Patient-specific beating heart simulator

Abstract

A synthetic patient specific A benchtop beating heart simulator that is custom-fabricated using 3D printing technology based on patient-specific imaging data has been developed. The simulator employs soft robotics to accurately mimic the motion and biomechanics of the left atrium, with a particular focus on the left atrial appendage and conditions associated with Heart failure with preserved ejection fraction (HFpEF). This simulator is designed to recreate hemodynamics in a controlled benchtop setting, allowing for precise testing and evaluation of medical device designs and interventions.

Description

[0001] PATIENT-SPECIFIC BEATING HEART SIMULATOR

[0002] Cross-Reference to Related Applications

[0003] This application claims the benefit of and priority to U. S. Provisional Application No. 63 / 721,942, filed November 18, 2024, the entire content of which is hereby incorporated herein by reference for all purposes in its entirety.

[0004] Statement Regarding Federally Sponsored Research or Development This invention was made with government support under CBET1847541 awarded by the National Science Foundation. The government has certain rights in the invention.

[0005] Field of the Invention

[0006] I’he invention pertains to the field of medical device testing and simulation, specifically to a benchtop simulator that mimics the human left heart's biomechanics and hemodynamics for testing and evaluating interventions and device designs related to left atrial appendage closure and treatment of heart failure with preserved ejection fraction.

[0007] Background of the Invention

[0008] Atrial fibrillation (AF) is the most common sustained cardiac arrhythmia and is associated with a significantly increased risk of stroke. This risk arises from the formation of blood clots within the left atrial appendage (LAA), a small, ear-shaped sac in the muscle wall of the left atrium. In patients with AF, blood flow can be sluggish, leading to clot formation in the LAA, which can then travel to the brain, causing a stroke. Stroke prevention is a primary goal in the management of AF, with LAA closure emerging as an important intervention strategy. LAA closure aims to seal off the LAA from the left atrium to prevent the embolization of clots. This procedure has become an increasingly important alternative to long-term anticoagulation therapy, especially for patients who are at high risk of bleeding or those for whom anticoagulation is contraindicated. The effectiveness of LAA closure in reducing stroke risk in AF patients has been supported by a growing body of clinical evidence, positioning it as a key therapeutic intervention in contemporary cardiac care.

[0009] Challenges Posed by LAA Variability

[0010] The anatomy and morphology of the LAA vary significantly among individuals, presenting a major challenge in the design and successful implementation of closure devices. The LAA can exhibit a wide range of shapes, sizes, and orientations, including but not limited to cactus, chicken wing, windsock, and cauliflower morphologies. This anatomical variability complicates the accurate sizing and placement of closure devices, as a device that fits well in one patient's LAA may be completely unsuitable for another's. The highly variable nature of LAA anatomy necessitates a personalized approach to device design and selection. However, current closure 1

[0011] 45802538 devices are limited in their adaptability, often designed to accommodate a "onc-sizc-fits-most" strategy rather than addressing the full spectrum of anatomical diversity. This limitation not only impacts the procedural success rates but also elevates the risk of complications, such as devicerelated thrombus formation, pericardial effusion, and residual leaks, which can negate the benefits of the procedure. Moreover, the importance of accurately assessing LAA morphology prior to closure procedures cannot be overstated. Conventional imaging techniques used in the planning and guidance of LAA closure, such as transesophageal echocardiography (TEE), computed tomography (CT), and magnetic resonance imaging (MRI), provide valuable insights but may not always capture the full complexity of the LAA's structure. This gap underscores the need for advanced modeling and simulation tools that can offer a more detailed and patient-specific understanding of LAA anatomy, facilitating the design and selection of more effective and safer closure devices. The development of a benchtop beating heart simulator, capable of replicating multiple, easily exchanged, patient specific LAA anatomies with appropriate biomechanics, stands as a promising solution to these challenges. By enhancing the ability to accurately model and test LAA closure devices in a broad spectrum of LAA morphologies and pathologies, one can move closer to achieving personalized and effective stroke prevention strategies for patients with AF.

[0012] The Complexity of HFpEF and its Treatment

[0013] Heart Failure with Preserved Ejection Fraction (HFpEF) represents a complex clinical syndrome characterized by symptoms of heart failure, a preserved ejection fraction, and evidence of diastolic dysfunction. Accounting for approximately half of all heart failure cases, HFpEF poses a significant challenge in cardiovascular medicine due to its heterogeneous nature and the multifaceted pathophysiology that underlies the condition. Unlike heart failure with reduced ejection fraction (HFrEF), where several evidence-based therapies have demonstrated significant benefits, effective treatment options for HFpEF remain limited, highlighting a substantial unmet medical need. The treatment of HFpEF is complicated by several factors. First, the condition encompasses a wide variety of pathophysiological mechanisms, including but not limited to left ventricular stiffness, impaired relaxation, and abnormal ventricular filling, as well as pulmonary congestion and increased left atrial pressures along with left atrial dilatation. This diversity in underlying causes and contributing factors makes the development of a one-size-fits-all treatment approach particularly challenging. Device-based interventions represent a promising avenue for innovation, offering potential to directly modify the disease process and provide symptomatic relief for patients with HFpEF. A recent approach to treating HFpEF involves designing devices to reduce diastolic pressure in the left ventricle by strategically relieving the preload from the left atrium. These devices function by creating a shunt that allows blood to flow directly from the left atrium to 2

[0014] 45802538.1 the aorta, effectively bypassing the left ventricle. The efficacy and safety of such devices are paramount, necessitating rigorous preclinical testing to optimize its design and functionality. Herein lies the role of the benchtop beating left heart simulator, specifically tailored to replicate the left atrium's biomechanics and hemodynamics. The simulator's capacity to mimic patient-specific left atrial anatomy and function allows for a comprehensive assessment of how the shunt device interacts with varying anatomical configurations and disease states. By simulating the precise anatomical and physiological conditions of the left atrium, including the unique hemodynamic challenges presented by HFpEF, the simulator can provide an invaluable resource for evaluating the novel shunt device's performance under highly controlled yet physiologically relevant conditions.

[0015] The development and testing of medical devices for cardiac interventions, such as left atrial appendage (LAA) closure and treatment of heart failure with preserved ejection fraction (HFpEF), requires accurate simulation of the human heart's dynamic environment. Current in vitro testing models for cardiac interventions and device designs do not adequately replicate the complex biomechanics and hemodynamics of the human heart, limiting the ability to accurately predict clinical outcomes and hemodynamic effects of interventions.

[0016] Furthermore, there is considerable intra-patient anatomical heterogeneity with respect to LAA geometry and size, which contributes to patient-device mismatch. Current in vitro models only capture a single LAA geometry, rendering them ineffective at encompassing broad patient heterogeneity, and therefore limited in scope for device development for diverse or challenging patient geometries.

[0017] Summary of the Invention:

[0018] A benchtop beating heart simulator that is custom- fabricated using 3D printing technology based on patient-specific imaging data has been developed. The simulator employs soft robotics to accurately mimic the motion and biomechanics of the left atrium, optionally with a particular focus on the left atrial appendage and conditions associated with Heart failure with preserved ejection fraction (HFpEF). This simulator is designed to recreate hemodynamics in a controlled benchtop setting, allowing for precise testing and evaluation of medical device designs and interventions. This simulator is also designed to allow rapid interchange of different Left Atrial Appendage (LAA) models (different geometries and material properties) that allows for device testing in multiple patient-derived anatomies, representative of both anatomical and pathological heterogeneity.

[0019] Brief Description of the Drawings

[0020] FIGs. 1A-1G are schematics of the location and anatomy of the left atrial appendage (“LAA”), and mechanism of thrombus formation during atrial fibrillation. This shows the anatomical and morphological 3

[0021] 45802538.1 overview of the LA and LAA, showing healthy and atrial fibrillation (AF) states (FIG. 1A) and schematics of LAA morphologies (FIGs. 1B-1E). The “chicken wing” morphology (FIG. IB) of an LAA with an obvious bend in the proximal or middle part of the dominant lobe is the most common variant found in 48% of patients. The “cactus” LAA (FIG. 1C) has a dominant central lobe with secondary lobes extending from the central lobe in both superior and inferior directions and is found in 30% of patients. The “windsock” morphology (FIG. ID) describes an LAA in which one dominant lobe of sufficient length is the primary structure, found in 19% of patients. The “cauliflower” LAA’s main characteristic is an LAA that has limited overall length with complex internal features, present in just 3% of patients (https: / / journals.viamedica.pl / folia_morphologica / article / view / 76067). Atrial fibrillation (AF) is caused by abnormal electrical activity. Disordered electrical propagation causes disorganized stimulation of the myocardium and subsequent arrhythmic contractions. AF decreases contractility, resulting in blood stasis and diminished peak flow velocities. AF is also associated with endothelial damage, fibrosis, and inflammation, especially within the LAA, which leads to a prothrombotic and hypercoagulable state. This association is consistent with “Virchow’s Triad (FIG. IF), which synthesizes the pathogenesis of clot formation in the LAA in patients with AF: abnormal blood flow (stasis), endocardial dysfunction (vessel wall injury), and altered hemostasis (hypercoagulability). FIG. 1G is a schematic showing an exemplary left atrial cardiac simulator that can be used to advance the field of LAAO. The simulator includes a LA model that is anatomically accurate and can accommodate a diverse array of patient specific, interchangeable LAA geometries. The system allows real-time analysis of hemodynamics that can be tuned to mimic physiological and pathological conditions. The simulator is compatible with multiple clinical imaging modalities and allows for repeatable device deployment with visual feedback and measurable hemodynamic changes, making it a highly useful tool for device development and user training.

[0022] FIG.2A shows the segmentation of CT data from patients (FIG. 2A), soft 3D-printed robotic actuators mimicking LA, LAA wall motion (FIG.2B), and the reservoir and the pump (FIG. 2C). FIG.2D-2H is a workflow diagram for 3D-CT imaging and segmentation in healthy and AF patients (FIG.2D), producing patient-specific LA and LAA geometries (FIG. 2E) for the simulator design. Clinical imaging-derived segmentation of LA and LAA structures (FIG.2F) from CT data, illustrating significant anatomical heterogeneity across patients. Soft 3D-printed model of the LA and LAA based on patient data (2G), fabricated to accurately replicate the flexibility and biomechanics of human cardiac tissue. Integration of soft robotic actuators within the LA and LAA model (FIG. 2H), which allows for biomimetic wall motion simulation, including atrial systole and diastole. FIG.21 shows anatomical and morphological overview of the LA and LAA, showing healthy and atrial fibrillation (AF) states, along with the challenges posed by geometric mismatches for LAAO due to inter-patient variability and LAA trabeculations. The illustrations presented in this figure panel were created using BioDigital (under a Personal Plus subscription). FIG. 2J shows clinical imaging-derived segmentation of LA and LAA structures from CT data, illustrating significant anatomical heterogeneity across patients. FIG.2K is a collection of images showing the muscular architecture of the LA, highlighting the myocardial fiber arrangement, modified from reference 19.

[0023] 4

[0024] 45802538.1 FIG.2L are illustrations of native myocardial fiber orientation in the LA and the corresponding placement of soft robotic actuators in four discrete regions, modified from reference 18. FIG.2M is a collection of dynamic images of the soft robotic LA and LAA during atrial diastole and atrial systole where inflation and deflation of the soft robotic actuators replicate native wall mechanics.

[0025] FIGs.3A-3D shows the workflow from clinical imaging FIG.3A to soft, patient-derived LAAs for attachment to rigid model. FIG.3A, Image segmentation, post-processing, and cutting at the ostium plane.

[0026] 3B, 3D printed mold for silicone casting of LAA geometry. Elastic modulus of porcine LAA. * Elastic modulus of 2mm thickness Ecoflex 00-30. FIG.3C, segmented LA anatomy, FIG.3D, soft, patient-derived LAA models.3E-3H are side, top and bottom view of LAA1 (FIG.3E), LAA2 (FIG.3F), LAA3 (FIG. 3G), and LAA4 (FIG.3H). FIGs.3I-3K shows workflow from clinical imaging to soft, patient-derived LAAs for attachment to rigid model. FIG.31 shows image segmentation, post-processing, and cutting at the ostium plane. FIG.3J shows an exemplary 3D printed mold for silicone casting of LAA geometry. FIG.3K is a table showing comparing measurements of the LAA models with the healthy population and patients with AF who experienced cardioembolic stoke. tValues for Healthy Population from Al-Saady et al.l ± Values for Cardioembolic Stroke with AF (n=57) from Jeong et al.31 * Elastic modulus of real LAA tissue at e=10-20% from Fanni et al.27**Elastic modulus of 2mm thickness Ecoflex 00-30.

[0027] FIGs.4A-4D show the WATCHMAN FLX (FIG. 4A), WAVECREST (FIG. 4B), AMULET (FIG.

[0028] 4C), AMPLATZER (FIG.4D), ULTRASEED (FIG.4E), and LAMBRE (FIG.4F) which are exemplary commercially available, CE-marked percutaneous LAAO devices.

[0029] FIG. 5A is schematic of circulatory flow loop and incorporation of patient-derived model to achieve LA cardiac simulator. Incorporation of patient-derived model into circulatory flow loop to achieve left atrial cardiac simulator. FIG. 5B is a schematic of a patient-derived 3DP LA model used in LA cardiac simulator. FIG.5C shows the cauliflower LAA model coupled to rigid patient-derived LA model. Simulator has four inflows (RI PV, RS PV, LS PV, LI PV) and one outflow (MV). Schematic of complete left atrial cardiac simulator. Simulator consists of pulse duplicator pump, patient-derived LA model, mitral valve, tunable resistance and compliance elements, pressure sensors, and flow probes. LA = left atrium; RS PV = right superior pulmonary vein; RI PV = right inferior pulmonary vein; LI PV = left inferior pulmonary vein; LS PV = left superior pulmonary vein; MV = mitral valve. FIGs. 5D-5G. Measured values in LA cardiac simulator: FIG. 5D, points measured, FIG. 5E, Wigger's Diagram (pressure over time); FIG. 5F, flow (L / min over time, seconds); FIG.5G, pressure (mmHg) over time, seconds. FIGs. 5H and 51 show schematics of exemplary circulatory flow loop and incorporation of patient-derived LA model to achieve LA cardiac simulator. FIG.5H is a schematic of an exemplary circulatory flow loop with patient-derived LA model, pulsatile pump, commercially available mechanical heart valve in the mitral position, and tunable resistance and Windkessel-based compliance elements to mimic pulmonary and systemic vasculature. Instrumentation and corresponding measured values indicated 5

[0030] 45802538.1 (Q indicates flow; P indicates pressure). Total flow (QTOT) is measured at the outflow of the pulsatile pump. Flow (Q) is measured at the four pulmonary vein inlets. Pressure (P) is measured at the four pulmonary vein inlets, the left atrium, the left atrial appendage, and the simulated left ventricle and aorta. FIG. 51 is a 3D rendering of LA model composed of patient-specific soft elastomeric LAA coupled to 3DP patient-derived rigid LA. The LA model has four inflows (RIPV, RSPV, LSPV, LIPV) and one outflow (MV). The model can be customized by varying the LA geometry (~ days) or interchanging the LAA geometry (~ minutes) from the library of LAA models of varying geometry and material properties.

[0031] FIG. 6A is a diagram of tunable parameters in the system including tunable pulsatile pump in LA cardiac simulator. The data obtained in the system is shown as the Wiggers diagram, FIG.

[0032] 6B; Pressure over time, FIG. 6C, flow over time, FIG.6D and pressure over time, FIG. 6E.

[0033] FIGs. 7A-7D Tunable compliance in LA cardiac simulator 7A, graph of LA pressure over time for LAA1 decreasing compliance, FIG. 7B, graph of LA pressure over time for LAI for Cl-C57, FIG. 7C, graph of pulse pressure for C1-C5, FIG. 7D. SIL30 and Ecoflex 30 LAA1 pressure over C1-C5, FIG. 7E, and graph of pulse pressure for SIL30 and Ecoflex 30 LAA1, FIG.7F.

[0034] FIG. 8A-8F. Tunable LAA geometry in LA cardiac simulator, diagram of tunable parameters LAA1 parameters, 8A, diagram of tunable parameters LAA4, 8B; graphs of pressure overtime for LAA1, 8C and LAA4, 8D; graphs of flow over time, 8E, LAA1 and LAA4, 8F.

[0035] FIGs. 9A-9D, acrylic cap: FIG. 9A, 24 mm Watchman FLX LAA2; FIG. 9B, no device LAA2; FIG. 9C, graph of pulse pressure in FIG.9D. FIG. 9D is a dot plot showing LA pressure amplitude (LAPpuisePressure = LAPmaxima - LAPminima) before (No Device) and after LAAO with a 24 mm FLX (24 mm FLX) or a silicone coated impermeable 24 mm FLX device (Coated 24 mm FLX), or exclusion with an acrylic cap (Acrylic Cap). All data collected at 60 BPM and pump SV of 90 mL. Dunnell’s T3 multiple comparisons test, **** <0.0001, *** <0.0002, ** <0.0021, * <0.0332, ns = not significant. FIG. 9E is a dot plot showing mean flow before (No Device) and after LAAO with a silicone-coated impermeable FLX device (Coated FLX). Inflow is defined as the sum of the mean flow through the four pulmonary veins. All data collected at 70 BPM and pump SV of 70 mL or 80 mL. FIG. 9F shows the mean difference in volume per beat between inflow and outflow before (No Device) and after LAAO with a silicone coated impermeable FLX device (Coated FLX). All data collected at 70 BPM and pump SV of 70 mL or 80 mL. Paired t test, p = 0.0546. LAAO: left atrial appendage occlusion: LA: left atrium; LAA: left atrial appendage: SV: stroke volume; PV: pulmonary vein. RSPV: right superior pulmonary vein; RIPV: right inferior pulmonary vein: LIPV: left inferior pulmonary vein; LSPV: left superior pulmonary vein; LA: left atrium; LAA: left atrial append-age; MV: mitral valve: LV: left ventricular.

[0036] 6

[0037] 45802538.1 FIGs. 10A-10D. positioning LAA2 in LA cardiac simulator of Watchman FLX (FIG. 10A) and coated Watchman (FIG. 10B) in cardiac simulator; graphs of LAP peak pressure over time uncoated (FIG. 10C) and coated watchman FLX (FIG. 10D).

[0038] FIGs. 11A-11D, uncoated, FIG. 11A, and coated Watchman FLX, FIG. 11B, intracardiac echography, 11C; diagrams of device and device canted, FIG. 11D.

[0039] FIGs. 12A-12F shows LAAO procedural training in LA cardiac simulator. Repeated device deployment and recapture performed by an experienced user using the Access System and Delivery System in the cardiac simulator. FIG. 12A is a schematic of LA model that allows septal crossing with a delivery sheath and device deployment into the LAA (Watchman FLX positioning). FIG. 12B is an exemplary LA model with delivery sheath advanced into the LAA. LAA model allows direct visualization for enhanced user-feedback (deployment). FIGs. 12C-12F show repeated device deployment and recapture until PASS™ (position, anchor, size, seal) criteria are met: recapture and deployment (FIG. 12C), recapture and deployment 2 (FIG. 12D), recapture and deployment (FIG. 12E), recapture and deployment (FIG. 12F).

[0040] Device release (FIG. 12F, right) when user is satisfied with device deployment.

[0041] FIGs. 13A-13B shows validation of LA cardiac simulator by demonstrating physiological and pathological hemodynamics. FIG. 13A show Mean LAP achieved in LA cardiac simulator validated against reference data (Table 2) from healthy adults (dotted green lines) and clinical data measured in patients with AF (n = 435), SEC (n = 32) and post-LAAO (n = 250). LAA1 (healthy patient) was used in the simulator to collect “Healthy” data. LAA4 (AF patient) was used in the simulator to collect “AF”, “SEC” and “Post-LAAO” data. The simulator was tuned by varying the pump heart rate, pump stroke volume, resistance, compli-ance and static reservoir height to achieve the target pressure values. Reference data for healthy adults from UpToDate (Table 4). Clinical data from collaborators at Mayo Clinic ( Table 5). LAP: left atrial pressure; AF: atrial fibrillation; SEC: spontaneous echo contrast; post-LAAO: post-left atrial appendage occlusion. Figure 13B shows mean LA inflow achieved in LA cardiac simulator at varying pump stroke volume (45-90 mL) and pump heart rate (60-160 BPM). Data collected with LAA1. Red shading indicates range of mean LA inflows in patients with AF referenced from the literature (Table 3). Green shading indicates range of LA inflows in healthy adults referenced from the literature (Table 3). LA: left atrium; SV: stroke volume.

