A Comparative Study of Machine Learning and Algorithmic Approaches to Automatically Identify the Yield Point in Normal and Aneurysmal Human Aortic Tissues.

The stress-strain curve of biological soft tissues helps characterize their mechanical behavior. The yield point on this curve is when a specimen breaches its elastic range due to irreversible microstructural damage. The yield point is easily found using the offset yield method in traditional engineering materials. However, correctly identifying the yield point in soft tissues can be subjective due to its nonlinear material behavior. The typical method for yield point identification is visual inspection, which is investigator-dependent and does not lend itself to automation of the analysis pipeline. An automated algorithm to identify the yield point objectively assesses soft tissues' biomechanical properties. This study aimed to analyze data from uniaxial extension testing on biological soft tissue specimens and create a machine learning (ML) model to determine a tissue sample's yield point. We present a trained machine learning model from 279 uniaxial extension curves from testing aneurysmal/nonaneurysmal and longitudinal/circumferential oriented tissue specimens that multiple experts labeled through an adjudication process. The ML model showed a median error of 5% in its estimated yield stress compared to the expert picks. The study found that an ML model could accurately identify the yield point (as defined) in various aortic tissues. Future studies will be performed to validate this approach by visually inspecting when damage occurs and adjusting the model using the ML-based approach.

Cerebral Aneurysm Wall Stress After Coiling Depends on Morphology and Coil Packing Density.

Endovascular coil embolization is now widely used to treat cerebral aneurysms (CA) as an alternative to surgical clipping. It involves filling the aneurysmal sac with metallic coils to reduce flow, induce clotting, and promote the formation of a coil/thrombus mass which protects the aneurysm wall from hemodynamic forces and prevents rupture. However, a significant number of aneurysms are incompletely coiled leading to aneurysm regrowth and/or recanalization. Computational models of aneurysm coiling may provide important new insights into the effects of intrasaccular coil and thrombus on aneurysm wall stresses. Porcine blood and platinum coils were used to construct an in vitro coil thrombus mass (CTM) for mechanical testing. A uniaxial compression test was performed with whole blood clots and CTM, with coil packing densities (CPDs) of 10%, 20%, and 30% to obtain compressive stress/strain responses. A fourth-order polynomial mechanical response function was fit to the experimentally obtained stress/strain responses for each CPD in order to represent their mechanical properties for computational simulations. Patient-specific three-dimensional (3D) geometries of three aneurysms with simple geometry and four with complex geometry were reconstructed from digital subtraction angiography (DSA) images. The CPDs were digitally inserted in the aneurysm geometries and finite element modeling was used to determine transmural peak/mean wall stress (MWS) with and without coil packing. Reproducible stress/strain curves were obtained from compression testing of CTM and the polynomial mechanical response function was found to approximate the experimental stress/strain relationship obtained from mechanical testing to a high degree. An exponential increase in the CTM stiffness was observed with increasing CPD. Elevated wall stresses were found throughout the aneurysm dome, neck, and parent artery in simulations of the CAs with no filling. Complete, 100% filling of the aneurysms with whole blood clot and CPDs of 10%, 20%, and 30% significantly reduced MWS in simple and complex geometry aneurysms. Sequential increases in CPD resulted in significantly greater increases in MWS in simple but not complex geometry aneurysms. This study utilizes finite element analysis to demonstrate the reduction of transmural wall stress following coil embolization in patient-specific computational models of CAs. Our results provide a quantitative measure of the degree to which CPD impacts wall stress and suggest that complex aneurysmal geometries may be more resistant to coil embolization treatment. The computational modeling employed in this study serves as a first step in developing a tool to evaluate the patient-specific efficacy of coil embolization in treating CAs.

An artificial intelligence based abdominal aortic aneurysm prognosis classifier to predict patient outcomes.

