Abstract
Background
Fetal ultrasound is an important component of antenatal care, but shortage of adequately trained healthcare workers has limited its adoption in low-to-middle-income countries. ...This study investigated the use of artificial intelligence for fetal ultrasound in under-resourced settings.
Methods
Blind sweep ultrasounds, consisting of six freehand ultrasound sweeps, were collected by sonographers in the USA and Zambia, and novice operators in Zambia. We developed artificial intelligence (AI) models that used blind sweeps to predict gestational age (GA) and fetal malpresentation. AI GA estimates and standard fetal biometry estimates were compared to a previously established ground truth, and evaluated for difference in absolute error. Fetal malpresentation (non-cephalic vs cephalic) was compared to sonographer assessment. On-device AI model run-times were benchmarked on Android mobile phones.
Results
Here we show that GA estimation accuracy of the AI model is non-inferior to standard fetal biometry estimates (error difference −1.4 ± 4.5 days, 95% CI −1.8, −0.9,
n
= 406). Non-inferiority is maintained when blind sweeps are acquired by novice operators performing only two of six sweep motion types. Fetal malpresentation AUC-ROC is 0.977 (95% CI, 0.949, 1.00,
n
= 613), sonographers and novices have similar AUC-ROC. Software run-times on mobile phones for both diagnostic models are less than 3 s after completion of a sweep.
Conclusions
The gestational age model is non-inferior to the clinical standard and the fetal malpresentation model has high AUC-ROCs across operators and devices. Our AI models are able to run on-device, without internet connectivity, and provide feedback scores to assist in upleveling the capabilities of lightly trained ultrasound operators in low resource settings.
The paper presents multi‐scale modeling of the step‐and‐flash imprint lithography, a modern patterning process, which depends on photopolymerization in order to replicate the topography of a template ...onto a substrate Colbum et al., J. Vac. Sci. Technol. B 2001, 19, 6. Multi‐scale modeling presented in this paper corresponds to densification of the feature inside the template as well as deformation of the feature after removal of the template. Linear elasticity with thermal expansion coefficient, discretized with the finite element method (FEM) Paszynski et al., IOP Conf. Series: Mat. Sci. Eng. 2010, 10, 012247 is utilized as the macro‐scale model. Molecular statics (MS) with quadratic and Lennard‐Jones potentials is utilized as the nano‐scale model Paszynski et al., ICES Report, 2005, 05‐38. Degrees of freedom from the macro‐scale model located on the interface have been identified with particles from nano‐scale domain. In order to improve the performance of the traditional approach, we propose an optimization technique for multi‐frontal direct solvers with constant coeffcients. The technique consists in reuse of sub‐branches of elimination trees over regular cube‐shaped grids built with hexahedral finite elements. To obtain an efficient reuse scheme, we construct an elimination tree in a specific way, so that all frontal matrices at certain level of the tree are identical. It is based on an observation that the solver tree for a uniform, fine 3D finite element grid is very regular, provided that introduction of boundary conditions as well as macro–nano‐scale interface conditions is postponed to the root of the tree (as opposed to applying them at the bottom nodes). Apart from a detailed description of the optimization, we offer a comprehensive estimation of the computational cost and memory usage benefits and showcase its accuracy.
The paper presents multi‐scale modeling of the step‐and‐flash imprint lithography, a modern patterning process. The internal part is modeled with finite elements while the outer part uses a more precise particle model. In order to improve the performance of software modeling of the finite element part of the domain, we propose an optimization technique for multi‐frontal direct solvers with constant coefficients. The technique consists in reuse of sub‐branches of elimination trees over regular cube‐shaped grids built with hexahedral finite elements.
This paper presents a graph-transformation-based multifrontal direct solver with an optimization technique that allows for a significant decrease of time complexity in some multi-scale simulations of ...the Step and Flash Imprint Lithography (SFIL). The multi-scale simulation consists of a macro-scale linear e lasticity model with thermal expansion coefficient and a nano-scale molecular statics model. The algorithm is exemplified with a photopolimerization simulation that involves densification of a polymer inside a feature followed by shrinkage of the feature after removal of the template. The solver is optimized thanks to a mechanism of reusing sub -domains with similar geometries and similar material properties. The graph transformation formalism is used to describe the algorithm - such an approach helps automatically localize sub-domains that can be reused.
In this paper we utilize the concept of the L2 and H1 projections used toadaptively generate a continuous approximation of an input material data inthe finite element (FE) base. This approximation, ...along with a correspondingFE mesh, can be used as material data for FE solvers. We begin with a brieftheoretical background, followed by description of the hp-adaptive algorithmadopted here to improve gradually quality of the projections. We investigatealso a few distinct sample problems, apply the aforementioned algorithms andconclude with numerical results evaluation.
The paper presents multi‐scale modeling of the step‐and‐flash imprint lithography, a modern patterning process, which depends on photopolymerization in order to replicate the topography of a template ...onto a substrate Colbum et al.,
J. Vac. Sci. Technol
. B
2001
,
19
, 6. Multi‐scale modeling presented in this paper corresponds to densification of the feature inside the template as well as deformation of the feature after removal of the template. Linear elasticity with thermal expansion coefficient, discretized with the finite element method (FEM) Paszynski et al.,
IOP Conf. Series: Mat. Sci. Eng
.