[0042] FIGs. 14A-14D shows simulation of AF hemodynamics by increasing pump heart rate. FIG. 14A shows LSPV flow measured in cardiac simulator as pump HR is increased from 70 BPM to 120 BPM in real time. FIG. 14B shows mean LSPV flow measured in cardiac simulator over 10 seconds as pump HR is increased from 70 BPM to 120 BPM to mimic AF. FIG. 14C show LA outflow measured in cardiac simulator at pump HR of 60 BPM and 120 BPM. FIG. 14D shows mean LA outflow measured in cardiac simulator over 10 seconds at pump HR of 60 BPM and 120 BPM to mimic AF. Welch's two-tailed t-test, **** <0.0001. LSPV: left superior pulmonary vein; LA: left atrium; HR: heart rate.

[0043] FIGs. 15A-15C show soft robotic actuator positioning influences left atrium (LA) and left atrial appendage (LAA) wall motion and displacement under varying pressures. FIG. 15A show 7

[0044] 45802538.1 different configurations of soft robotic actuators wrapped around the LA and LAA structures at distinct anatomical orientations (highlighted in orange), impacting the extent of wall motion and direction of simulated atrial contraction. FIG. 15B show M-mode ultrasound images capturing wall displacement at increasing actuator pressures. FIG. 15C show quantitative analysis of wall displacement in response to input pressure to each actuator, highlighting tunable mechanical response.

[0045] FIG. 16A shows pressure waveforms from the left ventricular pressure (LVP) and left atrial pressure (LAP) within the circuit, showing characteristic phases of the cardiac cycle (a, c, v waves) associated with left atrial hemodynamics. FIG. 16B shows pressure traces show LVP and aortic pressure (AoP) over several cardiac cycles, highlighting the model’s capacity to simulate systemic hemodynamics. FIG. 16C shows flow waveforms through the aortic ( AoV) and mitral (MV) valves. FIG. 16D shows pulsed-wave Doppler of LA flow velocities capturing the hemodynamic characteristics of the LA to mimic LA filling and emptying patterns.

[0046] FIGs. 17A-17C shows the soft robotic left ventricle (LV) model was validated against porcine. FIG.

[0047] 17A shows the experimental setup for testing the soft robotic LV’s ability to replicate systemic physiological flow and pressure parameters under physiological resistance, compliance, and native vasculature with a hybrid synthetic-biological flow loop configuration in a swine circulatory system. FIG. 17B shows invasive hemodynamic data recorded in a live swine with the native LV: systemic blood pressure (BP), cardiac outflow, and left ventricular pressure (LVP). FIG. 17C shows hemodynamic measurements with the soft robotic LV, showing BP, cardiac outflow, and LVP generated by the synthetic LV inside the porcine circulatory system with native vasculature.

[0048] FIG. 18 shows functional assessment of the soft robotic left atrium (LA), left atrial appendage (LAA), and left ventricle (LV) model demonstrates its ability to replicate realistic cardiac hemodynamics and valve mechanics. As shown in FIG. 18, soft robotic LA / LAA and LV models equipped with actuators designed to simulate contraction and relaxation with high anatomical and physiological fidelity shown in top panels. The illustrations presented in this figure panel were created using BioDigital (under a Personal Plus subscription).

[0049] FIGs. 19A-19J shows simulation of atrial fibrillation (AF) in the soft robotic left atrium (LA) and left atrial appendage (LAA) model demonstrates hemodynamic changes, contractility, and flow patterns associated with AF and sinus rhythm (SR). FIG. 19A show temporal velocity profile of LA and LAA flow velocities over time in the soft robotic model measured via 2D phase-contrast magnetic resonance imaging (MRI). FIG. 19B shows mean LA velocity under increasing actuation pressures (10, 12, and 15 psi). FIG 19C shows mean LAA velocity under the same range of pressures (10, 12, and 15 psi). FIG. 19D shows development of the soft robotic LA and LAA model derived from a patient with AF, showing enlarged LA and LAA geometries characteristic of AF-induced remodeling. FIG. 19E shows mitral flow and left atrial pressure (LAP) waveforms associated with atrial arrhythmia. FIG. 19F shows pressure-volume (PV) loops of the LA in sinus rhythm (60 bpm) vs. atrial flutter (150 bpm). FIG. 19G shows hemodynamic response to 8

[0050] 45802538.1 increasing heart rates (40-150 bpm), showing LAP, left ventricular pressure (LVP), and cardiac outflow (CO), associated with atrial arrhythmias, including slow ventricular response (SVR, 40 bpm), controlled ventricular response (CVR, 60 bpm), and rapid ventricular response (RVR, 150 bpm). FIG. 19H is a collection of graphs depicting the effect of LA contractility as a function of input pressure on peak A-wave velocity, a-wave pressure, and cardiac outflow (CO). FIG. 191 shows LAA ejection fraction as a function of input pressure (or LA / LAA contractility), mimicking states of both AF and SR. FIG. 19J shows color Doppler echocardiography images showing simulation of flow through the LAA ostium in healthy SR vs. atrial flutter.

[0051] FIGs.20A-20C shows assessment of left atrial appendage occlusion (LAAO) in the soft robotic simulator demonstrates device positioning and flow dynamics. FIG.20A shows imaging and visualization of the LAAO device (WATCHMAN FLX) within the LAA of the soft robotic simulator. Echocardiographic and camera views show the proper positioning of the device within the LAA. FIG.20B shows color flow Doppler mapping of the LAA with undersized (24 mm) and correctly sized (27 mm) occlusion devices. FIG.

[0052] 20C shows phase contrast magnetic resonance imaging (MRI)-based segmentation and imaging of the LAA and LA before and after LAAO. Orange arrows show the integrated soft robotic actuators around the underlying anatomy and blue arrows highlight the deployed WATCHMAN FLX device for LAAO.

[0053] FIGs.21A-21H are computational models analyze the impact of left atrial appendage occlusion (LAAO) on left atrial (LA) hemodynamics and the effect of atrial fibrillation (AF) on LA wall mechanics.

[0054] FIG.21A is a schematic of an exemplary lumped parameter model (LPM) developed to study and understand the effect of LAAO on LA hemodynamics. FIG.21B Hemodynamic data from the LPM model, showing left atrial pressure (LAP) before and after LAAO. FIG.21C shows mitral and pulmonary vein flow before and after LAAO. FIG.21D shows in vivo LAP data from live porcine models before and after LAAO, showing a similar increase in mLAP post-occlusion, aligning with predictions from the LPM. FIG.

[0055] 21E shows predicted LA wall stress distributions during the pump and reservoir phases of the cardiac cycle in sinus rhythm (SR) and AF, derived from a dynamic finite element analysis (FEA) model. The gray color in the contour plots indicates values that exceed the maximum range of the color bar. FIG. 21F shows simulated LAP and LA volume (LAV) during SR, showing characteristic a- and v-waves with corresponding minimum (Vmm) and maximum (Vmax) LA volumes. FIG.21G shows pressure-volume loops for SR and AF generated by FEA model. FIG. 21H shows validation of simulated pressure-volume loop shape with in vivo data from a healthy porcine heart in SR.

[0056] FIG.22A-22E shows the soft robotic left atrium (LA) and left atrial appendage (LAA) models were validated against porcine and human data, demonstrating realistic hemodynamics and biomechanics. FIG.

[0057] 22A shows echocardiographic images comparing LAA ostium motion in two human patients (Patients 4 and 7) and the soft robotic simulator during atrial systole and diastole, closely matching the dynamics observed in human cases. FIG.22B shows quantitative analysis of LAA ostium contraction in human patients compared with the range achieved by the soft robotic simulator. FIG. 22C shows views of the LAA ostium plane in the simulator (endoscopic camera) and in a human patient (Patient 1, echocardiography) during 9

[0058] 45802538.1 atrial systole and diastole. FIG. 22D shows comparison of LA area changes across porcine, human, and LA simulator models via echocardiographic images in atrial diastole and systole show comparable area changes in all models. The bar graph shows quantitative LA area contraction, with the simulator showing close alignment with both porcine and human data, indicating accurate replication of atrial mechanics. FIG.22E left atrial pressure (LAP) traces in porcine in vivo vs. soft robotic simulator.

[0059] FIG.23 shows an exemplarysoft robotic, patient-specific 3D-printed left atrium (LA) and left ventricle (LV) integrated with the mock circulatory flow loop.

[0060] FIG.24 show an exemplary patient-specific 3D-printed left atrium (LA) and left ventricle (LV) integrated with soft robotic synthetic myocardium, comprising of McKibben-type actuators to drive atrial and ventricular wall mechanics during systole. Photo credit: Jodi Hilton.

[0061] FIG.25 is a schematic of a portable and compact magnetic resonance imaging (MRI) compatible mock circulatory flow loop setup. LA; left atrium, LV; left ventricle, LAA; left atrial appendage, MV; mitral valve, AoV; aortic valve, DAQ; data acquisition. To further examine the distinct contributions of rate irregularity and loss of atrial contraction characteristic of AF, we implemented a 1D lumped parameter model of the left atrium. The model incorporated time-varying elastance and RR interval variability to simulate both normal sinus rhythm (SR) and atrial fibrillation (AF). In the AF condition, atrial elastance was held constant to represent loss of coordinated contraction, and randomized atrial impulses were introduced to replicate irregular conduction. Simulation outputs included time-varying left atrial pressure, volume, and trans-mitral flow. Compared with SR, AF resulted in elevated mean atrial pressures, increased atrial volumes, and impaired atrial emptying during ventricular filling. These hemodynamic alterations translated into reduced ventricular preload and diminished aortic outflow, consistent with a reduction in global cardiac output (data not shown). Together, these findings demonstrate that irregular activation and loss of atrial booster function independently impair systemic performance, even when mean heart rate is matched.

[0062] Detailed Description of the Invention

[0063] Atrial fibrillation (AF) is the most common clinically significant cardiac arrhythmia, affecting more than 33.5 million people globally. In the United States, AF afflicts 9% of the population aged 65 and older. Patients with AF have a five-fold increase in the incidence of stroke, attributed to the dislodgement and subsequent obstruction of a blood vessel by a blood clot. Studies have shown that 90% of these clots originate in the left atrial appendage (LAA), a tubular, blind-ended sac that extends from the left atrium. When the atrium does not contract effectively, as is the case in patients with nonvalvular atrial fibrillation (NVAF), LAA geometry promotes blood stasis and subsequent clot formation. Stroke occurs when these clots embolize and reach the brain.

[0064] Prevention against stroke is a priority in patients with NVAF. Currently, the gold-standard therapy is blood thinner medication, termed oral anticoagulation (OAC), however, patients taking OAC have a five- fold increase in the risk of intracranial hemorrhage, with mortality rates exceeding 50% in most studies. Non-pharmacological strategies that locally prevent the development and 10

[0065] 45802538.1 dislodgcmcnt of clots from the LAA arc an alternative to OAC in patients who arc unable or unwilling to take OAC, have a contraindication to OAC, or are at an elevated risk of major bleeding. Such strategies include surgical excision or exclusion, and minimally invasive occlusion, termed left atrial appendage occlusion (LAAO). Although several endocardial occlusion devices are available on the market, these devices are rigid and only come in standard, pre-defined geometries. These devices are not always amenable to the complex and diverse patient anatomy of the LAA, leading to patient preclusion from treatment, or inconsistent device implantation that may result in peri-device leakage, device related thrombosis (DRT), and device embolization (DE), which are all associated with significantly increased risk of morality and ischemic events.

[0066] The LAA is a highly anatomically variable structure that comes in many shapes and sizes, and displays a variety of convexities, concavities, lobes, and trabeculations. There is significant variation in the size, morphology, and internal anatomy of the LAA between patients, as well as heterogeneity in its interaction with adjacent cardiac and extra-cardiac anatomy. The LAA volume is reported to range from 0.7-9.2 mL, with ostial dimensions ranging from 10-40 mm, and LAA lengths of 16-51 mm. The ostium is typically elliptical in shape with mean long-axis diameters in the range of 16-23 mm and mean short-axis diameters in the range of 10-17 mm. Post-mortem studies have demonstrated the LAA to be a multi-lobed structure in 80% of cases. The diverse LAA anatomy is highly amenable to a personalized medicine approach in which each procedure is precisely tailored for each individual patient. Significant strides in personalized medicine, specifically within the field of structural heart disease (SHD), have been made possible by utilizing techniques such as 3D printing (3DP), computational modeling, and artificial intelligence (Al). 3DP can decrease the early-operator learning curve for new technology adaptation, computational fluid modeling can emulate dynamic physical and physiological properties of cardiac pathophysiology, and application of Al has potential for patient-specific anatomic replica procedural simulation training. Computational modeling is performed using numerical analysis methods such as finite element analysis (FEA) and computational fluid dynamics (CFD). CFD has proven to be a useful complementary technique to predict thrombus risk from LAA morphology and to indicate optimal device sizing and placement for LAA occlusion. However, a major limitation of CFD is the high sensitivity to numerical assumptions, such as boundary conditions and material properties, that are adopted to reduce computational cost and simplify the modeling process.

[0067] Given the anatomical complexity, numerous cardiac imaging techniques are currently used to assess the anatomy and size of the LAA, including two-dimensional transesophageal echocardiography (2D TEE), three-dimensional transesophageal echocardiography (3D TEE), and cardiac computed tomography (CCT). At the core of these technologies is the ability to evaluate the 11

[0068] 45802538.1 LAA in three dimensions. CT virtual and physical 3D imaging -based computational models have been used to simulate device deployment into patient-specific cardiac anatomies to predict optimal placement strategy, and to determine optimal device sizing to reduce the potential for peri-implant leakage.

[0069] Procedures that utilized a 3D printed model based on CT imaging were associated with improved device selection and procedural efficiency (reduced procedure time, reduced anesthesia time, and reduced fluoroscopy time). Devices sized using a model based on CT imaging were also associated with a lower probability of peri-device leakage. While imaging-based simulations and 3D printing have been shown to lead to better interventional planning by improving device sizing, elucidating appropriate landing zones in a patient-specific manner, and improving overall procedural efficiency, they lack the ability to replicate physiologically relevant motion and fluidstructure interactions in a setting amenable to hands-on, real-world training and research.

[0070] Patient-specific 3D printed models of the LAA improve precision in occlusion device sizing and placement, however, these models are static and fail to recapitulate the complex motion, fluid flow, and tissue mechanics of the beating heart. 3D printed models fail to emulate the dynamic physical and physiological principles governing cardiac function. Successful development of LAAO technologies requires appreciation of complex anatomic pathophysiology integrated with occlusion device mechanical deformation properties for a highly sophisticated tissue-fluid-structure interaction.

[0071] Benchtop circulatory flow loops are designed to simulate cardiovascular hemodynamics of physiological systems. To accurately model the human cardiovascular environment without relying on prohibitively expensive and time consuming in vivo large animal studies, three main experimental approaches have been described: the use of synthetic ventricles, the integration of passive excised biological samples into artificial in vitro setups, and the use of ex vivo beating heart models. Synthetic models allow for easily controlled and repeatable experimental conditions but fail to recapitulate the complex intracardiac physiology. High-fidelity intracardiac anatomy is important for evaluating the function and simulating the placement of intracardiac devices, such as LAAO devices. Ex vivo beating heart models provide highly realistic cardiac physiology. However, their price, complexity in model setup, and limited longevity (<1 day) due to muscle stiffening and decay, prevent widespread adaptation. In vitro approaches that use passive biological samples ensure preservation of cardiac anatomy, however, to generate flow, the ventricular chamber must be connected through the apex of the heart to an external pumping system. This pumping mechanism creates and altered, non-physiologic fluid dynamic field inside the left ventricle. Currently, there is

[0072] 12

[0073] 45802538.1 no valid bcnchtop model (synthetic or biological) that accurately mimics the complex motion, fluid flow, and anatomical features of the heart.

[0074] Researchers have tried multiple benchtop models for various types of cardiac simulation. A biohybrid approach that combines both organic and synthetic components. They demonstrate the fabrication of a biomimetic ventricular model that is composed of organic endocardial tissue and an active, soft-robotic myocardium that drives motion of the heart. This model provides representation of the anatomical details of intracardiac structures, while also recreating physiological cardiac motion using soft robotics. While this represents an important advance towards the need for high-fidelity cardiac simulators for pre-clinical intracardiac device testing, this model focuses primarily on ventricular motion and is not designed for modularity in which multiple, patient- specific anatomies can be interchanged and represented. Additionally, this model is time-consuming and technically difficult to manufacture, and necessitates appropriate handling, preservation and storage of the tissue to maintain integrity for simulation.

[0075] While these models excel at capturing singular aspects of cardiac physiology, such as pressure dynamics, motion, or geometric fidelity, they are primarily focused on ventricular function and fail to combine the dynamic physical and physiological principles that govern cardiac function, along with the diverse anatomical variations of the LAA, which together affect intracardiac device performance and post -procedural outcomes. There is a significant need for benchtop models that capture patient heterogeneity, while also replicating physiological hemodynamics and mechanical properties, to facilitate pre-procedural planning and device development and validation in a more accurate, inclusive, and physiologically relevant environment.

[0076] The left atrial cardiac simulator that has been developed captures intra-patient anatomical heterogeneity through a library of diverse left atrial models that can be easily incorporated into a benchtop circulatory flow loop capable of measurable and tunable pressure and flow waveforms. The left atrial cardiac simulator can be used to better understand the hemodynamics of left atrial appendage occlusion (LAAO), serve as a procedural training tool for and help in the development and testing of new interventions and procedures.

[0077] The simulator provides a realistic platform for the testing and validation of medical devices and surgical interventions, specifically targeting the patient variability for left atrial appendage closure, using a patient-specific and interchangeable approach.

[0078] The simulator can replicate the accurate biomechanical and hemodynamic conditions of the left atrium, facilitating the development of interventions for LAA closure and HFpEF treatments and evaluating their effects on global hemodynamics.

[0079] 13

[0080] 45802538.1 I. Definitions

[0081] As used herein, “Biohybrid” refers to a device: containing or composed of both biological and non-biological components, such as possessing a component of biological origin; containing biomaterial and non-biological biomaterial; and / or integrating synthetic material with animal (e.g., human) tissue (such as muscles, nerves, or bone).

[0082] As used herein, “intracardiac” refers to situated within, occurring within, introduced into, or involving entry into the heart, such as intracardiac surgery using an intracardiac catheter.

[0083] As used herein, “endocardial tissue scaffold” refers to the whole heart with removed ventricular myocardial tissues and including intact intracardiac structures, and / or other heart structures such as the atria and major vessels. This includes heart tissues of all forms, not just limited to chemically preserved heart tissues, including biological tissue, chemically preserved or decellularized matrix that retain internal structures of the heart.

[0084] As used herein, “synthetic myocardium” refers to cardiac muscle tissue-mimic or -substitute made out of soft synthetic matrix and / or containing actuatable components, such as pneumatic artificial muscles.

[0085] As used herein, “soft” refers to material mechanical properties with low modulus similar to most biological tissues in the range of 10 kPa to 10 MPa.

[0086] As used herein, “biorobotic hybrid” refer to devices containing or composed of passive biological structures combined with active synthetic materials that can be actuated to produce motion and forces.

[0087] As used herein, physiological hemodynamics means development of intraventricular pressures simultaneously with the reduction of ventricular volumes induced by the robotic actuation of the synthetic myocardium (i.e. high pressure / low volume as opposed to high pressure / high volume).

[0088] As used herein, “complex three-dimensional cardiac motion” and “physiological level of contractile motion” refer to three-dimensional contraction of the synthetic myocardium, resulting in the reduction of volume / volumes in one or more of the ventricular chambers (which may be left or right only).

[0089] II. Compositions

[0090] Left atrium simulators, left heart simulators, and systems thereof have been developed. In some forms, the left atrium simulators and left heart simulators can be derived and segmented from high-resolution, time-resolved dynamic computed tomography (CT) scans, allowing for an accurate representation of both healthy and atrial fibrillation-affected geometries.

[0091] 14

[0092] 45802538.1 A. Left Atrium Simulator

[0093] The left atrium (also referred to herein as “LA”) simulator includes a left atrium structure and one or more soft robotic actuators.

[0094] 1. Left Atrium Structure and left atrial appendage attachment The left atrium structure contains a receiving portion for attaching to a left atrial appendage (also referred to herein as “LAA”) attachment. Each of the one or more soft robotic actuators is attached to and conforms to a region of the left atrium structure. This allows a user to select and attach LAA of variable geometries to the LA structure for testing.

[0095] Optionally, a septal crossing window (see, e.g., Figure 5B, 150) measuring lmm-20 mm, such as about 10mm, in diameter is incorporated into the LA structure at the inferior fossa ovalis. The specific position of the septal crossing window is determined based on input from a clinician, ensuring anatomical relevance and accessibility for device delivery in variable LAA geometries. In some forms, the window can be sealed with a suitable material, such as circular silicone sheets of varying thicknesses and durometers or fitted with a self-sealing port to accommodate delivery sheath insertion.

