabling gantry-free imaging. XACT shows great promise for biomedical and nondestructive testing applications, potentially replacing conventional CT.

Transformer-based multiview spatiotemporal feature interactive fusion for human action recognition in depth videos

Transformer-based multiview spatiotemporal feature interactive fusion for human action recognition in depth videos

Contrastive message passing for robust graph neural networks with sparse labels

Contrastive message passing for robust graph neural networks with sparse labels

Measuring the temporal evolution of seagrass Posidonia oceanica coverage using autonomous marine robots and Deep Learning

Measuring the temporal evolution of seagrass Posidonia oceanica coverage using autonomous marine robots and Deep Learning

Boosting one-class transfer learning for multiple view uncertain data

Boosting one-class transfer learning for multiple view uncertain data

Abstract 4143939: A large-scale multi-view deep learning-based assessment of left ventricular ejection fraction in echocardiography

Introduction: Recent studies using deep learning techniques have demonstrated promising left ventricular ejection fraction (LVEF) assessment from transthoracic echocardiograms (TTEs). However, most prior studies have focused on videos from a single apical view, a technique known to be subject to limitations given the regionality of LV systolic function. We hypothesized that a deep learning model trained to include echocardiographic video clips from multiple views from a large dataset will improve accuracy in LVEF assessment. Methods: We identified all adult TTEs with a clinically reported LVEF at Columbia University between 2019-2024. A view classification model was trained to identify apical 4 and 2-chamber and parasternal long and short-axis views for LVEF assessment. The internal dataset was split into train, validation and test sets to train spatiotemporal convolutional models for each of the 4 views to assess LVEF for each video clip. The median clip-level LVEF within a study was used to derive a study-level LVEF. The model was evaluated on an internal test set and a large external test set, which included all available adult TTEs from Weill Cornell Medical Center since 2011. As benchmark comparison, the previously published EchoNet-Dynamic model was also evaluated on the external test set. Results: The model was trained and validated on 97,566 internal studies, comprising 1,424,265 videos from 60,741 unique patients. The model achieved state of the art performance on the internal test set (16,396 studies), with mean absolute error (MAE) of 3.4% and root mean squared error (RMSE) of 4.6%. Multi-view results were superior to all single-view models. Model showed robust predictions on external test set (179,298 studies), with MAE of 5.6% and RMSE of 7.1% and outperformed EchoNet-Dynamic (Table). Conclusions: We developed a deep learning model trained on multiple echocardiographic views using the largest dataset to date. Our model achieved state-of-the-art accuracy in assessing LVEF with a level of agreement between the AI and cardiologist LVEF assessments comparable to cardiologist interobserver variability. Further studies are underway to study the implementation of these models within clinical systems.

Abstract 4140219: Performance of a Novel Rheumatic Heart Disease Screening Protocol Led by Non-Expert Frontline Nurses in Uganda

Background: Poor healthcare access results in late- or non-diagnosis of rheumatic heart disease (RHD), perpetuating the burden of RHD in low-resource settings. The ADUNU program, a partnership with the Ugandan Ministry of Health in Kitgum, Uganda, aims to improve RHD case detection through decentralized screening led by primary care nurses, who independently perform and interpret brief screening echocardiograms using handheld echocardiography. Hypothesis: We hypothesized ADUNU’s simplified screening protocol would achieve sensitivity and specificity greater than 80% on confirmatory evaluation. Aim: To determine the health system impact of deploying a novel RHD screening protocol into the public health system in Uganda through a cross-sectional study of diagnostic accuracy. Methods: Primary healthcare nurses, certified to perform echocardiographic screening using a single parasternal long axis view in 2D and color Doppler, integrated screening into their clinical and outreach workflows. Community members with positive screens (mitral regurgitation jet ≥2cm or aortic regurgitation jet ≥1cm) were referred for confirmatory echocardiograms at the District Hospital. A random subset with negative screens were recruited for confirmatory echocardiograms at the time of screening as well. Sensitivity, specificity, predictive values, likelihood ratios, accuracy, and agreement (Cohen’s kappa) were calculated between the screening protocol and the confirmatory results. Results: Between May 2023 and April 2024, 3020 community screenings (ages 5-70 years) were conducted by 19 certified nurses. Among 113 positive screens, 61 (53.9%) were confirmed to have RHD. Among 430 negative screens, 14 (3.3%) had RHD. Screening sensitivity was 82.4% (95% CI 72.2-89.4%) and specificity 89.1% (85.9-91.6%). Positive and negative predictive values were 54.5% (45.2-63.4%) and 97.0% (94.9-98.2%). Likelihood ratios were 7.55(+) and 0.19(-). Accuracy was 88.3% (85.2 – 91.4%) and kappa was 0.59 (0.49-0.68). Conclusions: ADUNU’s novel approach to RHD active case finding achieved acceptable diagnostic performance. Nurse-led RHD screening programs that are integrated into routine clinical care shows potential for use in a comprehensive public health program. Very few RHD cases were missed, and under two referrals were generated for every positive case, an acceptable false positive rate. Further economic evaluation is underway to understand the budgetary impact and cost-effectiveness of this program.