Abdominal aortic aneurysms (AAA) have been rigorously investigated to understand when their clinically-estimated risk of rupture-an event that is the 13th leading cause of death in the US-exceeds the risk associated with repair. Yet the current clinical guideline remains a one-size-fits-all "maximum diameter criterion" whereby AAA exceeding a threshold diameter is thought to make the risk of rupture high enough to warrant intervention. However, between 7 and 23.4% of smaller-sized AAA have been reported to rupture with diameters below the threshold. In this study, we train and assess machine learning models using clinical, biomechanical, and morphological indices from 381 patients to develop an aneurysm prognosis classifier to predict one of three outcomes for a given AAA patient: their AAA will remain stable, their AAA will require repair based as currently indicated from the maximum diameter criterion, or their AAA will rupture. This study represents the largest cohort of AAA patients that utilizes the first available medical image and clinical data to classify patient outcomes. The APC model therefore represents a potential clinical tool to striate specific patient outcomes using machine learning models and patient-specific image-based (biomechanical and morphological) and clinical data as input. Such a tool could greatly assist clinicians in their management decisions for patients with AAA.

Aortic tissue stiffness and tensile strength are correlated with density changes following proteolytic treatment.

INTRODUCTION: Dissection or rupture of the aorta is accompanied by high mortality rates, and there is a pressing need for better prediction of these events for improved patient management and clinical outcomes. Biomechanically, these events represent a situation wherein the locally acting wall stress exceed the local tissue strength. Based on recent reports for polymers, we hypothesized that aortic tissue failure strength and stiffness are directly associated with tissue mass density. The objective of this work was to test this novel hypothesis for porcine thoracic aorta. METHODS: Three tissue specimens from freshly harvested porcine thoracic aorta were treated with either collagenase or elastase to selectively degrade structural proteins in the tissue, or with phosphate buffer saline (control). The tissue mass and volume of each specimen were measured before and after treatment to allow for density calculation, then mechanically tested to failure under uniaxial extension. RESULTS: Protease treatments resulted in statistically significant tissue density reduction (sham vs. collagenase p = 0.02 and sham vs elastase p = 0.003), which in turn was significantly and directly correlated with both ultimate tensile strength (sham vs. collagenase p = 0.02 and sham vs elastase p = 0.03) and tangent modulus (sham vs. collagenase p = 0.007 and sham vs elastase p = 0.03). CONCLUSIONS: This work demonstrates for the first time that tissue stiffness and tensile strength are directly correlated with tissue density in proteolytically-treated aorta. These findings constitute an important step towards understanding aortic tissue failure mechanisms and could potentially be leveraged for non-invasive aortic strength assessment through density measurements, which could have implications to clinical care.

The regulatory environment for artificial intelligence-enabled devices in the United States.

The regulatory environment in the United States has not kept pace with the rapidly developing market for artificial intelligence (AI)-enabled devices. The number of AI-enabled devices has increased year after year. All of these devices are registered or cleared by the US Food and Drug Administration through exempt or 510(k) premarket notification pathways, and the majority are related to the radiology or cardiovascular spaces. US Food and Drug Administration guidance has not yet addressed the unique challenges of AI-enabled devices, including development, comprehensibility, and continuously learning models. The liability aspects of AI-enabled devices deployed into use by clinicians in practice have yet to be addressed. Future guidance from government regulatory sources will be necessary as the field moves forward.

Mechanical and matrix effects of short and long-duration exposure to beta-aminopropionitrile in elastase-induced model abdominal aortic aneurysm in mice.