[0096] In some forms, the left atrium simulator includes the LAA attachment attached to the receiving portion of the LA structure. The LAA attachment can be attached to the receiving portion of the LA structure via any suitable means, such as a pair of mating features incorporated in the attachment end of the LAA attachment and the receiving portion of the LA structure. Tor example, the LAA attachment contains an attachment end (see, e.g., Figure 3E-3H, 121, 121’, 121”, and 121”’) that is configured to mate with and be received by the receiving portion (see, e.g.. Figure 5B, 111,) of the LA structure, such as by having protrusions on the receiving portion that align with and fit in corresponding grooves on the attachment end of the LAA, or vice versa. The LAA attachment then connects to the LA structure by aligning and mechanically pressing together the attachment end of the LAA attachment and the receiving portion of the LA structure. In some other examples, the attachment end and receiving portion are connected to each other by being screwed on using mating threads.

[0097] FIG. 1A-1F are schematics of the location and anatomy of the left atrial appendage (“LAA”), and mechanism of thrombus formation during atrial fibrillation. This shows the anatomical and morphological overview of the LA and LAA, showing healthy and atrial fibrillation (AF) states (1A) and schematics of LAA morphologies (1B-1E). The “chicken wing” morphology (IB) of an LAA with an obvious bend in the proximal or middle part of the dominant lobe is the most common variant found in 48% of patients. The “cactus” LAA (1C) has a dominant central lobe with secondary lobes extending from the central lobe in both superior and inferior directions 15

[0098] 45802538.1 and is found in 30% of patients. The “windsock” morphology (ID) describes an LAA in which one dominant lobe of sufficient length is the primary structure, found in 19% of patients. The “cauliflower” LAA’s (IE) main characteristic is an LAA that has limited overall length with complex internal features, present in just 3% of patients (https: / / journals.viamedica.pl / folia_moiphologica / article / view / 76067).

[0099] Atrial fibrillation (AF) is caused by abnormal electrical activity. Disordered electrical propagation causes disorganized stimulation of the myocardium and subsequent arrhythmic contractions. AF decreases contractility, resulting in blood stasis and diminished peak flow velocities. AF is also associated with endothelial damage, fibrosis, and inflammation, especially within the LAA, which leads to a prothrombotic and hypercoagulable state. This association is consistent with “Virchow’s Triad” (IF), which synthesizes the pathogenesis of clot formation in the LAA in patients with AF: abnormal blood flow (stasis), endocardial dysfunction (vessel wall injury), and altered hemostasis (hypercoagulability).

[0100] FIG. 3A-3D shows the workflow from clinical imaging FIG. 3A to soft, patient-derived LAAs for attachment to rigid model. 3A, Image segmentation, post-processing, and cutting at the ostium plane. 3B, 3D printed mold for silicone casting of LAA geometry. 3C, segmented LA anatomy. 3D, soft, patient-derived LAA models. FIG. 3E is a diagram of a soft robotic driven mock circulatory loop: robotic LV pump; robotic LA / LAA for active emptying. Workflow and design of a soft robotic beating heart simulator for left atrium (LA) and left atrial appendage (LAA) biomechanics replication, targeted for left atrial appendage occlusion (LAAO) and heart failure with preserved ejection fraction (HFpEF) testing, is depicted. This diagrams the workflow from clinical imaging to rigid, patient-derived LA model. This diagram shows image segmentation, postprocessing, and addition of model features for LAA attachment, septal crossing and integration with circulatory flow loop.

[0101] The LA structure and LAA attachment can each be formed of a suitable polymer. Polymers suitable for forming the LA structure and LAA attachment are typically soft and elastomeric, such as a soft and elastomeric resin, polysiloxane, polylactic acid, acrylonitrile butadiene styrene, polyethylene terephthalate glycol, polyamide, polycarbonate, thermoplastic polyurethane, and acrylonitrile styrene acrylate, and combinations thereof. In some forms, the polymers are suitable for use in 3D printing to construct the LA structure and LAA attachment. For example, the polymer forming the LA structure and LAA attachment is polysiloxane.

[0102] The heart simulator is constructed using patient-specific imaging data, allowing for an anatomically accurate model of the left atrium and LAA. This personalized approach ensures that the simulator reflects the wide range of anatomical variations seen in patients, improving the 16

[0103] 45802538.1 relevance of testing. Besides 3D printing, the models of the LA and LAA arc created using simultaneous utilization of additive manufacturing and traditional casting / molding techniques using elastomers (e.g., silicones) or polyvinyl alcohol (PVA) cryogels, to accurately replicate tissue mechanical properties and surface finish.

[0104] 2. Soft Robotic Actuators

[0105] Soft robotic actuators (also referred to herein as “actuators”) are integrated into the LA structure, such as 3D printed LA structure, to mimic the natural beating of the heart. This system is designed to replicate the complex motion and biomechanics of the left atrium, including the nuances of contraction patterns associated with healthy and diseased conditions.

[0106] The actuators can have any suitable shape and size configured for placement at desired regions of the LA structure and optionally LAA attachment. For example, flattened actuators are used for the circumferential arrangement, while cylindrical actuators are used for the helical inner layer.

[0107] Each soft robotic actuator contains three components: a bladder, a tubing, and an expandable braided mesh. The bladder can be formed of a thermoplastic elastomer, such as thermoplastic polyurethanes (TPU), thermoplastic vulcanizates (TPV), styrenic block copolymers (SBC), thermoplastic polyolefins (TPO), copolyester elastomers (COPE), or polyether block amide (PEBA), or a combination thereof. The bladder is typically formed by sealing two polymer layers, such as by heat-sealing two TPE layers. The tubing can be formed of a suitable polymer, such as thermoplastic polyurethane. The tubing is inserted into the bladder and sealed using a suitable adhesive. The braided mesh is optionally coated with a layer of silicone to prevent kinking. The actuator can be assembled using any suitable methods as long as structural integrity is achieved. For example, the actuator is assembled by hand-sewing with thread to ensure structural integrity.

[0108] In some forms, the LA structure, optionally with an LAA attachment attached thereto serves as the anatomical foundation for the integration of soft robotic actuators in alignment with the underlying anatomical features. These actuators placed biomimetically, guided by the native myocardial fiber orientations. For example, actuators are biomimetically placed along the anterior, posterior-inferior, and LAA regions of the LA structure, allowing coordinated contraction, expansion, and wringing motions for realistic atrial and appendage dynamics.

[0109] In some forms, the LA simulator includes at least two soft robotic actuators, where a first soft robotic actuator is attached to and conforms to the anterior region of the left atrium structure, and a second soft robotic actuator is attached to and conforms to the posterior-inferior region of the left atrium structure. In some forms, the first and second soft robotic actuators can attach to each other and optionally form a unitary piece that surrounds the circumference of the anterior and 17

[0110] 45802538.1 posterior-inferior regions. For example, the outer mesh of the first and second soft robotic actuators are connected, with 2 separate interior bladders so they can be individually controlled. Optionally, the LA simulator further includes a third soft robotic actuator that is attached to and conforms to the circumference of the receiving portion for attaching the left atrial appendage attachment thereto; and / or a fourth soft robotic actuator that is attached to and conforms to the circumference of the roof-anterior region of the left atrium structure.

[0111] FIG. 2A shows the segmentation of CT data from patients (2A), soft 3D-printed robotic actuators mimicking LA, LAA wall motion (2B), and the reservoir and the pump (2C). FIG. 2D-2H is a workflow diagram for 3D-CT imaging and segmentation in healthy and AF patients (2D), producing patient-specific LA and LAA geometries (2E) for the simulator design. The need for LAA occlusion in prothrombotic LAA is highlighted, along with the challenges posed by geometric mismatches due to inter-patient variability and LAA trabeculations. Clinical imaging-derived segmentation of LA and LAA structures from CT data, illustrates significant anatomical heterogeneity across patients. This variability is essential to consider in device testing to account for real-world application scenarios. Patient-derived 3D models of the LA and LAA are utilized to inform the structural design of the simulator, creating anatomically realistic geometries for intervention testing. A soft 3D-printed model of the LA and LAA is based on patient data (2G), fabricated to accurately replicate the flexibility and biomechanics of human cardiac tissue.

[0112] Integration of soft robotic actuators within the LA and LAA model (2H), allows for biomimetic wall motion simulation, including atrial systole and diastole.

[0113] As shown in Figure 2H, the left atrium simulator 100 contains a LA structure 110 and a LAA attachment 120 attached to the receiving portion (not shown) of the LA structure. Four soft robotic actuators (collectively referred to as 130) are integrated into the LA structure 110. A first soft robotic actuator 131 is attached to and conforms to the anterior region of the left atrium structure; a second soft robotic actuator 133 is attached to and conforms to the posterior-inferior region of the left atrium structure; a third soft robotic actuator 135 that is attached to and conforms to the circumference of the receiving portion for attaching the left atrial appendage attachment thereto; and a fourth soft robotic actuator 137 that is attached to and conforms to the circumference of the roof-anterior region of the left atrium structure. Although not illustrated in Figure 2H, the number and location of the actuators can be varied to capture the desired heart motion.

[0114] In some forms, the LA structure may contain the LAA feature as a unitary piece with the rest of the LA structure. Such an LA structure can be similarly produced using 3D print technique, optionally based on the imaging results of a patient’ s LA for a patient specific LA structure that includes the patient specific LAA geometry.

[0115] 18

[0116] 45802538.1 The need for LAA occlusion in prothrombotic LAA is highlighted, along with the challenges posed by geometric mismatches due to inter-patient variability and LAA trabeculations. Clinical imaging-derived segmentation of LA and LAA structures from CT data illustrates the significant anatomical heterogeneity across patients. This variability is essential to consider in device testing to account for real-world application scenarios. Patient-derived 3D models of the LA and LAA are utilized to inform the structural design of the simulator, creating anatomically realistic geometries for intervention testing. Soft 3D-printed models of the LA and LAA based on patient data accurately replicates the flexibility and biomechanics of human cardiac tissue. Integration of soft robotic actuators within the LA and LAA model allows for biomimetic wall motion simulation, including atrial systole and diastole.

[0117] 3. Kits

[0118] In some forms, the LA simulator is provided as a kit. The kit can include the LA structure, one or more LAA attachments, and one or more soft robotic actuators, that are provided in separate packages. A user can assemble the LA simulator by selecting an LAA of desired geometry; attaching the selected LAA attachment to the receiving portion of the LA structure; and placing the one or more actuators in the desired regions of the LA structure. Optionally, the kit further includes a user manual and optionally tools for assembling the LA simulator.

[0119] Typically, in the kit, each of the one or more LAA attachments has a different geometry, such as cactus, chicken wing, windsock, and cauliflower. Additional LAA geometries are illustrated in Figure 2J.

[0120] B. Left Heart Simulator

[0121] The LA simulator can be integrated with a left ventricle ( simulator, optionally both segmented from dynamic CT imaging data, to form a left hear simulator that is a comprehensive system capable of reproducing left-sided cardiovascular hemodynamics.

[0122] The left ventricle (also referred to herein as “LV”) simulator contains a left ventricle structure and one or more soft robotic left ventricle actuators. Each of the one or more soft robotic actuators is attached to and conforms to a region of the LV structure. The LV structure can be formed of a suitable polymer, such as any one of those describe above for the LA structure and LAA attachment. For example, the LV structure is formed of a polymer suitable for 3D printing, such as polysiloxane.

[0123] Soft robotic actuators are integrated into the LV structure to capture the key modes of heart conditions. For example, soft robotic actuators are integrated into the LV structure to capture the modes of myocardial deformation, with helical actuators providing ventricular twisting, circumferential actuators driving inward radial contraction, and longitudinal actuators contributing 19

[0124] 45802538.1 to atrioventricular (AV) plane displacement. The soft robotic actuators can be formed as described above for LA simulators and contain the same components, such as a bladder, a tubing, and braided mesh.

[0125] In some forms, the LV simulator includes three or more soft robotic actuators (see, e.g., Figure 17A) integrated therein, where a first actuator is attached to and conforms to the circumference at the base of the LV structure; a second actuator is attached to and conforms to the circumference of the middle section of the LV structure; and a third actuator is attached to and conforms to the circumference of the apical section of the LV structure. Optionally, the LV simulator further includes a fourth, a fifth, and a sixth actuators (see, e.g., Figure 17A), where each of the fourth, fifth, and sixth actuators is attached to and conforms to the surface of the LV structure along a helical axis at about a 60-degree angle from the basal plane of the LV structure.

[0126] An exemplary LV simulator is illustrated in Figure 17A. As shown in Figure 17A, the left ventricle simulator 200 contains a LV structure 210 and soft robotic actuators (collectively referred to as 230) placed thereon. Three circumference actuators, 231, 233, and 235, are placed and conform to the circumference of the LV structure 210. A first actuator 231 is attached to and conforms to the circumference at the base of the LV structure 210; a second actuator 233 is attached to and conforms to the circumference of the middle section of the LV structure 210; and a third actuator 235 is attached to and conforms to the circumference of the apical section of the LV structure 210. A fourth 232, a fifth 234, and a sixth 236 (not shown) longitudinal actuators (see, e.g., Figure 17A) are each attached to and conforms to the surface of the LV structure 210 along a helical axis at about a 60-degree angle from the basal plane of the LV structure 210. Although not illustrated in Figure 17A, the number and location of the circumference actuators and longitudinal actuators can be varied to capture the key modes of desired heart conditions and / or to match the specific ventricular geometry.

[0127] For example, the LV simulator includes two or more soft robotic actuators (see, e.g., Figure 23) integrated therein, where a first actuator is attached to and conforms to the circumference at the base of the LV structure; and a second actuator is attached to and conforms to the circumference of the middle section of the LV structure.

[0128] For example, the LV simulator includes four or more circumferential actuators and one or more extended longitudinal actuators (see, e.g., Figure 24) to accommodate a larger ventricular geometry.

[0129] The LA simulator can be attached to the left ventricle simulator using a pair of mating features, such as by using protrusions and corresponding grooves or mating threads on the connecting portions of the LA simulator and LV simulator, similar to that described above for the 20

[0130] 45802538.1 LA structure and LAA attachment. For example, as shown in Figure 18, the LA simulator 100’ is attached to the LV simulator 200’ by aligning and connecting the two connecting portions 140 and 240 via a pair of mating features. Optionally, for real-time hemodynamic measurements, the LA structure and / or LV structure contains a port configured for connecting to an endoscopic camera, providing visualization of endocardial structures.

[0131] C. Systems

[0132] The simulator is equipped with a fluidic system that replicates the hemodynamic conditions of the left heart. This includes the generation of physiological pressure gradients and flow rates, for evaluating the efficacy of LAA closure devices and HFpEF treatments. This system allows for the assessment of device performance under realistic physiological conditions.

[0133] The system includes the left hear simulator and a mock circulatory flow loop. The left heart simulator is in fluid connection with the mock circulatory flow loop. The mock circulatory flow loop contains hydraulic and mechanical elements, such as tube(s) and chamber(s) and optionally pump(s), that recreates clinically relevant hemodynamics in a dynamic and reproducible manner for benchtop testing.

[0134] Optionally, the system includes one or more valves, such as mechanical mitral valves, mechanical aortic valves, bioprosthetic valves, resistance valves, and / or on-off ball valves, for flow control. In some forms, the system includes a mechanical mitral valve (MV) at the interface of the left atrium simulator and the left ventricle simulator (see, e.g., Figure 5A, 301); and a mechanical aortic valve (AoV) (see, e.g., Figure 5A, 303) at the interface of the left ventricle simulator and a conduit of the mock circulatory flow loop, for simulating unidirectional flow. The type of valves used and placement of the valves in the system can be varied to study different diseases / conditions. For example, the mechanical MV in the system, which precisely replicates mitral inflow during diastole and occlusion during systole, can be substituted with bioprosthetic valves to study various mitral valve disease states and repair or replacement interventions. Optionally, one or more resistance valve(s) are incorporated at desired locations in the system, such as at a location (see, e.g., Figure 5A, 305) in the tube of the mock circulatory flow loop (also referred to herein as “MCFL”).

[0135] Optionally, the system includes one or more sensors, such as one or more pressure sensors and / or one or more flow sensors, placed at desired locations in the left heart simulator and / or the mock circulatory flow loop. For example, one or more pressure sensors and one or more flow sensors are placed in one or more tubes of the MCFL, one or more chambers of the MCFL, the LA simulator, the LAA attachment, and / or the LV simulator to record biphasic pressure waveforms and measure fluid flow at such locations.

[0136] 21

[0137] 45802538.1 Optionally, the system includes an electro-pneumatic control unit that allows synchronization of atrial and ventricular contraction at clinically relevant heart rates.

[0138] III. METHODS OF USING THE HEART SIMULATORS AND SYSTEMS

[0139] A. Device Testing

[0140] The simulator provides a platform for preclinical testing of cardiac devices, such as LAA closure mechanisms (including occluders and sealants), and improving cardiac diastolic dysfunction and the hemodynamics of HFpEF under realistic conditions and investigate their impact on cardiac function. This can significantly reduce the development cycle time and improve the safety and efficacy of new designs. Clinically available devices that are inserted into the LAA to reduce the risk of stroke are shown in FIG.4A-4E. These include the WATCHMAN FLX (4A), WAVECREST, AMULET (4B), AMPLATZER (4C), ULTRASEED (4D), and LAMB RE (4E). These are implanted through a vein in the groin using a catheter and then act by bblocking theopening into the LAA. These are not patient specific but generalized to treatment different defects associated with LAA. The device described herein provides a means to optimize efficacy and selection of these devices for patient specific defects.

[0141] In some forms, the method of using the left atrium simulator or the left heart simulator for testing a medical device includes: (i) implanting the medical device at a desired location in the LA structure and / or LV structure; (ii) optionally selecting a left atrial appendage attachment having desired shape and attaching the selected left atrial appendage attachment to the receiving portion of the left atrium structure; (iii) connecting the left atrium simulator or left heart simulator to a mock circulatory flow loop to form a system; and (iv) running the system. When the left atrium simulator includes the LAA as part of a unitary LA structure, then step (ii) is omitted.

[0142] Optionally, the method further includes assembling the left atrium simulator or left heart simulator. For example, the components of a left atrium simulator, including an LA structure, one or more actuators, and optionally one or more LAA attachments, are provided as separate parts; a user can place the actuators in the desired regions of the LA structure and optionally selecting and attaching an LAA attachment to the receiving portion of the LA structure (when the LAA is not part of an unitary LA structure). For example, for assembling the left hear simulator, a user connects the LA structure to the LV structure and then places the actuators in the desired regions of the LA structure and LV structure, or vice versa.

[0143] Typically, during step (iv), the user first starts the flow of a fluid through the system and then tunes one or more operation parameters to match a desired physiological condition. For example, the user tunes the flow pressure, flow resistance, and / or flow rate in the system to mimic that under a disease condition of a patient. Once a desired condition is established, the user can 22

[0144] 45802538.1 measure one or more system parameters for evaluating the performance of the medical device, such as fluid flow, pressure contraction, position of the medical device, and / or leakage around the medical device.

[0145] When the system including the simulator(s) (e.g., LA simulator, LV simulator, or left heart simulator) and the mock circulatory flow loop already connected and the system provided to a user as a complete system, then the user can (i) implanting the medical device at a desired location in the left atrium structure and / or the left ventricle structure; (ii) optionally selecting a left atrial appendage attachment having desired shape and attaching the selected left atrial appendage attachment to the attachment portion of the left atrium structure; and (iii) running the system. Optionally, the method further includes connecting the system to a electro-pneumatic control unit and / or connecting an endoscopic camera to the LA and / or LV structure to visualize the LA-facing portion of the device, such as for assessing appropriate device positioning and peri-device leak.

[0146] B. Surgical Planning and Medical Training

[0147] The simulator can be used for pre-operative planning and surgical training, allowing clinicians to practice interventions on anatomically accurate heart models that replicate the patient's specific conditions. The simulator can serve as an educational tool for medical professionals, offering hands-on experience with cardiac interventions in a risk-free environment. The simulators and system can be used for testing medical devices for a variety of cardiovascular procedures, such as atrial ablation, different types of left atrial appendage occlusion, left atrial appendage ligation, and mitral valve procedures.

[0148] In some forms, the method of using the left atrium simulator or the left heart simulator for pre-procedure testing of a medical device includes: (i) imaging the patient’s heart, particularly the patient’ s left atrial appendage; (ii) selecting a LAA attachment having a shape that generally corresponds with the shape of the image of the patient's left atrial appendage; (iii) implanting the medical device at a desired location in the LA structure and / or LV structure; (iv) attaching the selected LAA attachment to the receiving portion of the left atrium structure; (v) connecting the left atrium simulator to a mock circulatory flow loop to form a system; and (vi) running the system. When the left atrium simulator includes the LAA as part of a unitary LA structure, then steps (ii) and (iv) are omitted and the LA structure, including the LAA is provided to the user as a unitary piece.

[0149] Optionally, the method further includes assembling the left atrium simulator or left heart simulator, as described above.

