An official website of the United States government
Automated interpretation of ultrasound imaging of the heart (echocardiograms) could improve the detection and treatment of aortic stenosis (AS), a deadly heart disease. However, existing deep learning pipelines for assessing AS from echocardiograms have two key limitations. First, most methods rely on limited 2D cineloops, thereby ignoring widely available Spectral Doppler imaging that contains important complementary information about pressure gradients and blood flow abnormalities associated with AS. Second, obtaining labeled data is difficult. There are often far more unlabeled echocardiogram recordings available, but these remain underutilized by existing methods. To overcome these limitations, we introduce Semi-supervised Multimodal Multiple-Instance Learning (SMMIL), a new deep learning framework for automatic interpretation for structural heart diseases like AS. During training, SMMIL can combine a smaller labeled set and an abundant unlabeled set of both 2D and Doppler modalities to improve its classifier. When deployed, SMMIL can combine information from all available images to produce an accurate study-level diagnosis of this life-threatening condition. Experiments demonstrate that SMMIL outperforms recent alternatives, including two medical foundation models.
more »
« less
A soft robotic sleeve mimicking the haemodynamics and biomechanics of left ventricular pressure overload and aortic stenosis
Preclinical models of aortic stenosis can induce left ventricular pressure overload and coarsely control the severity of aortic constriction. However, they do not recapitulate the haemodynamics and flow patterns associated with the disease. Here we report the development of a customizable soft robotic aortic sleeve that can mimic the haemodynamics and biomechanics of aortic stenosis. By allowing for the adjustment of actuation patterns and blood-flow dynamics, the robotic sleeve recapitulates clinically relevant haemodynamics in a porcine model of aortic stenosis, as we show via in vivo echocardiography and catheterization studies, and a combination of in vitro and computational analyses. Using in vivo and in vitro magnetic resonance imaging, we also quantified the four-dimensional blood-flow velocity profiles associated with the disease and with bicommissural and unicommissural defects re-created by the robotic sleeve. The design of the sleeve, which can be adjusted on the basis of computed tomography data, allows for the design of patient-specific devices that may guide clinical decisions and improve the management and treatment of patients with aortic stenosis.
more »
« less
- Award ID(s):
- 1847541
- PAR ID:
- 10368171
- Author(s) / Creator(s):
- ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; more »
- Editor(s):
- Pamies, Pep
- Date Published:
- Journal Name:
- Nature Biomedical Engineering
- ISSN:
- 2157-846X
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Automated interpretation of ultrasound imaging of the heart (echocardiograms) could improve the detection and treatment of aortic stenosis (AS), a deadly heart disease. However, existing deep learning pipelines for assessing AS from echocardiograms have two key limitations. First, most methods rely on limited 2D cineloops, thereby ignoring widely available Spectral Doppler imaging that contains important complementary information about pressure gradients and blood flow abnormalities associated with AS. Second, obtaining labeled data is difficult. There are often far more unlabeled echocardiogram recordings available, but these remain underutilized by existing methods. To overcome these limitations, we introduce Semi-supervised Multimodal Multiple-Instance Learning (SMMIL), a new deep learning framework for automatic interpretation for structural heart diseases like AS. During training, SMMIL can combine a smaller labeled set and an abundant unlabeled set of both 2D and Doppler modalities to improve its classifier. When deployed, SMMIL can combine information from all available images to produce an accurate study-level diagnosis of this life-threatening condition. Experiments demonstrate that SMMIL outperforms recent alternatives, including two medical foundation models.more » « less
-
IntroductionPrimary pulmonary vein stenosis (PVS) is a rare congenital heart disease that proves to be a clinical challenge due to the rapidly progressive disease course and high rates of treatment complications. PVS intervention is frequently faced with in-stent restenosis and persistent disease progression despite initial venous recanalization with balloon angioplasty or stenting. Alterations in wall shear stress (WSS) have been previously associated with neointimal hyperplasia and venous stenosis underlying PVS progression. Thus, the development of patient-specific three-dimensional (3D)in vitromodels is needed to further investigate the biomechanical outcomes of endovascular and surgical interventions. MethodsIn this study, deidentified computed tomography images from three patients were segmented to generate perfusable phantom models of pulmonary veins before and after catheterization. These 3D reconstructions were 3D printed using a clear resin ink and used in a benchtop experimental setup. Computational fluid dynamic (CFD) analysis was performed on modelsin silicoutilizing Doppler echocardiography data to represent thein vivoflow conditions at the inlets. Particle image velocimetry was conducted using the benchtop perfusion setup to analyze WSS and velocity profiles and the results were compared with those predicted by the CFD model. ResultsOur findings indicated areas of undesirable alterations in WSS before and after catheterization, in comparison with the published baseline levels in the healthyin vivotissues that may lead to regional disease progression. DiscussionThe established patient-specific 3Din vitromodels and the developedin vitro–in silicoplatform demonstrate great promise to refine interventional approaches and mitigate complications in treating patients with primary PVS.more » « less
-
Abstract There is a growing research interest in quantifying blood flow distribution for the entire cerebral circulation to sharpen diagnosis and improve treatment options for cerebrovascular disease of individual patients. We present a methodology to reconstruct subject‐specific cerebral blood flow patterns in accordance with physiological and fluid mechanical principles and optimally informed byin vivoneuroimage data of cerebrovascular anatomy and arterial blood flow rates. We propose an inverse problem to infer blood flow distribution across the visible portion of the arterial network that best matches subject‐specific anatomy and a given set of volumetric flow measurements. The optimization technique also mitigates the effect of uncertainties by reconciling incomplete flow data and by dissipating unavoidable acquisition errors associated with medical imaging data.more » « less
-
Abstract Previous in vitro studies interrogating the endothelial response to physiologically relevant flow regimes require specialized pumps to deliver time‐dependent waveforms that imitate in vivo blood flow. The aim of this study is to create a low‐cost and broadly adaptable approach to mimic physiological flow, and then use this system to characterize the effect of flow separation on velocity and shear stress profiles in a three‐dimensional (3D) topology. The flow apparatus incorporates a programmable linear actuator that superposes oscillations on a constant mean flow driven by a peristaltic pump to emulate flow in the carotid artery. The flow is perfused through a 3D in vitro model of the blood–brain barrier designed to induce separated flow. Experimental flow patterns measured by microparticle image velocimetry and modeled by computational fluid dynamics reveal periodic changes in the instantaneous shear stress along the channel wall. Moreover, the time‐dependent flow causes periodic flow separation zones, resulting in variable reattachment points during the cycle. The effects of these complex flow regimes are assessed by evaluating the integrity of the in vitro blood–brain barrier model. Permeability assays and immunostaining for proteins associated with tight junctions reveal barrier breakdown in the region of disturbed flow. In conclusion, the flow system described here creates complex, physiologically relevant flow profiles that provide deeper insight into the fluid dynamics of separated flow and pave the way for future studies interrogating the cellular response to complex flow regimes.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Journal Article:
- https://doi.org/10.1038/s41551-022-00937-8