Objective: Evaluate the mechanical and matrix effects on abdominal aortic aneurysms (AAA) during the initial aortic dilation and after prolonged exposure to beta-aminopropionitrile (BAPN) in a topical elastase AAA model. Methods: Abdominal aortae of C57/BL6 mice were exposed to topical elastase with or without BAPN in the drinking water starting 4 days before elastase exposure. For the standard AAA model, animals were harvested at 2 weeks after active elastase (STD2) or heat-inactivated elastase (SHAM2). For the enhanced elastase model, BAPN treatment continued for either 4 days (ENH2b) or until harvest (ENH2) at 2 weeks; BAPN was continued until harvest at 8 weeks in one group (ENH8). Each group underwent assessment of aortic diameter, mechanical testing (tangent modulus and ultimate tensile strength [UTS]), and quantification of insoluble elastin and bulk collagen in both the elastase exposed aorta as well as the descending thoracic aorta. Results: BAPN treatment did not increase aortic dilation compared with the standard model after 2 weeks (ENH2, 1.65 ± 0.23 mm; ENH2b, 1.49 ± 0.39 mm; STD2, 1.67 ± 0.29 mm; and SHAM2, 0.73 ± 0.10 mm), but did result in increased dilation after 8 weeks (4.3 ± 2.0 mm; P = .005). After 2 weeks, compared with the standard model, continuous therapy with BAPN did not have an effect on UTS (24.84 ± 7.62 N/cm2; 18.05 ± 4.95 N/cm2), tangent modulus (32.60 ± 9.83 N/cm2; 26.13 ± 9.10 N/cm2), elastin (7.41 ± 2.43%; 7.37 ± 4.00%), or collagen (4.25 ± 0.79%; 5.86 ± 1.19%) content. The brief treatment, EHN2b, resulted in increased aortic collagen content compared with STD2 (7.55 ± 2.48%; P = .006) and an increase in UTS compared with ENH2 (35.18 ± 18.60 N/cm2; P = .03). The ENH8 group had the lowest tangent modulus (3.71 ± 3.10 N/cm2; P = .005) compared with all aortas harvested at 2 weeks and a lower UTS (2.18 ± 2.18 N/cm2) compared with both the STD2 (24.84 ± 7.62 N/cm2; P = .008) and ENH2b (35.18 ± 18.60 N/cm2; P = .001) groups. No differences in the mechanical properties or matrix protein concentrations were associated with abdominal elastase exposure or BAPN treatment for the thoracic aorta. The tangent modulus was higher in the STD2 group (32.60 ± 9.83 N/cm2; P = .0456) vs the SHAM2 group (17.99 ± 5.76 N/cm2), and the UTS was lower in the ENH2 group (18.05 ± 4.95 N/cm2; P = .0292) compared with the ENH2b group (35.18 ± 18.60 N/cm2). The ENH8 group had the lowest tangent modulus (3.71 ± 3.10 N/cm2; P = .005) compared with all aortas harvested at 2 weeks and a lower UTS (2.18 ± 2.18 N/cm2) compared with both the STD2 (24.84 ± 7.62 N/cm2; P = .008) and ENH2b (35.18 ± 18.60 N/cm2; P = .001) groups. Abdominal aortic elastin in the STD2 group (7.41 ± 2.43%; P = .035) was lower compared with the SHAM2 group (15.29 ± 7.66%). Aortic collagen was lower in the STD2 group (4.25 ± 0.79%; P = .007) compared with the SHAM2 group (12.44 ± 6.02%) and higher for the ENH2b (7.55 ± 2.48%; P = .006) compared with the STD2 group. Conclusions: Enhancing an elastase AAA model with BAPN does not affect the initial (2-week) dilation phase substantially, either mechanically or by altering the matrix content. Late mechanical and matrix effects of prolonged BAPN treatment are limited to the elastase-exposed segment of the aorta. Clinical Relevance: This paper explores the use of short- and long-term exposure to beta-aminopropionitrile to create an enhanced topical elastase abdominal aortic aneurysm model in mice. Readouts of aneurysm severity included loss of mechanical stability and vascular extracellular matrix composition reminiscent of what is seen in the course of human disease. Additionally, we show that the thoracic aorta, unlike the findings below the renal arteries, is not damaged in our animal model.

Artificial intelligence framework to predict wall stress in abdominal aortic aneurysm.

Abdominal aortic aneurysms (AAA) have been rigorously investigated to understand when their risk of rupture - which is the 13th leading cause of death in the US - exceeds the risks associated with repair. Clinical intervention occurs when an aneurysm diameter exceeds 5.5 cm, but this "one-size fits all" criterion is insufficient, as it has been reported thatup to a quarter of AAA smaller than 5.5 cm do rupture. Therefore, there is a need for a more reliable, patient-specific, clinical tool to aide in the management of AAA. Biomechanical assessment of AAA is thought to provide critical physical insights to rupture risk, but clinical translataion of biomechanics-based tools has been limited due to the expertise, time, and computational requirements. It was estimated that through 2015, only 348 individual AAA cases have had biomechanical stress analysis performed, suggesting a deficient sample size to make such analysis relevant in the clinic. Artificial intelligence (AI) algorithms offer the potential to increase the throughput of AAA biomechanical analyses by reducing the overall time required to assess the wall stresses in these complex structures using traditional methods. This can be achieved by automatically segmenting regions of interest from medical images and using machine learning models to predict wall stresses of AAA. In this study, we present an automated AI-based methodology to predict the biomechanical wall stresses for individual AAA. The predictions using this approach were completed in a significantly less amount of time compared to a more traditional approach (~4 hours vs 20 seconds).