[0150] Typically, during step (vi), the user first starts the flow of a fluid through the system, and then tunes one or more operation parameters to match a desired physiological condition. For 23

[0151] 45802538.1 example, the user tunes the flow pressure, flow resistance, and / or flow rate in the system to mimic that under a disease condition of a patient. Once a desired condition is established, the user can measure one or more system parameters for evaluating the performance of the medical device, such as fluid flow, pressure contraction, position of the medical device, and / or leakage around the medical device.

[0152] When the system including the patient specific simulator(s) (e.g., LA simulator, LV simulator, or left heart simulator, selected based on imaging results) and the mock circulatory flow loop already connected and the system provided to a user as a complete system, then the user can (i) implanting the medical device at a desired location in the LA structure and / or the LV structure; (ii) optionally attaching a selected LAA attachment to the receiving portion of the LA structure; and (iii) running the system. Optionally, the method further includes connecting the system to a electropneumatic control unit and / or connecting an endoscopic camera to the LA and / or LV structure to visualize the LA-facing portion of the device, such as for assessing appropriate device positioning and peri-device leak.

[0153] C. Research and Development

[0154] This technology facilitates research into the biomechanics and hemodynamics of the left atrium and the development of new treatments for stroke prevention and advancing the understanding of HFpEF.

[0155] D. Patient-Specific Modeling

[0156] By utilizing patient-specific clinical imaging data, the simulator offers unmatched anatomical accuracy, enhancing the reliability of device testing and intervention planning.

[0157] E. Realistic Biomechanical and Hemodynamic Replication

[0158] The integration of soft robotics and fluidic systems allows for a more complete representation of heart function, far surpassing the capabilities of traditional models.

[0159] FIG. 5A is schematic of circulatory flow loop and incorporation of patient-derived model to achieve LA cardiac simulator. Incorporation of patient-derived model into circulatory flow loop to achieve left atrial cardiac simulator. FIG.5B is a schematic of a patient-derived 3DP LA model used in LA cardiac simulator. FIG. 5C shows the cauliflower LAA model coupled to rigid patient-derived LA model. Simulator has four inflows (RI PV, RS PV, LS PV, LI PV) and one outflow (MV). Schematic of complete left atrial cardiac simulator. Simulator contains pulse duplicator pump, patient-derived LA model, mitral valve, tunable resistance and compliance elements, pressure sensors, and flow probes. LA = left atrium; RS PV = right superior pulmonary vein; RI PV = right inferior pulmonary vein; LI PV = left inferior pulmonary vein; LS PV = left superior pulmonary vein; MV = mitral valve. FIG. 5D-5G. Measured values in LA cardiac simulator: 5D,

[0160] 24

[0161] 45802538.1 points measured, 5E, Wigger's Diagram (pressure over time); 5F, flow (L / min over time, seconds);

[0162] 5G, pressure (mmHg) over time, seconds.

[0163] F. Versatility

[0164] The modular, interchangeable design allows the simulator to be customized for a wide range of cardiac conditions and interventions, making it a valuable tool across various stages of device development and medical training. The LAA anatomies can be easily interchanged, enabling testing in a diverse spectrum of geometries and pathologies. The hemodynamics of the circulatory loop and be finely tuned to allow testing in both physiological and pathological conditions.

[0165] FIG. 6A is a diagram of tunable parameters in the system including tunable pulsatile pump in LA cardiac simulator. The data obtained in the system is shown as the Wiggers diagram (6B), pressure over time (6C), flow over time (6D) and pressure over time (6E).

[0166] FIG. 7A-7D show the tunable compliance in an LA cardiac simulator 7A, graph of LA pressure over time for LAA1 decreasing compliance, 7B, graph of LA pressure over time for LAI for C1-C57, 7C, graph of pulse pressure for C1-C5, 7D. SIL30 and Ecoflex 30 LAA1 pressure over C1-C5, 7E.and graph of pulse pressure for SIL30 and Ecoflex 30 LAAL 7F.

[0167] FIG. 8A-8E. Tunable LAA geometry in LA cardiac simulator, diagram of tunable parameters LAA1 parameters, 8A, diagram of tunable parameters LAA4, 8B; graphs of pressure over time for LAA1, 8C and LAA4, 8D; graphs of flow over time, 8E, LAA11 andLAA4, 8F.

[0168] G. Accurate Imaging and Surface Properties

[0169] The use of PVA cryogels in the simulator to replicate the imaging and surface characteristics of tissues offers realistic models for medical training and device testing. These cryogels precisely mimic the visualization of human LA / LAA tissues using various medical imaging technologies, like ultrasound and MRI, facilitating the practice of complex image-guided procedures in the LA / LAA simulator. Their engineered surface smoothness and friction mimic the physical interaction between LA / LAA tissues and medical devices accurately, including clinically relevant friction levels and preventing inconsistent movement during device deployment. The mechanical properties of PVA cryogels are adjustable in the manufacturing process to match those of the target tissues, ensuring the devices' conformal deployment. This is essential for evaluating device performance and safety in a controlled, realistic setting.

[0170] By providing robust preclinical data on the device's performance, the left atrium simulator not only can aid in the design optimization process but can also support regulatory submissions by offering supplementary evidence of efficacy and safety, thereby smoothing the path from innovation to patient care.

[0171] 25

[0172] 45802538.1 The bcnchtop beating heart simulator represents a significant advancement in cardiac research and device testing. By closely mimicking the actual conditions of the left atrium and LAA, this technology offers a powerful platform for the development of more effective treatments for HFpEF and the enhancement of LAA closure techniques.

[0173] The disclosed compositions and methods can be further understood through the following numbered paragraphs.

[0174] Paragraph 1. A patient-specific benchtop beating heart simulator comprising soft robotics to mimic the motion and biomechanics of the left atrium and a patient specific left atrium formed of synthetic polymer.

[0175] Paragraph 2. The heart simulator of paragraph 1 mimicking the motion and biomechanics of the left atrial appendage.

[0176] Paragraph 3. The heart simulator of paragraph 2 mimicking conditions associated with heart failure with preserved ejection fraction (HFpEF).

[0177] Paragraph 4. The heart simulator of any of paragraphs 1-3 made using three dimensional printing or molding to make the patient-specific left atrium.

[0178] Paragraph 5. The heart simulator of any of paragraphs 1-4 wherein Left Atrial Appendage (LAA) models having different geometries and / or material properties can be tested.

[0179] The disclosed compositions and methods can be further understood through the following numbered paragraphs.

[0180] Paragraph 1. A left atrium simulator for a patient comprising:

[0181] a left atrium structure formed of a first polymer; and

[0182] one or more soft robotic actuators,

[0183] wherein the left atrium structure comprises a receiving portion for attaching to a left atrial appendage attachment,

[0184] wherein each of the one or more soft robotic actuators is attached to and conforms to a region of the left atrium structure.

[0185] Paragraph 2. The left atrium simulator of paragraph 1, further comprising the left atrial appendage attachment formed of a second polymer, wherein the left atrial appendage attachment is attached to the receiving portion.

[0186] Paragraph 3. The left atrium simulator of paragraph 1 or 2, comprising two or more soft robotic actuators, wherein a first soft robotic actuator is attached to and conforms to the anterior region of the left atrium structure, and wherein a second soft robotic actuator is attached to and conforms to the posterior-inferior region of the left atrium structure, optionally wherein

[0187] 26

[0188] 45802538.1 Paragraph 4. The left atrium simulator of paragraph 3, comprising a third soft robotic actuator that is attached to and conforms to the circumference of the receiving portion for attaching the left atrial appendage attachment thereto.

[0189] Paragraph 5. The left atrium simulator of paragraph 3 or 4, comprising a fourth soft robotic actuator that is attached to and conforms to the circumference of the roof-anterior region of the left atrium structure.

[0190] Paragraph 6. The left atrium simulator of paragraph 1 or 2, comprising four soft robotic actuators, wherein:

[0191] a first soft robotic actuator is attached to and conforms to the anterior region of the left atrium structure,

[0192] a second soft robotic actuator is attached to and confomis to the posterior-inferior region of the left atrium structure,

[0193] a third soft robotic actuator that is attached to and conforms to the circumference of the receiving portion for attaching the left atrial appendage attachment thereto, and

[0194] a fourth soft robotic actuator that is attached to and conforms to the circumference of the roof-anterior region of the left atrium structure.

[0195] Paragraph 7. The left atrium simulator of any one of paragraphs 3-6, the first and second soft robotic actuators are attached to each other and optionally form a unitary piece that surrounds the circumference of the anterior and posterior-inferior regions.

[0196] Paragraph 8. The left atrium simulator of any one of paragraphs 2-7, wherein the first and second polymers are the same and each of the first and second polymers is polysiloxane.

[0197] Paragraph 9. The left atrium left atrium simulator of any one of paragraphs 1-8, wherein each of the one or more soft robotic actuators comprises a bladder, a tubing, and a mesh.

[0198] Paragraph 10. A kit for assembling a left atrium simulator for a patient, comprising:

[0199] a left atrium structure formed of a first polymer;

[0200] one or more left atrial appendage attachments; and

[0201] one or more soft robotic actuators,

[0202] wherein the left atrium structure comprises a receiving portion,

[0203] wherein each of the one or more left atrial appendage attachments comprises an attachment end configured to be received by the receiving portion, and

[0204] wherein each of the one or more left atrial appendage attachments has a different shape. Paragraph 11. The kit of paragraph 10, wherein the one or more left atrial appendage attachments have a shape selected from the group consisting of cactus, chicken wing, windsock, and cauliflower.

[0205] 27

[0206] 45802538.1 Paragraph 12. The kit of paragraph 10 or 11, wherein each of the one or more soft robotic actuators has a dimension and shape configured to attach to a region of the left atrium structure.

[0207] Paragraph 13. A left heart simulator for a patient comprising:

[0208] the left atrium simulator of any one of paragraphs 1-9; and

[0209] a left ventricle simulator,

[0210] wherein the left ventricle simulator comprises:

[0211] a left ventricle structure: and

[0212] one or more soft robotic left ventricle actuators,

[0213] wherein each of the one or more soft robotic actuators is attached to and conforms to a region of the left ventricle structure,

[0214] wherein the left atrium simulator is attached to the left ventricle simulator.

[0215] Paragraph 14. The left heart simulator of paragraph 13, comprising a first soft robotic left ventricle actuator that is attached to and conforms to the circumference at the base of the left ventricle structure, a second soft robotic left ventricle actuator that is attached to and conforms to the circumference of the middle section of the left ventricle structure, and a third soft robotic left ventricle actuator that is attached to and conforms to the circumference of the apical section of the left ventricle structure.

[0216] Paragraph 15. The left heart simulator of paragraph 13 or 14, further comprising a fourth, a fifth, and a sixth soft robotic left ventricle actuator, wherein each of the fourth, fifth, and sixth soft robotic left ventricle actuators is attached to and conforms to the surface of the left ventricle structure along a helical axis at about a 60-degree angle from the basal plane of the left ventricle structure.

[0217] Paragraph 16. The left heart simulator of any one of paragraphs 13-15, wherein the left ventricle structure is fomied of a third polymer.

[0218] Paragraph 17. A system comprising:

[0219] the left heart simulator of any one of paragraphs 13-16; and

[0220] a mock circulatory flow loop,

[0221] wherein the left heart simulator is in fluid connection with the mock circulatory flow loop. Paragraph 18. The system of paragraph 17, further comprising a mechanical mitral valve (MV) at the interface of the left atrium simulator and the left ventricle simulator; and a mechanical aortic valve (AoV) at the interface of the left ventricle simulator and a conduit of the mock circulatory flow loop.

[0222] Paragraph 19. The system of paragraph 17 or 18, wherein the mock circulatory flow loop comprises one or more pressure sensors and / or one or more flow sensors.

[0223] 28

[0224] 45802538.1 Paragraph 20. The system of any one of paragraphs 17-19, wherein the mock circulatory flow loop comprises an electro-pneumatic control unit.

[0225] Paragraph 21. A method of using the left atrium simulator of any one of paragraphs 1-9 for testing a medical device, comprising:

[0226] (i) implanting the medical device at a desired location in the left atrium structure;

[0227] (ii) optionally selecting a left atrial appendage attachment having desired shape and attaching the selected left atrial appendage attachment to the receiving portion of the left atrium structure;

[0228] (iii) connecting the left atrium simulator to a mock circulatory flow loop to form a system; and

[0229] (iv) running the system.

[0230] Paragraph 22. A method of using the left heart simulator of any one of paragraphs 13-16 for testing a medical device, comprising:

[0231] (i) implanting the medical device at a desired location in the left atrium structure and / or the left ventricle structure;

[0232] (ii) optionally selecting a left atrial appendage attachment having desired shape and attaching the selected left atrial appendage attachment to the receiving portion of the left atrium structure;

[0233] (iii) connecting the left heart simulator to a mock circulatory flow loop to form a system; and

[0234] (iv) running the system.

[0235] Paragraph 23. A method of using the system of any one of paragraphs 17-20 for testing a medical device, comprising:

[0236] (i) implanting the medical device at a desired location in the left atrium structure and / or the left ventricle structure;

[0237] (ii) optionally selecting a left atrial appendage attachment having desired shape and attaching the selected left atrial appendage attachment to the receiving portion of the left atrium structure; and

[0238] (iii) running the system.

[0239] Paragraph 24. A method of using the left atrium simulator of any one of paragraphs 1-9 for preprocedure testing of a medical device for a patient, comprising:

[0240] (i) imaging the patient’s heart, particularly the patient’s left atrial appendage;

[0241] (ii) selecting a left atrial appendage attachment having a shape that generally corresponds with the shape of the image of the patient’s left atrial appendage;

[0242] 29

[0243] 45802538.1 (iii) implanting the medical device at a desired location in the left atrium structure;

[0244] (iv) attaching the selected left atrial appendage attachment to the receiving portion of the left atrium structure;

[0245] (v) connecting the left atrium simulator to a mock circulatory flow loop to form a system; and

[0246] (vi) running the system.

[0247] Paragraph 25. A method of using the left heart simulator of any one of paragraphs 13-16 for preprocedure testing of a medical device for a patient, comprising:

[0248] (i) imaging the patient’s heart, particularly the patient’s left atrial appendage;

[0249] (ii) selecting a left atrial appendage attachment having a shape that generally corresponds with the shape of the image of the patient’s left atrial appendage;

[0250] (iii) implanting the medical device at a desired location in the left atrium structure and / or the left ventricle;

[0251] (iv) attaching the selected left atrial appendage attachment to the receiving portion of the left atrium structure;

[0252] (v) connecting the left heart simulator to a mock circulatory flow loop to form a system; and (vi) running the system.

[0253] Paragraph 26. A method of using the system of any one of paragraphs 17-20 for pre -procedure testing of a medical device for a patient, comprising:

[0254] (i) imaging the patient’s heart, particularly the patient’s left atrial appendage;

[0255] (ii) selecting a left atrial appendage attachment having a shape that generally corresponds with the shape of the image of the patient’s left atrial appendage;

[0256] (iii) implanting the medical device at a desired location in the left atrium structure and / or the left ventricle;

[0257] (iv) attaching the selected left atrial appendage attachment to the receiving portion of the left atrium structure; and

[0258] (v) running the system.

[0259] Paragraph 27. The method of any one of paragraphs 21-26, wherein the system running step comprises (a) starting flow of a fluid through the system; (b) tuning one or more operation parameters to match desired physiological conditions; and (c) measuring one or more system parameters.

[0260] Paragraph 28. The method of paragraph 27, wherein the one or more operation parameters comprise flow pressure and / or flow resistance.

[0261] 30

[0262] 45802538.1 Paragraph 29. The method of paragraph 27 or 28, wherein the one or more system parameters comprise fluid flow, pressure contraction, position of the medical device, and / or leakage around the medical device.

[0263] Paragraph 30. The method of any one of paragraphs 24-29, wherein the procedure is atrial ablation, left atrial appendage occlusion, left atrial appendage ligation, or mitral valve procedure.

[0264] More specific exemplary compositions and methods are described in the Examples below.

[0265] Examples

[0266] Example 1: Manufacture and Testing Rigid Patient-Derived LA Model

[0267] Material and Methods

[0268] Rigid Patient-Derived LA Model:

[0269] An open-source database of 100 MRI scans from patients with atrial fibrillation (AF)30was reviewed to select a single patient geometry (left atrium: LA) for modification. The dataset included 43 paroxysmal AF, 41 persistent AF, and 16 longstanding persistent AF cases, resulting in 100 patient-specific models. All subjects provided informed consent, and the models were irreversibly anonymized to facilitate confidentiality before being made publicly accessible through the database. The chosen geometry exhibited characteristic LA dilation, a common morphological feature in patients with chronic AF

[0031] . Selection criteria included an LA volume greater than 10 mL, representative of patients with AF and cardioembolic stroke

[0032] , and a long-axis dimension exceeding 32 mm to allow sufficient space for coupling with a library of LAAs of varying sizes. The selected LA geometry was modified in CAD software (MESHMIXER®) to isolate the LA remove the existing LAA. A custom connector (fitting) that allows for interchangeable attachment of multiple LAAs with ostium sizes encompassing about over 95% of reported LLA anatomies (i.e., from 14-34 mm) was created. The connector’s dimensions were selected to accommodate varying ostial diameters while maintaining compatibility with the underlying LA anatomy. To facilitate procedural simulation, a septal crossing window measuring 10 mm in diameter was incorporated into the model at the inferior fossa ovalis. Its position was determined based on input from a clinician, ensuring anatomical relevance and accessibility for device delivery in variable LAA geometries. The window can be sealed with circular silicone sheets of varying thicknesses and durometers or fitted with a self-sealing port to accommodate delivery sheath insertion. Additional modifications included the incorporation of 3 / 8 inch OD barbed connectors (5218K748, McMaster) at the pulmonary veins and a 3 / 4-inch OD barbed connector (5218K75, McMaster) at the mitral annulus to allow integration into the benchtop circulatory flow loop. The resulting model was 3D printed on a Connex500 (Stratasys, Eden Prairie, Minnesota) inkjet-based multi-material 3D printer from a polypropylene-like photopolymer resin (RGD450) with a wall thickness of 1.5 mm.

[0270] 31

[0271] 45802538.1 Soft Patient-Derived LAA Models:

[0272] Patient- specific CT scans were segmented in CAD software (Meshmixer), isolating the LAA. Four patient-specific LAA geometries were selected from MRI data to represent a range (captured) a range of anatomical variability, including both heterogenous patient anatomy and challenging to occlude anatomy (LAA 1, LAA 2, LAA 3, LAA 4). LAA geometries were isolated using 3D Slicer and refined in Meshmixer, permitting preservation of the ostial landing zone important for device deployment. The geometries were chosen in collaboration with clinicians and industry partners, incorporating criteria such as ostial dimensions, length, and overall morphology. The final library included geometries spanning 19.2 to 32.8 mm in ostial diameter and 23.5 to 44.8 mm in length. These dimensions reflect clinically relevant variations and were chosen to encompass both typical and anatomically complex cases to support robust device evaluation and procedural simulation. Each LAA geometry was used to make a 3D printed mold, having a positive core and a negative shell, fabricated using CAD software (SOLIDWORKS®). These molds facilitated silicone casting, with 2mm wall thickness, ECOFLEX 00-30, selected for its translucency and rapid fabrication process. Tensile testing guided wall thickness to match the elastic modulus of the native LAA tissue

[0028] . LAA geometry 1 was also 3D printed directly using a stiffer silicone material at 1.5 mm wall thickness (SIL30, Carbon) to mimic pathological compliance changes observed in patients with AF [34, 35], LAAs were integrated onto the rigid 3DP LA model using the custom fitting (connector). Following attachment, LAAs were secured with a silicone rubber adhesive (SIL-POXY®, SMOOTH-ON®), to facilitate maintenance of a watertight seal.

[0273] Mock Circulatory Flow Loop-.

[0274] A robotic LV pump and a robotic LA / LAA for active emptying was developed. Workflow and design of a soft robotic beating heart simulator for left atrium (LA) and left atrial appendage (LAA) biomechanics replication, targeted for left atrial appendage occlusion (LAAO) and heart failure with preserved ejection fraction (HFpEF) testing. The need for LAA occlusion in prothrombotic LAA is important, but generally has challenges due to geometric mismatches due to inter-patient variability and LAA trabeculations. Clinical imaging-derived segmentation of LA and LAA structures from CT data illustrated significant anatomical heterogeneity across patients. This variability is important to consider in device testing to account for real-world application scenarios.

[0275] 3D-CT imaging and segmentation in healthy and AF patients produces patient-specific LA and LAA geometries for the simulator design. Patient-derived 3D models of the LA and LAA are utilized to inform the structural design of the simulator, creating anatomically realistic geometries for intervention testing. Soft 3D-printed models of the LA and LAA based on patient data were fabricated to accurately replicate the flexibility and biomechanics of human cardiac tissue.