Graph Convolutional Networks With Adaptive Neighborhood Awareness.

Graph convolutional networks (GCNs) can quickly and accurately learn graph representations and have shown powerful performance in many graph learning domains. Despite their effectiveness, neighborhood awareness remains essential and challenging for GCNs. Existing methods usually perform neighborhood-aware steps only from the node or hop level, which leads to a lack of capability to learn the neighborhood information of nodes from both global and local perspectives. Moreover, most methods learn the nodes' neighborhood information from a single view, ignoring the importance of multiple views. To address the above issues, we propose a multi-view adaptive neighborhood-aware approach to learn graph representations efficiently. Specifically, we propose three random feature masking variants to perturb some neighbors' information to promote the robustness of graph convolution operators at node-level neighborhood awareness and exploit the attention mechanism to select important neighbors from the hop level adaptively. We also utilize the multi-channel technique and introduce a proposed multi-view loss to perceive neighborhood information from multiple perspectives. Extensive experiments show that our method can better obtain graph representation and has high accuracy.

Sociolinguistics of hope: Language between the no-more and the not-yet

Abstract Sociolinguistics has recently turned its attention to the production of hope in language. Although hope is dismissed in several everyday and academic discourses as escapism or cruel optimism, if investigated ethnographically, the affect and practice of hope emerge contextually as both practical reason and semiotic ideology with important political implications. The articles in this special issue variously engage with hope as situated action whereby individuals and communities struggle for material resources, reorient temporality, recalibrate registers, create alliances, and reflexively engage with social practice to build forms of life that in many ways resist despair and paralysis. While the collection of articles gathered here does not share a single view of hope, a common thread is that hope in different ways coheres with the Brazilian Portuguese esperançar—that is, hope not as sheer or passive waiting but as pragmatic and reflexive action. (Sociolinguistics of hope, affect, practical reason, language ideology, ethnography)

DeepMesh: Differentiable Iso-Surface Extraction.

Geometric Deep Learning has recently made striking progress with the advent of continuous deep implicit fields. They allow for detailed modeling of watertight surfaces of arbitrary topology while not relying on a 3D euclidean grid, resulting in a learnable parameterization that is unlimited in resolution. Unfortunately, these methods are often unsuitable for applications that require an explicit mesh-based surface representation because converting an implicit field to such a representation relies on the Marching Cubes algorithm, which cannot be differentiated with respect to the underlying implicit field. In this work, we remove this limitation and introduce a differentiable way to produce explicit surface mesh representations from Deep Implicit Fields. Our key insight is that by reasoning on how implicit field perturbations impact local surface geometry, one can ultimately differentiate the 3D location of surface samples with respect to the underlying deep implicit field. We exploit this to define DeepMesh - an end-to-end differentiable mesh representation that can vary its topology. We validate our theoretical insight through several applications: Single view 3D Reconstruction via Differentiable Rendering, Physically-Driven Shape Optimization, Full Scene 3D Reconstruction from Scans and End-to-End Training. In all cases our end-to-end differentiable parameterization gives us an edge over state-of-the-art algorithms.