An Objective and Repeatable Sac Isolation Technique for Comparing Biomechanical Metrics in Abdominal Aortic Aneurysms.

(1) Abdominal aortic aneurysm (AAA) biomechanics-based metrics often reported may be over/under-estimated by including non-aneurysmal regions in the analyses, which is typical, rather than isolating the dilated sac region. We demonstrate the utility of a novel sac-isolation algorithm by comparing peak/mean wall stress (PWS, MWS), with/without sac isolation, for AAA that were categorized as stable or unstable in 245 patient CT image sets. (2) 245 patient computed tomography images were collected, segmented, meshed, and had subsequent finite element analysis performed in preparation of our novel sac isolation technique. Sac isolation was initiated by rotating 3D surfaces incrementally, extracting 2D projections, curve fitting a Fourier series, and taking the local extrema as superior/inferior boundaries for the aneurysmal sac. The PWS/MWS were compared pairwise using the entire aneurysm and the isolated sac alone. (3) MWS, not PWS, was significantly different between the sac alone and the entire aneurysm. We found no statistically significant difference in wall stress measures between stable (n = 222) and unstable (n = 23) groups using the entire aneurysm. However, using sac-isolation, PWS (24.6 ± 7.06 vs. 20.5 ± 8.04 N/cm2; p = 0.003) and MWS (12.0 ± 3.63 vs. 10.5 ± 4.11 N/cm2; p = 0.022) were both significantly higher in unstable vs. stable groups. (4) Our results suggest that evaluating only the AAA sac can influence wall stress metrics and may reveal differences in stable and unstable groups of aneurysms that may not otherwise be detected when the entire aneurysm is used.

Development of a Semi-Automated, Bulk Seeding Device for Large Animal Model Implantation of Tissue Engineered Vascular Grafts.

Vascular tissue engineering is a field of regenerative medicine that restores tissue function to defective sections of the vascular network by bypass or replacement with a tubular, engineered graft. The tissue engineered vascular graft (TEVG) is comprised of a biodegradable scaffold, often combined with cells to prevent acute thrombosis and initiate scaffold remodeling. Cells are most effectively incorporated into scaffolds using bulk seeding techniques. While our group has been successful in uniform, rapid, bulk cell seeding of scaffolds for TEVG testing in small animals using our well-validated rotational vacuum technology, this approach was not directly translatable to large scaffolds, such as those required for large animal testing or human implants. The objective of this study was to develop and validate a semi-automated cell seeding device that allows for uniform, rapid, bulk seeding of large scaffolds for the fabrication of TEVGs appropriately sized for testing in large animals and eventual translation to humans. Validation of our device revealed successful seeding of cells throughout the length of our tubular scaffolds with homogenous longitudinal and circumferential cell distribution. To demonstrate the utility of this device, we implanted a cell seeded scaffold as a carotid interposition graft in a sheep model for 10 weeks. Graft remodeling was demonstrated upon explant analysis using histological staining and mechanical characterization. We conclude from this work that our semi-automated, rotational vacuum seeding device can successfully seed porous tubular scaffolds suitable for implantation in large animals and provides a platform that can be readily adapted for eventual human use.

Iron nanoparticle contrast enhanced microwave imaging for emergent stroke: A pilot study.