[0276] 32

[0277] 45802538.1 Integration of soft robotic actuators within the LA and LAA model allowed for biomimetic wall motion simulation, including atrial systole and diastole. The model mimicked the dynamic motion required for realistic LAAO device testing. Schematic of the soft robotic heart simulator setup, containing a soft robotic left ventricle (LV) and soft robotic LA / LAA for active emptying. Insets displayed the simulator’s capabilities, including biomimetic wall motion, real-time echocardiography imaging capability, and LAAO device testing. The simulator also facilitated hemodynamic simulation of the left-sided cardiac circulation and velocity-encoded MRI for flow assessment.

[0278] The LA model and interchangeable LAAs were integrated into a benchtop circulatory flow loop designed to mimic physiological and pathological hemodynamics. The system included: (1) a pulsatile pump (BDC Laboratories Pulse Duplicator 1100 or Harvard Apparatus pump, used interchangeably) capable of variable stroke volumes and heart rates; (2) an artificial mechanical valve (Masters Series Mechanical Heart Valve, Abbott) in the mitral position; and (3) tunable resistance valves and compliance chambers mimicking pulmonary and systemic vasculature. Differences between pump type may involve reconfiguration of the mitral valve, resistance valves, and compliance chambers to account for variations in system pulsatility. Resistance and compliance elements between the LA and mitral valve mitigated non-physiological suction artifacts from the pump. All tubing was Tygon S3™ B-44-3, with 3 / 8-inch ID tubing for the pulmonary veins (6516T27, McMaster) and 3 / 4-inch ID tubing for the systemic circulation (6516T33, McMaster). A 1:4 straight- flow rectangular manifold (1023N244, McMaster) distributed flow to the four pulmonary veins. Pressure was measured at the pulmonary vein inlets, within the LA, and within the LAA using PendoTech pressure sensors (PRESS-S-000, PendoTech). For pressure measurements within the LA and LAA, the sensors were connected to 5F umbilical vessel catheters (Cardinal Health) via Luer locks and positioned in the respective chambers to allow accurate readings. Flow was quantified using inline Transonic flow probes (ME10PXN, Transonic) coupled to a Transonic flow console (T403 with TS410 modules, Transonic). Data were acquired with a multi-channel data acquisition (DAQ) system (DewesoftX or LabChart, depending on setup) and exported for analysis in MATLAB (MathWorks).

[0279] The models were integrated into a benchtop circulatory flow loop with a pulse duplicator (BDC Laboratories), an artificial mechanical valve in the mitral position (Masters Series Mechanical Heart Valve, Abbott), and tunable resistance valves and compliance chambers to mimic pulmonary and systemic vasculature. Pressure and flow were measured at the four pulmonary vein inlets, the left atrium, and the simulated left ventricle and aorta (PendoTech Pressure Sensors; Transonic Flow Probes). Total cardiac output was measured by the pulse duplicator (BDC

[0280] 33

[0281] 45802538.1 Laboratories). Data were compiled in a multi-channel input data acquisition (DAQ) system (DEWEsoft) and viewed in the corresponding DAQ software (DewesoftX). Data were exported as.txt files and analyzed in MATLAB (Math Works).

[0282] Left Atrial Appendage Occlusion:

[0283] Hemodynamic Changes: To investigate acute hemodynamic changes following left atrial appendage occlusion (LAAO), pressure and flow were measured before and after device deployment using a 24mm Watchman FLX™ (Boston Scientific). The device size was selected according to manufacturer guidelines based on LA A dimensions. To simulate the hemodynamic changes associated with surgical excision, complete occlusion, or device endothelialization (i.e., scenarios in which the LAA is entirely removed or isolated from the LA), pressure and flow were measured after placement of an acrylic cap over the ostium or deployment of an impermeable occlusion device (24 mm coated WATCHMAN FLX™, Boston Scientific). The impermeable device contained a WATCHMAN FLX™ with a silicone coating on the polyethylene terephthalate (PET) covering to prevent flow through the fabric to mimic device endothelialization. The WATCHMAN device size was selected according to manufacturer guidelines based on LAA dimensions. Devices were deployed manually prior to coupling the LAA to the LA model to facilitate the desired positioning for accurate hemodynamic measurements. Changes in LA pulse pressure were used as a quantifiable measure of occlusion, capturing the effect of device placement on LA compliance. Additionally, changes in LA inflow and outflow, expressed as volume per beat, were assessed to quantify the impact of occlusion on LA compliance. These specific pressure and flow metrics are not traditional clinical measurements for LAAO but are employed here as research tools to assess and quantify occlusion in a controlled experimental setting.

[0284] Device Testing: To investigate the impact of device material and assess the degree of “occlusiveness” and peri-device leak (PDL), two identically sized devices were tested: a normal (standard) WATCHMAN FLX™ with a permeable PET fabric cover, and a WATCHMAN FLX™ with an impermeable silicone coating applied on the PET fabric cover (Figure 12A). Following device deployment, green dye was injected into the LAA into the LAA, distal to the device while pressure was monitored in the LAA and dye washout was monitored visually via video capture (iPhone 14). Following dye injection, the time to return to baseline pressure within the LAA before dye injection was used as a surrogate for degree of occlusiveness. An endoscopic camera was used to visualize the LA-facing portion of the device. Captured videos were processed frame-by-frame, focusing on a central region of interest (ROI) within the LAA distal to the device. Dye intensity within the ROI was quantified using a color thresholding method in HSV (Hue, Saturation, Value) color space. Intensity values were normalized, with pre-injection baseline intensity set to 0.0 and 34

[0285] 45802538.1 the maximum intensity following dye injection set to 1.0. The time to achieve a 40% reduction in intensity (color intensity = 0.6) was used as a surrogate marker for occlusion effectiveness. An endoscopic camera (SpyGlass DS Direct Visualization System, Boston Scientific) was employed to visualize the LA surface of the device, providing additional qualitative data on device position and potential PDLs. Devices were deployed manually prior to coupling the LAA to the LA model to facilitate desired placement for accurate measurement of hemodynamic effects. It is important to note that dye clearance time is not a clinical marker of occlusion and would not typically be measured in practice. Rather, it serves as a quantifiable surrogate metric for use in this simulatorbased research setting, thereby allowing standardized comparisons of occlusion effectiveness under controlled conditions.

[0286] Imaging Compatibility: To evaluate imaging compatibility and demonstrate the utility of imaging modalities as a means for assessing occlusion and successful device placement, devices (WATCHMAN FLX™, Boston Scientific) were deployed and visualized using three imaging techniques: an endoscopic camera (1,080-P HD endoscopic camera, NIDADE; 30 frames per second), intracardiac echocardiography (ICE; AcuNav Ultrasound Catheter), and standard echocardiography (Philips Epiq CVx Cardiovascular Ultrasound System with XL14-3 and X5-1 transducers). For direct visualization, an endoscopic camera was used to assess the LA-facing portion of the device (i.e., the LA surface of the device), confirming appropriate positioning and peri-device leak. PDLs were visualized by injecting colored dye behind the device; the dye’s passage through gaps or around the device was directly observed via the endoscopic camera. ICE was utilized to visualize and monitor dye injection into the LAA distal to the device, with dye shooting through the permeable fabric towards the ICE probe positioned within the LA. Flow through the permeable fabric was observed as flow jets directed toward the ICE probe positioned within the LA. Standard echocardiography (2D B-mode and PW mode) was used to measure peak velocity at the ostium (or ostial) edge before and after device placement, as well as during deliberate device mispositioning. To facilitate ultrasound propagation, the model was submerged in a water bath during echocardiographic image acquisition. The ultrasound probe was placed directly on the model, and images were obtained by a clinician to obtain desired views for analysis.

[0287] Deliberate device misplacement, characterized by significant canting of the device, was performed to induce PDL. This approach allowed a clear distinction between well-placed and mispositioned devices. PDL was quantitatively assessed using echocardiography, with data analyzed in Q-Vue 2.2 (Philips), demonstrating the capability of standard echocardiography to provide measurable data in this research context. These imaging modalities were not generally used in their traditional clinical

[0288] 35

[0289] 45802538.1 capacities. Instead, they were adapted in this research setting to evaluate and quantify occlusion, device placement, and PDL under controlled, simulated conditions.

[0290] Procedural Training: To demonstrate the simulator’s capability for procedural training and simulation, an experienced operator repeatedly deployed and recaptured the device (Watchman FLX™, Boston Scientific) under direct visualization. Deployment and recapture were performed using the Access System (Access Sheath and Dilator, Boston Scientific) and Delivery System (Delivery Catheter and Closure Device, Boston Scientific). Each deployment was assessed based on the PASS™ (Position, Anchor, Seal, Size) Criteria established by Boston Scientific, ensuring adherence to standard clinical protocols (Table 1).

[0291] Table 1: Endovascular Left Atrial Appendage Closure Devices

[0292] Device Name Company Design Device Sizes (mm)

[0293] Single-lobe occluder. 21, 24, 27, 30, 33 Nitinol frame with PET

[0294] WATCHMANa'bBoston Scientific

[0295] membrane.

[0296] 10 fixation anchors.

[0297] Single-lobe occluder. 20, 24, 27, 31, 35 Nitinol frame with PET

[0298] WATCHMAN FLXa’bBoston Scientific

[0299] membrane.

[0300] 12 fixation anchors.

[0301] Lobe and disc (polyester 16, 18, 20, 22, 24, 26, 28, Amplatzer Cardiac Plug mesh in both) nitinol mesh 30

[0302] Abbott

[0303] (ACP)bconstruct.

[0304] Stabilizing wires.

[0305] Similar design as Amplatzer 16, 18, 20, 22, 25, 28, 31, Cardiac Plug, but wider lobe 34

[0306] Amuleta-bAbbott

[0307] and disc and more

[0308] stabilizing wires.

[0309] Single-lobe occluder. 22, 27, 32

[0310] Nitinol frame with

[0311] WaveCrestbCoherex Medical polyurethane foam, ePTFE

[0312] membrane and retractable

[0313] anchors.

[0314]

[0315] 36

[0316] 45802538.1 Single-lobe occluder. 15, 18, 21, 24, 27, 30, 33, Occlutech LAA Nitinol wire mesh with 36, 39

[0317] Occlutech

[0318] Occluderbstabilizing loops and

[0319] nanomaterial covering.

[0320] Lobe and disc nitinol frame 16 to 36

[0321] LambrebLifetech with PET membrane.

[0322] Distal barb anchors.

[0323] aFDA approved.

[0324] bCE mark. Refers to any produce that may be freely traded in any part of Europe.

[0325] FDA: United States Food and Drug Administration.

[0326] CE: Conformite Europeenne (French for “European Conformity”).

[0327] PET: polyethylene terephthalate.

[0328] ePTFE: polytetrafluoroethylene.

[0329] LAA: left atrial appendage.

[0330]

[0331] Results and Discussion

[0332] The first step in developing the left atrial cardiac simulator was designing and fabricating a rigid, patient-derived left atrial model that could accommodate attachment of soft, patient-specific LAAs of varying shapes and sizes (Figures 3A-2H, Figure 5C). Using an open-source database of 100 MRI scans from patients with AF, a single LA for modification was selected and segmented. The segmented model was post-processed to include additional features for LAA attachment, septal crossing, and integration within a circulatory flow loop (Figure 5C). The LAA attachment was designed with a round geometry and a universal diameter to accommodate a wide range of LAA sizes and shapes. Its symmetrical design supports various LAA geometries with differing short and long-axis ostial dimensions and allows for rotation about the central axis, offering flexibility to simulate varying levels of difficulty in device deployment. The septal crossing window was positioned based on clinician input, so that it represented a realistic location for transeptal puncture. This window can be sealed with circular silicone sheets of varying thickness or durometer, allowing for customizable simulation of transeptal crossing during procedural training. The final model was then printed using a 500 inkjet-based multi-material 3D printer (Stratasys) from a polypropylene-like photopolymcr resin (RGD450), thereby creating the foundation for the left atrial cardiac simulator. The LAA attachment feature was designed to accommodate attachment of the complete spectrum of LAA sizes and shapes. The septal crossing feature was designed to allow procedural simulation and training and visualization of the ostium and LAA during and after device deployment.

[0333] 37

[0334] 45802538.1 Next, a library of soft patient-derived left atrial appendage models that could be attached onto the rigid left atrial model were developed (Fig. 3E-3H, Figure 31 and 3J). Using the same open-source database of 100 MRI scans from patients with AF, a diverse range of LA anatomies with diverse LAA geometries was selected. Following image segmentation and post-processing, the LAAs were isolated by making plane cuts through the LA proximal to the ostium to isolate the LAA, allowing preservation of the landing zone important for device deployment (Figure 31). Custom molds were created for each LAA geometry to produce soft silicone castings (Figure 3 J). The “Cauliflower” geometry (LAA1) falls with the appropriate size range and elasticity for the LAA in the healthy population. These models captured the size and elasticity ranges representative of LAAs in both healthy adults and patients for whom LAAO would have been appropriate (data not shown). This workflow was repeated to yield a comprehensive library of soft appendages, reflecting the spectrum of LAA geometries in both healthy patients and patients for whom LAAO would have been appropriate (pathological LAA anatomies for coupling with the LA model).

[0335] Following the design and fabrication of the LA model and library of LAA models that captured intra-patient heterogeneity, these components were incorporated into a mock circulatory flow loop (Fig. 2). A patient-derived model was incorporated into circulatory flow loop to achieve left atrial cardiac simulator. A cauliflower LAA model was coupled to rigid patient-derived LA model. The simulator has four inflows (RI PV, RS PV, LS PV, LI PV) and one outflow (MV). The resulting simulator contain of pulse duplicator pump, patient-derived LA model, mitral valve, tunable resistance and compliance elements, pressure sensors, and flow probes. The flow loop comprises a pulsatile pump, the custom LA model, a mechanical mitral valve, and tunable resistance and WindkesseLbased compliance elements (Figure 5H). The LA model includes four inflows, representing the pulmonary veins, and a single outflow, representing the mitral valve (Figure 51). The system is equipped with pressure sensors and flow probes to measure total LA inflow, as well as pressure at the pulmonary veins, within the LA and LAA, and in the simulated left ventricle (LV) and aorta. Different soft LAA geometries were easily attached to the rigid LA model, with LAA1 (data not shown). The successful integration of these models into the circulatory flow loop allowed us to fully realize the LA cardiac simulator as designed (data not shown).

[0336] The circulatory flow loop is composed of a pulsatile pump, the custom LA model, a mechanical mitral valve, and tunable resistance and compliance elements. The LA model has four inflows, representing the four pulmonary veins, and a single outflow, representing the mitral valve. This system as designed was easy to attach different soft LAA geometries onto the rigid LA model.

[0337] The entire system is equipped with pressure sensors and flow probes to measure total flow into the model, pressure and flow at the four pulmonary veins, pressure within the left atrium,

[0338] 38

[0339] 45802538.1 pressure within the LAA, and pressure in the simulated left ventricle and aorta. To validate the simulator, it was demonstrated that physiologically relevant pressure values could be achieved for the LA, LV and Aorta, closely mirroring the idealized curves of the Wigger diagram using this system. The simulator’s ability to replicate both physiological and pathological LA hemodynamics was evaluated (Figures 13A and 13B). First, left atrial pressure (LAP) was validated against reference data from healthy adults

[0037] (dotted green lines) and clinical data from patients with atrial fibrillation (AF), spontaneous echo contrast (SEC), an imaging marker associated with increased thrombus formation

[0038] - and post-LAAO conditions. Benchmark pressure data are provided in Table 2 with additional pressure data in Tables 4-5. Data collection was performed using LAA1 (healthy patient model) for “Healthy” data and LAA4 (AF patient model) for “AF,” “SEC,” and “Post-LAAO” conditions. By tuning the pump heart rate, pump stroke volume, system resistance, compliance, and static reservoir height, target pressure values consistent with these conditions were achieved (Figure 13 A). The simulator also replicated physiologically relevant left ventricle (LV) and aortic pressure profiles, closely matching idealized Wiggers diagram curves (Figures 5D and 5F). Next, mean LA inflow against literature-reported data for healthy adults [39-43] and patients with AF [41-43] were evaluated. By varying pump heart rate and stroke volume, a range of inflows between 2.0 and 5.7 L / min (Figure 13B) were achieved. Benchmark flow data are provided in Table 3 with additional flow data in Table 6. To simulate AF hemodynamics, increased pump heart rates (120 BPM) were utilized (Figures 14A-14D).

[0340] Physiological flow and pressure were tested with the LA model, which was benchmarked against standard values (Table 1).

[0341] The LAA in the model is not actively contractile (i.e., it does not contract itself, but exhibits passive motion driven by pulsatile flow) and therefore lacks the characteristic a-wave caused by the “atrial kick” or atrial contraction at the end of ventricular diastole just before the mitral valve closes.

[0342] A key feature of the simulator is its tunability and flexibility, allowing for adjustments and modifications to flow loop dynamics and model geometry to achieve different target hemodynamic values. Following the successful validation of physiological and pathological LA hemodynamics, the complete functionality of the simulator was evaluated. This included exploring all available pressure and flow measurement points and assessing the impact of pump parameters on LA hemodynamics (data not shown), varying compliance through either Windkessel elements or the material properties of the LAA model, and assessing the impact on LAP (data not shown) and exploring the effect of different LAA geometries on LA hemodynamics (data not shown). LAP, LA inflow, and LA outflow by varying pump HR and pump SV were easily tunable (data not shown).

[0343] 39

[0344] 45802538.1 Compliance was decreased and a characteristic increase in amplitude of the LAP waveform was observed [34, 35] (i.e., increased LA pulse pressure) (data not shown). Measurable differences in LA hemodynamics with different LAA geometries were observed (Supplementary Fig. 8), highlighting the importance of using patient-specific models for more rigorous testing and evaluation.

[0345] After demonstrating the tunability of the system, the simulator was used to investigate hemodynamic changes following LAAO (Figures 9A-9F). LAP was measured under three conditions: complete exclusion of the LAA using an acrylic cap (Figure 9A), deployment of a standard WATCHMAN device (24 mm WATCHMAN FLX, Boston Scientific), and deployment of a modified WATCHMAN device with an impermeable silicone coating (not intended for clinical use) to simulate a “healed” device after endothelialization (Figure 9B). These were compared to baseline LAP before occlusion (Figure 9C). The findings revealed a significant increase in LA pulse pressure following LAAO across all three techniques (Figure 9D). The observed increase in pulse pressure reflects a decrease in compliance (data not shown), as excision / occlusion reduces the total LA chamber volume and eliminates the LAA as a compliance reservoir. Among the techniques, complete exclusion with the acrylic cap resulted in the highest pulse pressure increase, likely because it entirely sealed the ostium and removed the LAA as a compliance vessel. In contrast, the standard WATCHMAN device, being compliant and permitting residual flow shortly postimplantation primarily through the fabric of the device, caused a smaller increase in pulse pressure. Over time, endothelialization would restrict flow through the standard device, approximating the hemodynamics of the impermeable modified WATCHMAN device. The modified device, intended to mimic the “healed” device, resulted in higher pulse pressure compared to the standard WATCHMAN, more closely resembling the complete occlusion achieved with the acrylic cap. However, the increase was modest, suggesting residual compliance in the “healed” device / LAA, potential minor leaks around the device, allowing residual LAA compliance function, or that the LAA’s role as a capacitance chamber may be minor. These findings highlight the utility of the simulator for measuring parameters that can be used as indices of occlusion effectiveness, permitting differentiation between levels of occlusion across devices and techniques.

[0346] LA inflow and outflow was also measured following LAAO and observed findings indicative of decreased compliance post-occlusion (Figures 9E and 9F). Specifically, pulmonary venous flow into the LA (LA inflow) and flow immediately distal to the LA (LA outflow) were measured (Figure 9E). LA inflow was assumed to be the total sum of flow from all four pulmonary veins. Data were collected both before and after device deployment at two pump settings (70 mL and 80 mL SV at 70 BPM). To assess compliance, the difference between inflow and outflow were 40

[0347] 45802538.1 compared, expressed as volume per beat (Figure 9F). It was postulated that a more compliant LA / LAA would result in larger differences between inflow and outflow, as the chamber could accommodate and store more volume. Conversely, a less compliant LA / LAA would yield smaller differences, reflecting a more rigid struct lire unable to expand with increased flow. A decrease in volume per beat was observed after device placement, consistent with reduced compliance. These findings further validate the utility of our simulator for measuring quantifiable markers of occlusion. This capability allows differentiation between occlusion states or pathological conditions that affect LA compliance.