2010
,
10
, 012247 is utilized as the macro‐scale model. Molecular statics (MS) with quadratic and Lennard‐Jones potentials is utilized as the nano‐scale model Paszynski et al.,
ICES Report
,
2005
, 05‐38. Degrees of freedom from the macro‐scale model located on the interface have been identified with particles from nano‐scale domain. In order to improve the performance of the traditional approach, we propose an optimization technique for multi‐frontal direct solvers with constant coeffcients. The technique consists in reuse of sub‐branches of elimination trees over regular cube‐shaped grids built with hexahedral finite elements. To obtain an efficient reuse scheme, we construct an elimination tree in a specific way, so that all frontal matrices at certain level of the tree are identical. It is based on an observation that the solver tree for a uniform, fine 3D finite element grid is very regular, provided that introduction of boundary conditions as well as macro–nano‐scale interface conditions is postponed to the root of the tree (as opposed to applying them at the bottom nodes). Apart from a detailed description of the optimization, we offer a comprehensive estimation of the computational cost and memory usage benefits and showcase its accuracy.
Predictive artificial intelligence (AI) systems based on deep learning have been shown to achieve expert-level identification of diseases in multiple medical imaging settings, but can make errors in ...cases accurately diagnosed by clinicians and vice versa. We developed Complementarity-Driven Deferral to Clinical Workflow (CoDoC), a system that can learn to decide between the opinion of a predictive AI model and a clinical workflow. CoDoC enhances accuracy relative to clinician-only or AI-only baselines in clinical workflows that screen for breast cancer or tuberculosis (TB). For breast cancer screening, compared to double reading with arbitration in a screening program in the UK, CoDoC reduced false positives by 25% at the same false-negative rate, while achieving a 66% reduction in clinician workload. For TB triaging, compared to standalone AI and clinical workflows, CoDoC achieved a 5-15% reduction in false positives at the same false-negative rate for three of five commercially available predictive AI systems. To facilitate the deployment of CoDoC in novel futuristic clinical settings, we present results showing that CoDoC's performance gains are sustained across several axes of variation (imaging modality, clinical setting and predictive AI system) and discuss the limitations of our evaluation and where further validation would be needed. We provide an open-source implementation to encourage further research and application.
In the first part of the paper we present the multi-scale simulation of the Step-and-Flash Imprint Lithography (SFIL), a modern patterning process. The simulation utilizes the hp adaptive Finite ...Element Method (hp-FEM) coupled with Molecular Statics (MS) model. Thus, we consider the multi-scale problem, with molecular statics applied in the areas of the mesh where the highest accuracy is required, and the continuous linear elasticity with thermal expansion coefficient applied in the remaining part of the domain. The degrees of freedom from macro-scale element's nodes located on the macro-scale side of the interface have been identified with particles from nano-scale elements located on the nano-scale side of the interface. In the second part of the paper we present Unified Modeling Language (UML) description of the resulting multi-scale application (hp-FEM coupled with MS). We investigated classical, procedural codes from the point of view of the object-oriented (O-O) programming paradigm. The discovered hierarchical structure of classes and algorithms makes the UML project as independent on the spatial dimension of the problem as possible. The O-O UML project was defined at an abstract level, independent on the programming language used.
In this paper we present an agent-based algorithm for the spatial distribution of objects. The algorithm is a generalization of the bubble mesh algorithm, initially created for the point insertion ...stage of the meshing process of the finite element method. The bubble mesh algorithm treats objects in space as bubbles, which repel and attract each other. The dynamics of each bubble are approximated by solving a series of ordinary differential equations. We present numerical results for a meshing application as well as a graph visualization application.
Fetal ultrasonography is essential for confirmation of gestational age (GA), and accurate GA assessment is important for providing appropriate care throughout pregnancy and for identifying ...complications, including fetal growth disorders. Derivation of GA from manual fetal biometry measurements (ie, head, abdomen, and femur) is operator dependent and time-consuming.
To develop artificial intelligence (AI) models to estimate GA with higher accuracy and reliability, leveraging standard biometry images and fly-to ultrasonography videos.
To improve GA estimates, this diagnostic study used AI to interpret standard plane ultrasonography images and fly-to ultrasonography videos, which are 5- to 10-second videos that can be automatically recorded as part of the standard of care before the still image is captured. Three AI models were developed and validated: (1) an image model using standard plane images, (2) a video model using fly-to videos, and (3) an ensemble model (combining both image and video models). The models were trained and evaluated on data from the Fetal Age Machine Learning Initiative (FAMLI) cohort, which included participants from 2 study sites at Chapel Hill, North Carolina (US), and Lusaka, Zambia. Participants were eligible to be part of this study if they received routine antenatal care at 1 of these sites, were aged 18 years or older, had a viable intrauterine singleton pregnancy, and could provide written consent. They were not eligible if they had known uterine or fetal abnormality, or had any other conditions that would make participation unsafe or complicate interpretation. Data analysis was performed from January to July 2022.
The primary analysis outcome for GA was the mean difference in absolute error between the GA model estimate and the clinical standard estimate, with the ground truth GA extrapolated from the initial GA estimated at an initial examination.
Of the total cohort of 3842 participants, data were calculated for a test set of 404 participants with a mean (SD) age of 28.8 (5.6) years at enrollment. All models were statistically superior to standard fetal biometry-based GA estimates derived from images captured by expert sonographers. The ensemble model had the lowest mean absolute error compared with the clinical standard fetal biometry (mean SD difference, -1.51 3.96 days; 95% CI, -1.90 to -1.10 days). All 3 models outperformed standard biometry by a more substantial margin on fetuses that were predicted to be small for their GA.
These findings suggest that AI models have the potential to empower trained operators to estimate GA with higher accuracy.