Emergent stroke is mostly evaluated using hospital based imaging. Quick imaging allows for rapid administration of IV thrombolysis and outcome improvement. Microwave imaging (MI) is an emerging portable imaging modality. Iron oxide nanoparticles are known to interact with microwave frequency electromagnetic radiation. In this manuscript, we provide proof of concept for a novel iron oxide nanoparticle enhanced microwave imaging device for differentiating emergent ischemic stroke from hemorrhagic stroke. A MI device was constructed. Attenuation of the microwave signal transmitted with or without iron oxide nanoparticles was measured over a 1-2 GHz frequency range in a silicone brain phantom, in New Zealand white rabbits, and in a human. Observed differences in signal attenuation were used to reconstruct an image following induction of a left sided anterior circulation stroke in a New Zealand white rabbit. An increase in microwave signal attenuation exists across a frequency range of 1.3-2 GHz when iron oxide nanoparticles are introduced into a silicone phantom model, in New Zealand white rabbits, and in a human volunteer. Using this increase in signal attenuation following nanoparticle administration, we localize induced ischemia in a New Zealand white rabbit. To the best of out knowledge, we provide the first evidence that superparamagnetic Iron oxide nanoparticles may be used as contrast in the setting of MI. Our data suggest infusion of intravenous iron oxide nanoparticles with follow on microwave imaging may ultimately allow for more timely administration of thrombolytic mediation in the setting of acute ischemic stroke.

Does elevated wall tension cause aortic aneurysm rupture? Investigation using a subject-specific heterogeneous model.

OBJECTIVE: To investigate whether peak wall tension in abdominal aortic aneurysm occurs at the site of rupture to test for a causative relationship. METHODS: Four ruptured and nine unruptured AAA were harvested whole from cadavers, followed by regional measurements of wall thickness, elastic parameters and failure tension. Finite element models were developed with subject-specific load-free AAA morphology and heterogeneous properties interpolated using a geodesic distance weighted approach from the measurements. The wall tension under uniform pressure and tension to failure tension ratio as an index of susceptibility to rupture were computed. As a secondary aim, the peak wall tension using this heterogeneous model approach was compared to the traditional homogeneous model approach in order to evaluate the reliability of the latter. RESULTS: The average peak wall tension in the ruptured group was 43% higher than in the unruptured group without statistical significance even though it was 54% larger in diameter. The site of peak wall tension was in the vicinity of the site of rupture in two ruptured AAA. The peak tension did not breach failure tension at the rupture site in any of the AAA. The traditional population-wide homogeneous model approach overestimated peak wall tension by just 3% compared to the subject-specific heterogeneous model approach. CONCLUSION: We failed to find adequate evidence of a causative relationship between peak wall tension and AAA rupture. The findings are not conclusive owing to study limitations such as ignoring intraluminal thrombus, sparse distribution of specimens procured and small study population.

Modeling Outcomes: Modified Aortic Arch Advancement for Neonatal Hypoplastic Arch.

OBJECTIVE: Numerous surgical approaches regarding aortic arch advancement for neonatal arch hypoplasia have been described. These repairs can be classified into two categories: those that incorporate a patch and those that do not. The decision between repairs remains largely experiential, rather than empirical, because of the limited number of reported outcomes. We report early outcomes from neonates undergoing modified aortic arch advancement with an anterior patch and our experience using computational fluid dynamic modeling to better understand the hemodynamic consequences associated with this repair. METHODS: A retrospective review of neonates undergoing aortic arch advancement with anterior patch in 2014 at a single institution was performed. Anatomical, perioperative, and follow-up data were collected. Three-dimensional cardiac magnetic resonance images were used to generate computational fluid dynamic models of the modified anterior patch and direct end-to-side repairs. Cardiac waveform inputs were simulated and hemodynamic analyzed. RESULTS: Ten neonates underwent modified aortic arch advancement. No hemodynamically significant gradients were observed at a median follow-up of 0.77 (0.30-1.2) years. Asymmetrical flow was observed in the end-to-side repair, whereas more concentric laminar flow was observed throughout the modified model. Spatial variations in velocities immediately distal to the anastomosis were greater in the end-to-side model (0.35 vs 0.17 m/s, P < 0.001). Time-averaged variations in wall shear stress during systole were greater in the end-to-side model at the same location (3.44 vs 1.98 dynes/cm, P < 0.001). CONCLUSIONS: Early outcomes after the use of an anterior patch for neonatal hypoplastic aortic arch repair show favorable hemodynamic outcomes.