[0348] After demonstrating that the simulator could measure changes in occlusion states, it was used to evaluate specific device performance (Figures 10A-10D; Figures 11A-11D). To assess the impact of device material and the degree of “occlusiveness” or PDL, important markers of procedural success, two different devices were tested and dye clearance was monitored within the LA A distal to the device (Figure 10A). Dye clearance time, measured as the duration for peak dye intensity in the LAA to return to baseline, served as a quantitative marker for occlusion efficacy. Two identically sized devices were deployed, selected per manufacturer guidelines based on LAA dimensions. The first was a standard WATCHMAN FLX with a permeable PET fabric cover, and the second was a modified WATCHMAN FLX with an impermeable silicone coating applied to the PET cover. Following deployment, green dye was injected into the LAA distal to the device, and dye dilution was visually monitored (Fig. 10B-10C). An endoscopic camera captured images of the LA portions of the standard (Fig. 10A, right) and modified devices (Fig. 10C, right). Color intensity in the LAA was measured and normalized from pre-injection baseline intensity to maximum intensity following dye injection (Figure 10C). For the standard device, peak dye intensity decreased by 40% within 22.9 s and returned to baseline within 40 s, indicating flow through the permeable PET cover (Figure 10D). This reflects the device’s behavior immediately postdeployment, before endothelialization. In contrast, for the coated device, peak dye intensity decreased by 40% over 70.4 s and remained above baseline through 240 s, indicating a higher degree of occlusion (Figure 10D). The impermeable coating prevented flow through the device cover, simulating long-term device behavior after endothelialization. However, the steady dye intensity decline suggests incomplete occlusion, likely due to PDL around the coated device. These results demonstrate the simulator’s capability to evaluate device-specific performance, including occlusiveness and PDL, under both acute and chronic conditions.

[0349] After deploying the devices into the simulator, clinical imaging techniques were used to visualize the devices and evaluate their performance (Fig. HA, 11C). Using intracardiac echocardiography (ICE; Fig. 11 A.i), flow through the permeable PET fabric cover toward the ICE 41

[0350] 45802538.1 probe following dye injection distal to the device was observed (Fig. 11 B.ii-iii). Standard echocardiography was also used to assess device placement and measure PDL(Fig. HC.i). Flow measurements were conducted at the LAA ostium wall under three conditions: (1) no device present, measuring baseline flow into and out of the LAA (Fig. 1 IC.ii), (2) device deployed, measuring a reduction in flow at the same region indicative of successful occlusion (Fig. HC.iii), and (3) deliberate device misplacement, intentionally canting the device into the LA, resulting in a return of flow at the same region indicative of PDL due to poor positioning (Fig. 1 IC.iv). An endoscopic camera provided additional visualization of the LA portion of the device, confirming proper placement or deliberate misplacement (Fig. 1 ID). These experiments highlight the simulator’s compatibility with imaging techniques for quantitative assessment of device performance, including occlusion efficacy and potential complications such as PDL from improper device positioning.

[0351] Finally, it was demonstrated that the simulator could serve as a valuable tool for procedural planning and user training in a realistic, clinically relevant physical environment (Fig. 12a). A trained operator utilized the simulator to practice procedural steps, including septal crossing and repeated device deployment and recapture, using the Access System (Access Sheath and Dilator, Boston Scientific) and Delivery System (Delivery Catheter and Closure Device, Boston Scientific) (Fig. 12B). The model allowed the operator to practice deploying and repositioning the WATCHMAN FLX device with direct visualization of the device within the LAA (Fig. 12B). The use of clear fluid further allowed intracardiac visualization using an endoscopic camera inserted through one of the pulmonary veins. This setup allowed the operator to repeatedly deploy and recapture the device until achieving satisfactory placement, assessed by the PASS Criteria (Fig.

[0352] 12C-12F). Each criterion could be evaluated using the following approaches: Position by visual inspection externally or via endoscopy / echocardiography; Anchor by pullback testing and inspection for device movement externally and internally; Seal by dye injection or hemodynamic changes as demonstrated in Figures 9A-9F using measurable indices for occlusion and seal; and Size by external measurement of device compression or internal measurement via

[0353] endoscopy / echocardiography.

[0354] 42

[0355] 45802538.1 Table 2. Benchmark pressure data for validation of LA cardiac simulator.

[0356] Value Healthy Patients AF Patients (n=435) SEC Patients Post-LAAO (n=32) Patients (n=250)

[0357] Mean Range Mean 95% Cl Mean 95% Cl Mean 95% Cl Pressure (mmHg)

[0358] 14.5- Left Atrium (LA) Mean 7 6 to 15 14.6 14.1-15.2 17.7 15.4-20.0 15.2 15.9 Left Atrium (LA)

[0359] Maximum 13 6 to 20

[0360] Left Atrium (LA)

[0361] Minimum 3 -2 to +9

[0362] Pulmonary Artery Wedge (" PCWP")

[0363] PCWP Mean 9 6 to 15

[0364] PCWP Maximum 16 9 to 23

[0365] PCWP Minimum 6 1 to 12

[0366] AF: atrial fibrillation

[0367] SEC: spontaneous echo contrast.

[0368] LAAO: left atrial appendage occulsion

[0369]

[0370] Table 3. Benchmark flow data for validation of LA cardiac simulator.

[0371] Category Age TVI (cm) Velocity (cm / s) Heart Estimate Source Rate dLA

[0372] (bpm Inflow

[0373] ) (L / min)

[0374] Systoli Diastoli Systoli Diastoli

[0375] c c c c

[0376] Healthy 64.

[0377] Adults 7 ± Basnight et 7.3 14.8 9.7 56.4 44.3 72.9 6.8 al. (1991) 61 Chao et al. ± 8 10.4 5.9 45 35 78.6 4.9 (2000)

[0378] Paraskevaidi 55 s et al.

[0379] ± 6 10.6 7.4 48 43.7 73 5 (1994)

[0380]

[0381] 43

[0382] 45802538.1 48

[0383] + Ren et al. 15 13 7 54 43 73 5.6 (1993) SO- Gentile et al.

[0384] 59 13.9 8 53 41 77.7 6.5 (1996) 60- Gentile et al.

[0385] 69 14.8 7.5 54 37 72.3 6.2 (1996) 70- Gentile et al.

[0386] 80 13.6 6 51 35 79 5.9 (1996) Atrial

[0387] 63 Chao et al. Fibrillatio

[0388] ± 8 4.2 5.7 30 40 85 3.2 (2000) n

[0389] Paraskevaidi 58 s et al.

[0390] ± 7 4.4 6.9 35.5 48.2 76.5 3.3 (1994) 61

[0391] + Ren et al. 12 5 8 31 42 76 3.8 (1993) TVI: time velocity integral

[0392] LA: left atrium

[0393] Estimated LA Inflow Calculation:

[0394] Qpv— TVI x A x HR

[0395] A

[0396] / D\2

[0397] — Tt x 1 — 1 where D is the diameter of the PV (mean PV diameter assumed to be ~ 11 m Qtotai = 4 x QPV= Estimated LA inflow

[0398]

[0399] From Kim et al., (2005)

[0400] Table 4: Hemodynamic values of normal recumbent adults.

[0401] Value Mean Range

[0402] Cardiac Index (L / min / m2) 3.4 2.8 to 4.2

[0403] Stroke volume index (mL / m2 / beat) 47 30 to 65

[0404]

[0405] 44

[0406] 45802538.1 Pressure (mmHg)

[0407] Left Ventricle

[0408] Systolic 130 90 to 140 End-diastolic 7 4 to 12 Left Atrium

[0409] Maximum 13 6 to 20 Minimum 3 -2 to +9 Mean 7 6 to 15 Pulmonary Artery Wedge (“PC”)

[0410] Maximum 16 9 to 23 Minimum 6 1 to 12 Mean 9 6 to 15 Pulmonary Artery

[0411] Systolic 24 15 to 28 Diastolic 10 5 to 16 Mean 16 10 to 22 Right Ventricle

[0412] Systolic 24 15 to 28 End-diastolic 4 0 to 8 Right Atrium

[0413] Maximum 7 2 to 14 Minimum 2 -2 to +6 Mean 4 -1 to +8 Venae Cavae

[0414] Maximum 7 2 to 14 Minimum 5 0 to 8 Mean 6 1 to 10

[0415]

[0416] 45

[0417] 45802538.1 Resistance - Wood Units (dyn*s / cms)

[0418] Total systemic 14.4 (1150) 11.3 to 17.5 (900 to 1400)

[0419] Systemic arteriolar 10.6 (850) 7.5 to 11.3 (600 to 900)

[0420] Total pulmonary 2.5 (200) 1.9 to 3.1 (150 to 250)

[0421] Pulmonary arteriolar 0.9 (70) 0.6 to 1.5 (45 to 120)

[0422]

[0423] Table 5. Pressure values in healthy patients and patients pre- and post-LAAO. Values from Mayo Clinic.

[0424] Pressure (mmHg)

[0425] v-wave

[0426] LA Mean LA Mean Range a-wave v-wave Range Normal Adult 7.9 2-12 7.1 12.8 6-21

[0427] Pre-LAAO

[0428] Atrial Fibrillation (AF) 14.6 95% CI: 14.1-15.2 14.7 24.9 95% CI: 24.0-25.9 Spontaneous Echo Contrast

[0429] 17.7 95% CI: 15.4-20.0 16.7 30.3 95% CI: 26.3-34.4 Post-LAAO 15.2 95% CI: 14.5-15.9 15.6 25.3 95% CI: 24.0-26.5 LA: left atrium.

[0430] LAAO: left atrial appendage occlusion.

[0431]

[0432] Table 6. Velocity values in healthy patients and patients pre-LAAO. Values from Mayo Clinic.

[0433] Velocity (cm / sec)

[0434] LAA Mitral E / A Ratio Contraction Filling Emptying E-Wave Peak A-Wave Peak

[0435] 60 ± 14 52 ± 13 20 ± 11

[0436] Normal Adult 60-80 20-40 1-1.5

[0437] 64 ± 19 46 + 12 38 ± 11

[0438] Pre-LAAO

[0439] 12.0 1.3 Atrial Fibrillation 13.3

[0440] 40.0 ± 27.5 95% CI: 2.6- 95% CI: 1.2- (AF) 95% CI: 5.5-21.2

[0441] Spontaneous Echo Contrast 9.7

[0442] 95% CI: 7.9-11.5

[0443] LAA: left atrial appendage.

[0444]

[0445] 46

[0446] 45802538.1 LAAO: left atrial appendage occlusion.

[0447] E / A Ratio: Ratio of peak velocity blood flow from left ventricular relaxation in early diastole (E-wave) to peak velocity flow iastole caused by atrial contraction (A-wave).

[0448]

[0449] Cun-ent state-of-the-art cardiac simulators fall short of replicating the intricate anatomical and hemodynamic features of the left atrium and the left atrial appendage. Existing simulators are primarily focused on valvular or ventricular dynamics with applications in valve repair or replacement and ventricular functional assessment. There is a lack of dedicated simulators specifically designed for LAAO that adequately capture the unique structural complexity and hemodynamic physiology of the LA and LAA such that they are suitable for robust research and device testing. To advance the field of LAAO, there is a need for simulators that combine anatomical fidelity with physiologically accurate hemodynamics. Such simulators could significantly enhance preprocedural planning and support the development and evaluation of novel occlusion devices and techniques in a realistic and inclusive environment.

[0450] While existing static 3D printed LA models have been associated with improved outcomes, such as device selection, reduced procedure times, and lower probabilities of PDL [44-47], these models lack features that are important to LA anatomy and physiology. They cannot replicate physiologically relevant hemodynamics or allow fine-tuning of parameters such as heart rate, pulmonary venous return, and compliance to mimic healthy and pathological states. Additionally, these models are often single prints that do not offer the facile exchange of LAA geometries to capture inter-patient anatomical diversity. Chaging these models by integrating dynamic flow systems, tunable hemodynamics, and modularity to accommodate various LAA geometries could yield even greater improvements in procedural planning, device development, and clinical training. For example, the ability to test devices across a spectrum of anatomies and physiological conditions could lead to more refined device designs and improved procedural techniques. Ultimately, these changes could translate to better patient outcomes, reduced complication rates, and a more comprehensive understanding of LAAO under diverse clinical scenarios.

[0451] In this study, a left atrial cardiac simulator designed to address the limitations of current models was developed. The simulator integrates a rigid, patient-derived LA model with soft, interchangeable patient-specific LAA geometries. This modular design is incorporated in a circulatory flow loop equipped with a pulsatile pump, a mitral valve, and tunable vascular resistance and compliance. The simulator was validated by replicating left atrial pressure and left atrial inflow conditions representative of both healthy adults and patients with atrial fibrillation. By adjusting parameters such as heart rate, stroke volume, and compliance, we successfully recreated a spectrum of physiologically realistic hemodynamic states, ranging from healthy to pathological.

[0452] 47

[0453] 45802538.1 While maximum achieved mean inflow (5.7 L / min) is below the upper range reported for healthy adults at rest, LA inflow is significantly reduced in AF due to shorter diastolic filling times and the absence of atrial contraction. Thus, our model provides a reasonable approximation of LA inflow conditions in patients with AF who may require LA AO. To simulate AF, the pump heart rate was increased to reflect shorter diastolic filling times.

[0454] In addition, the simulator was used to evaluate the hemodynamic changes induced by LAAO using various occlusion methods. Quantitative metrics, such as pulse pressure changes and compliance reduction, could be measured, providing valuable data to assess the effectiveness of occlusion techniques and devices. Hemodynamic analysis revealed increased LAP following LAAO, with the greatest increase occurring after complete exclusion of the LAA. These data corroborate recent clinical findings of a small but significant increase in LAP immediately post-LAAO

[0048] . When evaluating pressure and flow waveforms, it is notable that the LA in the present model is not actively contractile; instead, it exhibits passive motion driven by pulsatile flow and inherent elasticity. As a result, the model lacks the characteristic a-wave associated with the “atrial kick,” which occurs at the end of ventricular diastole just before mitral valve closure.

[0455] Following validation and demonstration that the simulator could be used to collect measurable, quantifiable data relevant for LAAO, it was used to evaluate clinically approved occlusion devices. Device performance was assessed under different conditions, including material properties, device sizing, and positioning. Metrics such as dye clearance time and detection of PDL provided valuable insights into occlusion effectiveness. Importantly, the simulator supported various imaging modalities, enabling visualization of device placement and occlusion outcomes, as well as the quantification of performance metrics like seal efficacy and flow reduction. These findings highlight the potential of the present simulator as a valuable tool for device research and development.

[0456] Finally, the simulator proved to be a robust procedural training tool, offering clinicians a realistic environment to practice device deployment, repositioning, and recapture. Using direct and imaging-guided feedback, the simulator facilitated the assessment of key procedural criteria, such as the PASS Criteria, enhancing operator learning and confidence. These iterative procedures highlight the simulator’s utility as a training tool, providing clinicians with a realistic environment to practice device manipulation, change sizing and positioning, and troubleshoot potential challenges. Importantly, the simulator supported the implantation of multiple device sizes into a single LAA anatomy. Following implantation, percent compression, an important metric for successful device fit, can be assessed externally through direct measurement or internally via endoscopy / echocardiography. Additional changes could include the integration of sensors into the 48

[0457] 45802538.1 model’s walls to measure force and test device compression in real-time, further increasing the simulator’s utility for device evaluation and procedural optimization. Additionally, the simulator offers potential applications in educational settings, including industry-sponsored workshops and conference demonstrations, where it can be used to train clinicians and showcase new devices and procedural techniques.

[0458] References:

[0459] I. Al-Saady, N. M., (1949). J Am Med Assoc.

[0460] 3. Lloyd-Jones, Det al. (2004). Circulation.

[0461] 4. Heeringa, J., (2006). study. Eur Heart J

[0462] 5. Di Biase, L., (2012). J Am Coll Cardiol.

[0463] 6. Lee, J. M., (2014). American Journal of Cardiology 113.

[0464] 7. Viles-Gonzalez, (2012). J Am Coll Cardiol 59, 923-929.

[0465] 8. Blackshear, J. L., and Odell, J. A. (1996). Preprint

[0466] 9. Fender, E. A., (2016). Left Atrial Appendage Closure for Stroke Prevention in Atrial Fibrillation. Preprint

[0467] 10. Alli, O. O., and Holmes, D. R. (2015). Left Atrial Appendage Occlusion for Stroke Prevention. Cun Probl Cardiol.

[0468] II. Hart, R. G., Diener, H. C., Yang, S., Connolly, S. J., Wallentin, L., Reilly, PA., Ezekowitz, M. D., and Yusuf, S. (2012). The RE-LY trial. Stroke.

[0469] 12. De Backer, O., (2014). Percutaneous left atrial appendage occlu-sion for stroke prevention in atrial fibrillation: An update. Preprint at BMJ Publishing Group

[0470] 13. Singh, I. M., and Holmes, D. R. (2010). Loll atrial appendage clo-sure. Preprint

[0471] 14. Wunderlich, N. C., Beigel, R., Swaans, M. J., Ho, S. Y, and Siegel, R. J. (2015). Percutaneous Interventions for Left Atrial Append-age Exclusion. JACC Cardiovasc Imaging.

[0472] 15. Glikson, M., Wolff, R., Hindricks, G., Mandrola, J., Camm, A. J., Lip, G. Y. H. H., Fauchier, L., Betts, T. R., Lcwaller, T, Saw, J., et al. (2020). EHRA / EAPCI expert consensus statement on cathe-ter-based left atrial appendage occlusion - an update. EuroIntervention 15

[0473] 16. Collado, F. M. S (2021). Left Atrial Appendage Occlusion for Stroke Preven-tion in Nonvalvular Atrial Fibrillation. J Am Heart Assoc 10.

[0474] 17. Samaras, A., (2024). Eur Heart J 45

[0475] 18. Alkhouli, M., (2023). Peridevice I.oak After Left Atrial Appendage Occlusion. JACC Cardiovasc Interv 16, 627-642

[0476] 19. Fauchier, L., et al. (2018). Device-Related Thrombosis After Percutaneous Left Atrial Appendage Occlusion for Atrial Fibrillation. J Am Coll Car-diol 71, 1528-1536.

[0477] 49

[0478] 45802538.1 20. Scdaghat, A., ct al. (2021). Dcvicc-Rclatcd Thrombus After Left Atrial Appendage Closure: Data on Thrombus Characteristic Treatment Strategies, and Clinical Outcomes From the EUROC-DRT- Registry. Circ Cardiovasc Interv 14.

[0479] 21. Simard, T, et al. (2021). Predictors of Device-Related Throm-bus Following Percutaneous Left Atrial Appendage Occlusion. J Am Coll Cardiol 78, 297-313.

[0480] 22. Dukkipati, S. R., (2018). Device-Related Thrombus After Left Atrial Append-age Closure. Circulation 138, 874-885.

[0481] 23. Alkhouli, M., (2018). Incidence and Clinical Impact of Device-Related Thrombus Following Percutaneous... JACC Clin Electrophysiol 4, 1629-1637.

[0482] 24. Bosi, G. M., (2018). Front Cardiovasc Med 5.

[0483] 25. Mill, J., (2022). Patient-specific flow simu-lation analysis to predict device-related thrombosis in left atrial appendage occluders. REC: interventional cardiology (English Edition).

[0484] 26. Danielli, F., Be (2024). Int J Numer Method Biomed Eng 40.

[0485] 27. mentice (2023). Left Atrial Appendage Occlusion, mentice.

[0486] 28. Fanni, B. M., (2024) Rapid Prototyp J

[0487] 29. Vogl, B. J., et al. (2024). JACC: Advances 3, 101339.

[0488] 30. Roney, C. H., et al. (2022). Circ Arrhythm Electrophysiol 15.

[0489] 31. van de Vegte, Y. J (2021). Sei Rep 11, 8431.

[0490] 32. Jeong, W. K., (2016). Heart Rhythm 13, 820-827.

[0491] 33. Veinot, (1997). Anatomy of the Normal Left Atrial Appendage. Circulation 96, 3112-3115.

[0492] 34. Park, J., (2015). PLoS One 10, e0143853.

[0493] 35. Meskin, M„ (2024). Sci Rep 14, 1864.

[0494] 36. Roney, C. H., (2019). Med Image Anal 55, 65-75.

[0495] 37. Fleitman, J. (2024). Pulmonary artery catheterization: Interpretation of hemodynamic values and waveforms in adults. UpToDate.

[0496] 38. BLACK, I. W. (2000). Echocardiography 17, 373-382.

[0497] 39. Basnight, M. A., et al., Journal of the American Society of Echocardiography 4, 547-558. 40. Gentile, F„ et al. (1997). Eur Heart J 78, 148-164.

[0498] 41. Paraskevaidis, (1994). Am J Cardiol 73, 392-396.

[0499] 42. REN, W. D., VISENTIN, P„ NICOLOSI, G. L., CANTERIN, F. A., DALL’AGLIO, V., LESTUZZI, C„ MIMO, R„ PAVAN, D„ SPARACINO, L„ CERVESATO, E„ et al. (1993). Effect of atrial fibrillation on pulmonary venous flow patterns: transoesophageal pulsed Doppler echocardiographic study. Eur Heart J 14, 1320- 1327.

[0500] 43. Chao, T.-H.. (2000). Chest 117, 1546-1550.

[0501] 50

[0502] 45802538.1 44. Obasarc, E (2017). Int J Cardiovasc Imaging.

[0503] 45. Otton, J. M., (2015). JACC Cardiovasc Interv 8, 1004-1006.

[0504] 46. Liu, P., (2016). Cardiology 135.

[0505] 47. Wang, D. D., et al. (2016). JACC Cardiovasc Interv 9, 2329-2340

[0506] 48. Alarouri, H. S., et al. (2024). Heart Rhythm 21, 1024— 1031.

[0507] 49. Meskin, M„ (2024). Sci Rep 14, 1864.

[0508] 50. Park, C„ Singh, M„ Saeed, M. Y., Nguyen, C. T, and Roche, E. T (2024). Device 2, 100217. 51. Singh, M., et al. (2023). Nature Cardiovascular Research 2, 1310-1326.

[0509] 52. Alkhouli, M„ (2023). Am Coll Cardiol 87, 1063-1075.

[0510] 53. Raphael, C., (2017). EuroIntervention 13, 1218-1225.

[0511] 54. Piayda, K., et al. (2021). EuroIntervention 77, el033-el040.

[0512] Example 2: Integrating soft robotics and computational models to study left atrial hemodynamics and device testing in sinus rhythm and atrial fibrillation

[0513] Materials and Methods

[0514] Study Overview

[0515] A multimodal framework combining physical and computational models to investigate atrial fibrillation (AF) hemodynamics and device interactions was implemented. The workflow began with a soft robotic benchtop model, which reproduced left atrial and appendage motion under controlled actuation patterns. This platform allowed direct measurement of flow fields, pressures, and device deployment behavior in a physiologically relevant but experimentally accessible setting. The benchtop findings were then extended using a lumped parameter model (LPM) of the atrium and circulation. The LPM provided a system-level perspective, allowing examination of how alterations in atrial function and pacing rate influenced global hemodynamic indices such as cardiac output, atrial pressures, and trans-mitral flow. These quantities are difficult to access experimentally in the benchtop model but are important for interpreting the broader circulatory consequences of AF. Finally, finite element analysis (FEA) was employed to resolve local mechanics that neither the benchtop model nor the LPM can capture. The FEA quantified atrial wall stresses, strains, and deformation patterns under varying loading conditions, thereby offering tissue-scale insight into the mechanical environment during arrhythmia and device deployment. By integrating these three levels of analysis, experimental (benchtop), systemic (LPM), and structural (FEA), the study design allowed for each method to address complementary aspects of atrial biomechanics. This sequential strategy allowed us to anchor simulations in experimentally observed behavior while extending the findings to hemodynamic and structural domains not accessible to direct measurement.

[0516] 51

[0517] 45802538.1 Fabrication of patient-specific models and integration with McKibben soft robotic actuators

[0518] Retrospective cardiac-gated computed tomography angiography (CTA) images were obtained from one healthy individual and one patient with atrial fibrillation (AF). These images were provided by Hospital de la Santa Creu i Sant Pau (Barcelona, Spain) following approval by the institutional Ethics Committee and after obtaining informed consent from the patients. Imaging was performed using a Somatom Force scanner (Siemens Healthineers, Erlangen, Germany) with a biphasic contrast injection protocol. For the control case, full cardiac phase reconstructions were performed at every 1% of the cardiac cycle, except for the 34%- 48% interval, which was not acquired. In the AF patient, images were acquired at every 5% of the cardiac cycle, covering the entire R-R interval (0% to 99%). The left atria (LA) from both cases were segmented from the CTA images using semi-automatic tools available in 3D Slicer v4.10.11, representing the 0% of the R-R interval, which corresponds to the onset of ventricular systole. From these segmentations, 3D surface meshes were generated using the flying edges algorithm in 3D Slicer. Post-processing steps were applied to refine the LA surface meshes. MeshLab v2021-07 was used to correct intersecting faces and non-manifold edges. Additionally, Autodesk Meshmixer v3.3.15 was employed to define planar surfaces at the ends of the pulmonary veins and mitral valve to facilitate anatomical consistency.

[0519] The LA models from healthy subjects were fabricated using SIL30, a soft, flexible silicone material provided by Carbon, Inc., and integrated with soft robotic pneumatic actuators. Similarly, the LA and LV models from AF patients were also printed in SIL30.

[0520] Soft robotic actuators were customized for specific cardiac regions. For the LA from a healthy patient, three actuators were designed for the anterior and posterior-inferior walls, and a single actuator was created for the left atrial appendage wall. For the LA model derived from an AF patient, only two actuators were implemented in the anterior and posterior-inferior regions to replicate the diminished atrial contractility characteristic of AF, omitting any actuator for the left atrial appendage. Dimensions were carefully measured so that the precise fabrication of these elements, and the actuators were fabricated with dimensions tailored for their specific regions. The LV actuators included three circumferential actuators placed at the basal, middle, and apical sections and three helical actuators positioned at an approximate 60-degree angle from the basal plane21. The actuator framework was designed to capture the key modes of myocardial deformation, with helical actuators providing ventricular twisting, circumferential actuators driving inward radial contraction, and longitudinal actuators contributing to atrioventricular (AV) plane displacement. All prototypes, including the one shown in Figure 23, contained these longitudinal actuators, though 52

[0521] 45802538.1 they are difficult to discern in the figure and video due to their darker color and shorter length. The model in Figure 24 represents a later variant in which the number of circumferential actuators were increased and the longitudinal actuators were extended to accommodate a larger ventricular geometry. These refinements were guided by prior computational modeling framework21, which demonstrated that higher circumferential actuator density and longer longitudinal actuators improved AV plane motion and hemodynamic performance. While the helical and circumferential actuators allow twisting and inward shortening, the present setup did not fully reproduce physiological AV plane displacement. This underscores the importance of actuator placement and scaling, and addressing AV plane motion will be a focus of future iterations.

[0522] Flattened McKibben actuators were employed for the circumferential arrangement, while cylindrical McKibben actuators were used for the helical inner layer. Each soft robotic actuator contained three core components: a thermoplastic elastomer (TPE) bladder, thermoplastic polyurethane tubing, and a PET expandable braided mesh. The TPE bladder (Stretchion 200, sourced from FibreGlast Developments Corp.) was formed by heat-sealing two TPE layers on a 3D-printed mold at 300°F for 4 seconds. A 1 / 8-inch thermoplastic polyurethane tube (from McMaster-Carr) was inserted into the bladder and sealed using Ure-Bond II adhesive (Smooth-On, Inc.). The braided mesh (1 / 4-inch PET mesh, TechFlex, Inc.) was coated with Ecoflex 00-30 silicone to prevent kinking, and the actuator assembly was hand-sewn with Kevlar thread for structural integrity. The actuators underwent rigorous quality assurance by cycling them approximately 500 times at a pressure of 20 psi, using a custom-built electro-pneumatic control system51’54. This system incorporated electropneumatic pressure regulators and valves (SMC Pneumatics), which allowed precise control of pressure waveforms, heart rate, and systolic / diastolic ratios through analog inputs. To integrate the actuators with 3D-printed cardiac anatomies, they were first encapsulated within a passive silicone matrix layer approximately 3-5 mm thick. After curing, this structure, referred to as the soft robotic myocardium, was adhered to the LV model using Sil-Poxy adhesive (Smooth-On). In the case of the EA, actuators were directly attached to the printed components using Sil-Poxy, and their ends were secured with Kevlar thread for a conformal fit.

[0523] Design and configuration of the mock circulatory flow loop and hemodynamics acquisition

[0524] The 3D-printed cardiac models with integrated soft robotic myocardium were connected to a mock circulatory flow loop, simulating left-sided cardiovascular circulation. In this setup, the soft robotic left atrium and left atrial appendage acted as a reservoir, conduit, and pump during atrial diastole and systole, while the soft robotic left ventricle drove systemic blood flow. Directionality of flow was controlled using mechanical valves (St. Jude Medical) positioned as mitral and aortic 53

[0525] 45802538.1 valves. The mock circulatory loop incorporated hydraulic and mechanical elements, including inhouse acrylic compliance chambers to represent systemic and pulmonary compliance. On-off ball valves (McMaster-Carr, 4796K71) were employed to simulate systemic vascular resistance, while a 1:4 straight-flow rectangular manifold (McMaster, 1023N244) divided flow from the reservoir to the four pulmonary veins. During sinus rhythm simulations, the models were actuated at 60 bpm using a custom-built electro-pneumatic control box. The system synchronized atrial contraction (atrial systole) and ventricular systole with programmed time delays. Adjustments to actuation, resistance, and compliance within the loop allowed the simulation of desired left-sided hemodynamics. A blood mimic fluid containing 40% propylene glycol by volume in deionized water, with a dynamic viscosity of 4.3 ± 0.8 mPa s, was circulated through the system.

[0526] Hemodynamic parameters were measured using pressure sensors (PRESS-S-000, PendoTECH) positioned at key locations, pulmonary veins, LA, LAA, LV, and aorta, via 5 F umbilical vessel catheters (CardinalHealth). The sensors recorded biphasic pressure waveforms, while flow measurements at the pulmonary veins, mitral valve conduit, and aorta were taken using an ultrasonic flow probe (ME 13 PXN, Transonic) connected to a T420 multichannel research console (Transonic Systems Inc.). To simulate hemodynamic conditions associated with atrial dysfunction and pathophysiology, the control system was utilized to adjust the actuation input pressure of the soft robotic myocardium, replicating reduced contractility. The code for the control system has been previously described53. Alternatively, the control system was programmed to vary the beating rate (40-150 bpm), mimicking atrial arrhythmias. To visualize valve and wall motion, a 1080P HD endoscopic camera (NIDAGE) recorded videos at 30 frames per second. This comprehensive setup provided detailed insights into the mechanical and hemodynamic performance of the models within the simulated cardiovascular environment. The following equation was utilized for evaluating left atrial appendage emptying.

[0527] LAA ejection fraction %

[0528]

[0529] — lM21mjnjmuin are£I) / L2424maxjmum area) X 100

[0530] Echocardiographic imaging for structural and functional analysis The Philips Epiq CVx cardiovascular ultrasound system, equipped with XL14-3 and X5-1 transducers, was utilized for imaging and analysis. Echocardiographic assessments of wall mechanics and valvular motion were conducted using 2-dimensional B-mode imaging. Flow velocity within the left atrium was measured through pulsed wave (PW) Doppler, while the LAA region was evaluated with color Doppler echocardiography to map flow patterns. For precise visualization of flow dynamics, wall mechanics, and valve motion, the transducer was placed directly on the soft robotic heart for epicardial imaging during benchtop experiments. Wall displacement linked to each actuator's activity in the left atrium and appendage walls was quantified 54

[0531] 45802538.1 by capturing repetitive wall motion in the region of interest with M-mode echocardiography. Data analysis and visualization were conducted using Q-Vue 2.2 software (Philips), facilitating detailed interpretation of actuator-induced wall mechanics.

[0532] Phase-contrast 2D flow MRI for intra-atrial velocity measurements

[0533] To facilitate compatibility with the MR environment, all ferromagnetic components in the mock circulatory loop were substituted with non-ferromagnetic plastic parts. The pulmonary vein flow splitter was fabricated using VeroBlue material (Stratasys Objet) via 3D printing. The control system responsible for the cyclic actuation of the robotic heart was positioned outside the magnetic resonance imaging (MRI) room to avoid magnetic interference. Imaging was performed using an MRI scanner (3T MAGNETOM Prisma-Fit, Siemens Healthineers). The 2D flow MRI was acquired with a voxel resolution of 1 x 1 x 1 mm, synchronized with a simulated electrocardiogram (ECG) signal mimicking the robotic heart's beating pattern (60 bpm). Velocity encoding was performed in three directions, focusing on the through-plane component. Imaging parameters for the 2D flow sequence included a repetition time of 126.4 ms, echo time of 5.37 ms, flip angle of 7°, velocity encoding (VENC) range from 40 cm / s to 60 cm / s, temporal resolution of 28 frames per cycle, and a 2D matrix size of 256 x 256. The MRI phase-contrast 2D flow data inside the LA and LAA was analyzed using GTFlow 4.9.21 software (GyroTools). Potential sources of phase offsets, such as eddy currents, were addressed during data processing. This process involved noise masking, velocity anti-aliasing, and eddy current correction for accuracy and reliability of the velocity maps.

[0534] Device testing for left atrial appendage occlusion and performance assessment Commercially available occlusion devices, including the WATCHMAN FLX, sizes 24 mm and 27 mm (Boston Scientific), were employed to evaluate hemodynamic changes pre- and post-LAAO and determine the appropriate device size for the patient-specific anatomy derived from AF cases. Device placement within the benchtop simulator was guided in real-time by the clinical user using either echocardiography (Philips Epiq CVx cardiovascular ultrasound system) or a 1080P HD endoscopic camera recording at 30 frames per second. Post-occlusion efficacy was assessed through color Doppler echocardiography, which mapped the flow to verify proper device sizing and positioning. Structural MRI (3T MAGNETOM Prisma-Fit, Siemens Healthineers) was used to confirm and validate the device's positioning at the LAA ostium. Although MRI is not a standard clinical modality for assessing LAA occlusion efficacy, typically evaluated using post-operative cardiac-gated CT or echocardiography, it was employed in this experimental research setting due to the lack of cardiac-gated CT equipment. MRI allowed for the observation of potential leakage through the device under pulsatile flow conditions and provided qualitative measurements in a controlled environment. To simulate endothelialization and prevent leakage through the

[0535] 55

[0536] 45802538.1 polyethylene terephthalate (PET) mesh of the occlusion device, the WATCHMAN FLX™ devices were coated with a thin layer of silicone (Ecoflex 35-00 fast, Smooth-On).

[0537] Lumped parameter modeling for systemic hemodynamic analysis

[0538] A lumped parameter model was developed to simulate cardiovascular hemodynamics pre-and post-LAAO, leveraging Simulink / Matlab (MathWorks, Natick, MA, USA) for system modeling and numerical simulations. The cardiovascular system was represented as a network of resistive, capacitive, and inertial elements, capturing the hydraulic and mechanical properties of the systemic and pulmonary circulations, modified from previously published work55,59. The two key modifications included: (i) incorporation of a dynamic left atrial elastance component to capture reservoir, conduit, and pump functions, and (ii) parameter adjustments to simulate post-LAAO conditions by altering atrial compliance and reservoir stiffness. Parameters for atrial and ventricular elastance, valvular properties, and vascular compliance were calibrated based on physiological data to ensure accuracy using the parameter estimation tool Simulink / Matlab. The LA was modeled using dynamic elastance, incorporating its three primary functions: reservoir, conduit, and pump. A time-varying elastance model was used to represent active contraction during sinus rhythm, expressed as: LA_Ees x (LA_Vt - LA_V0) for LA pressure. The atrial reservoir function was captured using the following equation: LA_A_res x (exp(LA_B_res x (LA_Vt - LA_V0)) - 1). Parameter values used for LA are LA_Ees = 0.45 mmHg / mL (end-systolic elastance), LA_V0 = 0 mL (volume at zero pressure), LA_A_res = 0.5, LA_B_res = 0.049 (dimensionless stiffness coefficients). To simulate post-LAAO hemodynamics, several parameter adjustments were made to reflect the physiological changes. The stiffness for the reservoir function of the left atrium was increased by raising LA_B_res to 0.07. Reduced reservoir capacity was modeled by reducing LA_A_res to 0.35. A slight reduction in atrial compliance was modeled by increasing LA_Ees to 0.7 mmHg / mL. Each simulation was run for 30 s and an ode15s solver was used. Parameters for systemic vascular resistance (SVR), pulmonary vascular resistance (PVR), total arterial compliance, ventricular elastances, and atrial elastances were initialized from literature values and iteratively tuned against porcine in vivo recordings of LA pressure and trans-mitral / pulmonary venous flows. A full list of parameter values used in the final model is provided in Table 8.

[0539] Table 8: Lumped parameter model (LPM) parameters for baseline case.

[0540] Parameter Symbol Unit Baseline

[0541] (pre-LAAO)

[0542] Sampling time Tsmp s 0.001

[0543] Sampling frequency Fs Hz 1000.0

[0544] Blood density RO g / mm30.000787

[0545]

[0546] 56

[0547] 45802538.1 End-systolic elastance Ees_LV mmHg / mL 3.0

[0548] Minimum elastance LA_Emin mmHg / mL 0.125

[0549] End-systolic elastance LA_Ees mmHg / mL 0.45

[0550] Reservoir function A LA_A_res - 0.5

[0551] Nonlinear PV alpha LA_alpha - 2.9417e-05

[0552] Nonlinear PV beta LA_beta - 3.0

[0553] Systemic arterial compliance Ca_s mL / mmHg 0.155

[0554] Pulmonary arterial compliance Ca_p mL / mmHg 0.7

[0555] Systemic arterial resistance Ra_s mmHg-s / mL 1.066

[0556] Pulmonary arterial resistance Ra_p mmHg-s / mL 0.11

[0557] Pulmonary venous resistance Rv_p mmHg-s / mL 0.075

[0558] Systemic venous compliance Cv_s mL / mmHg 50.0

[0559] Pulmonary venous compliance Cv_p mL / mmHg 10.0

[0560] Systemic venous unstressed volume Vso_s mL 2200.0

[0561] Pulmonary venous unstressed volume Vso_p mL 200.0

[0562] Mean circulatory filling pressure MCFP mmHg 10.0

[0563]

[0564] Finite element modeling for left atrial biomechanics and stress analysis

[0565] Finite element analysis was employed to study left atrial biomechanics and the impact of atrial fibrillation versus sinus rhythm on pressure-volume relationships. The dynamic cardiac FEA model was developed using the SIMULIA Living Heart Human Model (LHHM) 2023 and Abaqus 2023, Simulia, Dassault Systemes60,61. The LHHM provides a dynamic, four-chamber human heart with coupled electrophysiology, mechanics and a lumped / 3D hybrid circulatory model. Electrical analyses (monodomain) were run in Abaqus / Standard; mechanical analyses were run in Abaqus / Explicit for three cardiac cycles to reach steady state. The baseline LHHM mesh and analysis sequence (PRE-LOAD, BEAT / RECOVERY with SA-node pacing) were used and are summarized in the vendor documentation. Core model components, including electrophysiology (monodomain formulation), active contraction (time-varying elastance law), and electromechanical 57

[0566] 45802538.1 coupling are implemented and validated by the vendor. 3D atrial geometry was generated from the SIMULIA Living Heart Human Model geometry (NURBS at -70% ventricular diastole) licensed from Zygote Media Group and meshed in Abaqus; the LA part is included and discretized for both electrical and mechanical analyses. Local fiber / sheet / normal directions were assigned from LHHM Discrete Fields (R_Atrium, L_Atrium, Ventricles). Atrial orientations follow the euHeart atlas used in LHHM; ventricular helix angles vary from — 60° epicardium to +60° endocardium. Orientations were verified via LHHM’s data-check display and, when remeshing, regenerated element- wise orientations per the documented workflow. For electrophysiology modeling, LHHM uses a monodomain formulation, which solves for transmembrane potential V with a diffusion tensor D aligned to fibers and a local recovery variable r governing restitution. SA-node pacing was applied as a smoothed pulse (-80 to +20 mV); electrical field output drives active stress in mechanics. LHHM’s electrical step length control and confirmed temporal / mesh adequacy for the atria. Monodomain allowed consistent voltage-driven electromechanics at our mesh density;

[0567] LHHM documentation provides electrical and mechanical convergence data supporting this choice. Atrial heterogeneity was represented by spatially varying (1) the principal values of D (fiber / sheet / normal), and (2) restitution parameters c, γ, μ1, μ2. a. Atrial anisotropy was aligned with the local fiber field. Region sets (LA roof, posterior wall, septal wall, appendage, PV antra) received distinct parameter scalings to reproduce observed slower, more heterogeneous conduction in AF substrates. Active tissue response is modeled through a time-varying elastance formulation, which captures the Frank-Starling effect. The model adds an active stress component to the passive fiber and sheet stresses. Active stress is a function of calcium dynamics and sarcomere length, with constitutive parameters including: Tmax (scaling factor governing peak contractility and ejection fraction), CaO / Ca_max (intracellular calcium concentrations), B (shape of tension-sarcomere length relation), sarcomere thresholds for activation, timing constants for rise and relaxation. Stress was applied primarily in the fiber direction, with a fraction coupled into the sheet direction.

[0568] Parameters are chamber-specific and tuned to match physiological ejection fractions and twisting behavior. Electromechanical coupling is achieved by running two linked analyses: (1) Electrical simulation (ELEC model) computes nodal potentials across the myocardium. (2) Mechanical simulation (MECH model) then applies those potentials as time-dependent activation signals to generate active fiber stresses. In practice, during each BEAT step of the mechanics analysis, the fiber stress is driven by the electrical activation wavefront; during RECOVERY steps, the potentials are reset to resting values so the chambers exhibit passive filling behavior. Nonlinear explicit dynamic analyses were performed to simulate the time-dependent mechanics of the LA under different (patho)physiological conditions. To represent blood flow and simulate its interaction with 58

[0569] 45802538.1 LA wall mechanics, each cardiac compartment and the systemic circulation were modeled as hydrostatic fluid cavities. Surface-based fluid cavities were incorporated to simulate LA inflow and outflow interactions, enabling dynamic coupling of atrial contraction and relaxation with systemic circulation. The passive mechanical behavior of the LA wall was modeled using the anisotropic hyperelastic material formulation proposed by Holzapfel and Ogden for cardiac tissue60’62. Active LA tissue mechanics were described using a time-varying elastance model that simulated atrial contraction during sinus rhythm (60 bpm) and the absence of coordinated contraction during AF (120-160 bpm). The mechanical model was meshed using 639176 elements and 199317 nodes and the element types included C3D4, CONN3D2, DCOUP3D, MASS, S3R, S4, SFM3D3, SFM3D4R, and T3D2, for appropriate representation of the complex anatomical and functional structures in the model. Although linear tetrahedral (C3D4) elements can be overly stiff, their use in the Living Heart Human Model is mitigated by mesh density and validated through convergence studies, which demonstrated stable physiologic outputs across coarse, medium, and fine discretization. Boundary conditions followed the default LHHM implementation. The pericardial constraint was modeled as a distributed spring-dashpot applied normal to the epicardial surface, providing physiologic tethering while permitting limited motion. At the atrioventricular plane, basal nodes were coupled to a reference node through axial connectors with tuned stiffness, allowing physiological atrioventricular plane displacement during ventricular systole and atrial filling. The LA cavity pressure and volume changes were simulated over three cardiac cycles of 1 second each to achieve steady-state conditions. For sinus rhythm, LA contraction was dynamically modeled, capturing reservoir, conduit, and pump functions. The electrical model was meshed with 114617 nodes, 424706 elements with DC1D2 and DC3D4 elements type. The baseline model was modified to incorporate electrophysiological and structural remodeling characteristics of AF. The baseline model was modified to incorporate both electrophysiological and structural remodeling characteristics of atrial fibrillation. Irregular activation patterns were generated by applying a sequence of shortened and variable cycle lengths (0.375 s, 0.40 s, and 0.50 s), each having a beat and recovery phase, to mimic the cycle -to-cycle variability typical of AF. Electrical remodeling was introduced through anisotropic reductions in atrial conductivity, shortened refractoriness, and perturbed activation thresholds, which together produced slowed conduction, chaotic re-entry, and rapid recovery consistent with AF physiology. These modified electrical signals were then mapped through electromechanical coupling to drive left atrial contraction. To simulate atrial fibrillation, the electrical properties of the left atrium were modified from those used to represent normal sinus rhythm. For example, anisotropic conductivity values were reduced to reflect AF-associated slowed conduction and heterogeneity: 20.0, 4.5, 4.5 mm2 / ms for 120 bpm, 16.5, 4.2, 4.2 mm2 / ms for 150

[0570] 59

[0571] 45802538.1 bpm, and 15.0, 4.0, 4.0 mm2 / ms for 160 bpm. Electrical material properties, including refractoriness (y), scaling (c), restitution time constants (μ1, μ2), and oscillation threshold (a), were adjusted to mimic AF-related rapid recovery, flattened restitution curves, and chaotic dynamics, for an accurate representation of cycle durations and irregular atrial activity across heart rates. The sinoatrial (SA) node amplitude varied irregularly between -80 mV and 20 mV, with cycle durations of 500 ms (120 bpm), 400 ms (150 bpm), and 375 ms (160 bpm). Atrioventricular (AV) node parameters reflected disrupted conduction, including reduced delay times (e.g., tdelay = 42.5 ms for 120 bpm, 31.9 ms for 160 bpm) and shorter repolarization periods (te = 95 ms for 120 bpm, 71.25 ms for 160 bpm). Rise times (t2) ranged from 1.9 to 7.5 ms, and threshold activation potentials (EPOT Act = 10 mV ± 1 mV) included random perturbations to simulate chaotic atrial activity. The parameters have been summarized in Table 7. The active and passive material properties of the LA were also modified to reflect the stiffness, reduced relaxation, and increased passive tension characteristic of AF.

[0572] Mechanical analysis was achieved by mapping electrical outputs to drive myocardial contraction dynamics. Biomechanical Mises stress distributions in the LA wall were analyzed at end-diastole and peak systole, while pressure-volume data were extracted to compare functional differences between sinus rhythm and AF conditions. Formal sensitivity or uncertainty analyses was not performed in this study. The irregularity of atrial activation was validated by comparing simulated activation sequences and pressure-volume dynamics to published experimental and clinical AF data, confirming that the model reproduced hallmark features such as variable cycle lengths, disorganized conduction, and impaired atrial contractility.

[0573] Validation of left atrial model functionality using human imaging data Retrospective anonymized human echocardiographic imaging data (n = 7) acquired through transesophageal echocardiography (TEE) and tabular hemodynamic data (including LA volume, left atrial mean pressure values, and pressure for v-wave and a-wave; n = 3) were provided via a collaborative partnership with Cedars-Sinai, Los Angeles, USA (IRB approval number STUDY00002705). Left atrial (n=3) and LAA (n=7) wall motion and contractility were analyzed using Q-Vue 2.2 software (Philips). All these patients had a history of atrial fibrillation and were considered candidates for left atrial appendage occlusion device implantation. The analysis focused on calculating the percentage area change of the left atrium between systolic and diastolic phases to validate the contractility range achieved by the soft robotic actuators.

[0574] 60

[0575] 45802538.1 Table 7: Parameter Modifications for Atrial Fibrillation Simulation in FEA model.

[0576] Parameter Baseline (Sinus Modified (AF simulation)

[0577] Rhythm)

[0578] Cycle periods Regular 1.0 s (60 Irregular 0.375 s, 0.40 s, 0.50 s

[0579] bpm)

[0580] SA node amplitude -80 to +20 mV, -80 to +20 mV, irregular between beats regular

[0581] Atrial conduction (anisotropic 20.0 / 4.5 / 4.5 20.0 / 4.5 / 4.5 @120 bpm; 16.5 / 4.2 / 4.2 @150 bpm; conductivity, mm2 / ms) (normal atrium) 15.0 / 4.0 / 4.0 @160 bpm

[0582] AV node conduction delay (t_delay) ~60 ms (normal) 42.5 ms @120 bpm; 31.9 ms @160 bpm AV node repolarization (t_e) -120 ms 95 ms @120 bpm; 71.25 ms @160 bpm Rise time (tz) -7.5 ms 1.9-7.5 ms

[0583] Activation threshold (E_POT Act) 10 mV 10 ± 1 mV, random perturbations

[0584]

[0585] Animal handling, surgical procedures, and in vivo hemodynamics in porcine models To confirm the physiological capabilities of the model and validate its biomechanics and hemodynamics, experiments were conducted using an acute porcine model (n=3). The animal experiments adhered to ethical guidelines outlined in the National Research Council’s Guide for the Care and Use of Laboratory Animals and were conducted under MIT Institutional Animal Care and Use Committee protocol #2311000601. Three Yorkshire swine (60-80 kg, sourced from CBSET, Inc.) were used. Animals were intubated, placed on mechanical ventilation, and maintained under general anesthesia (2-3% isoflurane). Arterial and venous femoral lines were inserted for systemic blood pressure monitoring and medication delivery, respectively, and a median sternotomy was performed to access the thoracic cavity. To measure pressures and volumes, a Transonic Scisense ADV 500 Large Animal pressure-volume (PV) System console with a 5F straight-tip PV loop catheter (Model: SCISENSE ADVANTAGE Large Animal PV Foundation System, V.5.0) was inserted through a catheter secured in the left atrium to collect data from the LA and LV respectively as needed. Catheters were secured with purse-string sutures, and flow probes (ME 13 PXN, Transonic connected to a T420 multichannel research console, Transonic Systems Inc.) were positioned around the aorta to measure cardiac outflow. Data acquisition was conducted with a PowerLab 35 series system (ADInstruments) at a sampling frequency of 1 kHz. Real-time monitoring and analysis were performed using LabChart Pro v8.1.16 (ADInstruments), and pressure and flow data were processed with a 10 Hz low-pass digital filter to remove high-frequency noise. Epicardial imaging was performed using 2D echocardiography (Philips Epiq CVx cardiovascular ultrasound system) to validate LA contractility, with visualization and analysis conducted in Q-Vue 2.2 software (Philips).

[0586] 61

[0587] 45802538.1 The validation process began with collecting invasive hemodynamic data, including left atrial pressure, LA volume, left ventricular pressure, systemic blood pressure, and cardiac outflow during native, healthy conditions. To mimic left atrial appendage occlusion and evaluate its hemodynamic effects for validating the lumped parameter model, the pig’s LAA was externally clipped using a hemostat clamp. Pre- and post-clamp hemodynamics were measured using the PV catheter positioned within the left atrium. After collecting native baseline hemodynamic and wall biomechanics data from live pigs, the native left heart was surgically bypassed with the soft robotic LV to assess the model’s capability to replicate systemic hemodynamics in the porcine circulatory system. The bypass procedure involved directing inflow to the soft robotic LV by cannulating the pig’s left atrium and connecting it to the inflow graft of the robotic LV. This graft included a St. Jude mechanical valve functioning as a surrogate mitral valve to regulate flow directionality. The outflow graft, also equipped with a St. Jude mechanical valve to mimic the aortic valve, was connected to the pig’s native aorta via a cannula. The animals were anticoagulated with 5,000 units of heparin IV, and after the bypass circuit was established, the animals were humanely euthanized with pentobarbital at a dose of 100 mg / kg body weight. This step allowed the exclusive evaluation of the soft robotic LV’s performance by eliminating any contributions from the native LV. The soft robotic LV was then actuated right away to assess its ability to generate systemic pressures and flows. This experiment tested whether the soft robotic LV could independently recreate systemic hemodynamics within the swine circulatory system of realistic preload, afterload, vascular resistance, and compliance values.

[0588] Results

[0589] Design of a synthetic, patient-specific soft robotic left atrium and atrial appendage Atrial fibrillation (AF) is characterized by a highly irregular heart rhythm due to uncoordinated atrial signals. This disorganized activity impairs effective atrial contraction, leading to blood stasis, particularly in the left atrial appendage (LAA), where 90% of thrombi originate17. As a result, LAA occlusion is sometimes pursued for stroke prevention in patients who are not ideal candidates for long-term anticoagulation (Figure 21). Analysis of patient imaging data confirmed the significant anatomical inter-patient variability of the LAA (Figure 2J), which contributes to challenges in implantation and peri-device leakage. Patient-specific left atrium models were derived and segmented from high-resolution, time-resolved dynamic computed tomography (CT) scans, allowing for an accurate representation of both healthy and atrial fibrillation-affected geometries, all at the 0% of the R-R interval, corresponding to the onset of ventricular systole. (Figure 2D-2F). These segmented models were then used to create 3D-printed LA structures using a soft, elastomeric commercial resin including detailed left atrial appendage morphology (Figure 2G). The 62

[0590] 45802538.1 resulting models served as the anatomical foundation for the integration of McKibben soft robotic actuators in alignment with the underlying anatomical features (Figure 2H). These actuators were placed biomimetically, guided by the native myocardial fiber orientations derived from anatomical studies in clinical literature (Figure 2K). Native myocardial fibers in the LA are organized into distinct bundles with anisotropic alignment, transitioning from oblique to transverse and horizontal in the lower anterior and posterior-inferior walls, and oblique or longitudinal in the roof-anterior wall18,19. These complex orientations facilitate coordinated inward contraction for the atrial kick during systole and stretch-recoil dynamics during diastole, supporting reservoir and conduit functions of the LA for pulmonary venous return. The LAA has an encircling spiral myocardial fiber architecture that generates a wringing motion for effective blood ejection18 19. To mimic these mechanics, actuators were biomimetically placed along the anterior, posterior-inferior, and LAA regions of the LA (Figure 2K), allowing coordinated contraction, expansion, and wringing motions for realistic atrial and appendage dynamics.

[0591] The global myocardial fiber orientation map was simplified into four discrete regions of interest foractuator placement and provide precise control over contractile motion (Figure 2L). Digital photos of the assembled soft robotic LA model demonstrate the anatomical accuracy and the integration of the soft robotic actuators, showing their placement along the atrial and appendage walls (Figure 2M). The actuators allowed biomimetic wall motion, including atrial contraction ("kick") during the systolic phase and relaxation during diastole, simulating native atrial pumping and filling functions.

[0592] The motion generated from the placement of actuators along these key regions, guided by fiber orientation, was validated through M-mode ultrasound imaging (Figure 15A). The input pressure for the actuators was systematically varied from 0 to 40 psi to assess the range of wall displacement. Ultrasound imaging (Figure 15B) confirmed the effectiveness of individual actuators in producing tunable wall motion. For the LA wall, actuators achieved displacements ranging from 2 to 12 mm with input pressures of 2 to 20 psi and for the LAA wall, displacements of up to 9 mm were achieved with pressures up to 40 psi (Figure 15C). These results demonstrated that the actuators could reliably replicate both healthy and impaired contractile behaviors, providing the flexibility to model a variety of atrial pathophysiologies. This biomimetic contractile motion can support the ejection of blood from the LA and LAA and prevent stagnation, important for reducing thrombus formation in healthy physiology. Diseased conditions were also mimicked, such as AF, where the LA and LAA lose the coordinated contractile motion, leading to blood stasis and an increased risk of thromboembolism.

[0593] 63

[0594] 45802538.1 Recreating healthy left-sided hemodynamics with atrial contraction

[0595] The soft robotic left atrium model was integrated with a soft robotic left ventricle model, both segmented from dynamic CT imaging data, to create a comprehensive system capable of reproducing left-sided cardiovascular hemodynamics (data not shown). The design and placement of actuators in the LV model were previously developed through computational and experimental methods, allowing realistic contractile motion and hemodynamic performance (Example 1)20,21. The soft robotic LV eliminates the need for an external pulsatile pump by generating systemic cardiovascular hemodynamics on its own. While the primary focus of this study is on recreating pressures and flows associated with the LA, including the precise replication of the “atrial kick” and A-wave characteristic of atrial contraction, the use of a patient-specific dynamic LV model reinforces the system's ability to replicate the entire left-sided cardiac and circulatory hemodynamics. This integration is important for the successful development of an LA simulator, as it allows for studying the effects of atrial contraction and wall mechanics on overall cardiac function in both healthy and diseased states.

[0596] The mock circulatory flow loop was specifically designed to simulate systemic circulation, incorporating adjustable parameters such as preload, afterload, vascular compliance, and resistance (Figure 5A, Data not shown). The system utilized two clinically standard mechanical valves (mitral and aortic) to allow unidirectional flow through the circuit, with the soft robotic LV functioning as the primary pump to drive fluid flow (FIG. 23). The LA and LV models were controlled using a custom electropneumatic control box (FIG. 24), for synchronization of atrial and ventricular contraction at clinically relevant heart rates (e.g., 40-150 bpm). The loop featured integrated pressure and flow sensors for real-time hemodynamic measurements and the model was equipped with a port for an endoscopic camera, providing visualization of endocardial structures. The actuation of soft robotic elements on the LA and LAA successfully replicates the atrial kick, a dynamic contraction crucial for active ventricular filling, which cannot be achieved with passive 3D-printed models (data not shown). Figure 16A and data not shown highlight the physiologic left atrial pressure waveforms generated by the system, including the characteristic a-, c-, and v-waves, corresponding to the pump, reservoir, and conduit phases of the LA cycle. These pressure waveforms were synchronized with the LV contraction to replicate realistic cardiac cycles, demonstrating their ability to simulate the functional interplay between atrial contraction and ventricular filling. The a-wave reflects atrial contraction during late ventricular diastole, the c-wave represents a transient increase in left atrial pressure caused by ventricular contraction and mitral valve bulging, and the v-wave arises from passive atrial filling during ventricular systole, collectively illustrating the pump, reservoir, and conduit phases of the left atrial cycle22. The soft 64

[0597] 45802538.1 robotic LV generated biphasic ventricular pressures (120 / 6 mmHg) and drove systemic circulation, producing phasic aortic systemic pressures of 115 / 60 mmHg at a sinus rhythm of 60 bpm (Figure 16B). When combined with the soft robotic LA, the system reproduced physiologically accurate flow waveforms for left-sided circulation. Figure 16C demonstrates alternating mitral and aortic flow, with a cardiac outflow of over 5 L / min. The mitral valve flow waveform exhibited distinct E and A wave regions, representing passive ventricular filling during early diastole and active filling driven by atrial contraction, respectively. These findings confirm the capability of the soft robotic LA to contribute to active ventricular filling, mimicking the functional role of atrial contraction in the cardiac cycle. The soft robotic LA-LV system is compatible with clinical imaging modalities such as echocardiography (Figure 16C). Pulsed wave Doppler imaging was used to measure fluid flow velocities in the soft robotic LA, visualizing E and A wave patterns associated with mitral flow23. This capability allows for real-time, clinically relevant assessments, making the system suitable for studying both physiological and pathological atrial mechanics and hemodynamics in a controlled environment.

[0598] Integration and validation of the soft robotic LV in swine circulatory system

[0599] To validate the functionality of the soft robotic left ventricle model to drive systemic hemodynamics, it was integrated into a swine circulatory system, forming a hybrid synthetic-biological configuration. As shown in Figure 17A, the native LV was bypassed, and the soft robotic LV was connected to the native vasculature via inflow cannulas from the LA to the robotic LV and outflow grafts from the robotic LV to the ascending aorta. Systemic blood pressure and invasive left-sided hemodynamics were first recorded in a live porcine model with native LV (Figure 17B). The animals were then euthanized and the setup aimed to replicate the function of the native left-sided heart by pumping blood systemically with soft robotic LV. The soft robotic LV actuation was triggered using a custom electro-pneumatic control system at the matching heart rate to the data collected in the living model, ensuring seamless comparison between the biological and synthetic LV. Figure 17C demonstrates the hemodynamic performance of the soft robotic LV, generating clinically relevant systemic blood pressures (~ 68 / 40 mmHg), ventricular outflow (~ 4 L / min), and left ventricular pressures (~ 75 / 2 mmHg). The soft robotic LV successfully sustained systemic circulation, replicating the hemodynamic profiles of the native LV, including the phasic patterns of pressure and flow. The successful validation of the soft robotic LV into a swine circulatory system establishes its capability to replicate left-sided cardiac function under physiological conditions and provides a versatile platform for its use in mock flow loops for preclinical studies to model and address complex cardiovascular conditions and test interventions.

[0600] 65

[0601] 45802538.1 Multimodality imaging confirms mimicry of physiological wall and valve function The biomimetic, 3D-printed design of the soft robotic models successfully replicates the anatomy and physiological motion of the left atrial and ventricular wall and the function of heart valves (Figure 18 and movie not provided). Epicardial echocardiography was employed to capture real-time images of the LA and LAA during their dynamic actuation in tandem with the soft robotic LV and mechanical valves. The echocardiographic images illustrate the displacement of the LA wall between atrial diastole and systole (data not shown). During atrial systole, the inflation of the soft robotic actuators mimics native atrial contraction, resulting in inward displacement of the LA wall. Conversely, deflation during diastole allows the wall to return to its resting position, replicating the passive reservoir phase of the LA. This biomechanical and functional mimicry, achieved via the inflation and deflation of the soft robotic actuators, aligns closely with physiological motion patterns observed in the native atrium. Furthermore, the LAA ostium shows a measurable reduction (> 66%) in diameter during systole, as highlighted in the echocardiographic images, effectively replicating the contractile motion that promotes blood ejection from the LAA. The synchronized function of the soft robotic LV and LA were measured (data not shown).

[0602] Echocardiographic images captured during ventricular systole and diastole illustrate the ability of soft robotic muscles to create coordinated motion of the LV wall, mirroring native ventricular contraction and relaxation. This synchronization, achieved through the electropneumatic control system, allows a seamless interplay between the LA and LV, ensuring effective blood flow within the system. Endoscopic and echocardiographic images confirm the opening and closing of the valves in synchrony with the actuation of the soft robotic LA and LV. During ventricular diastole, the mitral valve (MV) opens to facilitate passive filling of the LV, observed in clear circular aperture (data not shown). During systole, the MV closes tightly to prevent retrograde flow, for unidirectional blood movement. Similarly, the aortic valve (AoV) exhibits characteristic behavior, opening fully during ventricular systole to allow forward blood flow into the systemic circulation and closing during diastole to maintain systemic pressure (data not shown). The anatomical orientation of the soft robotic LA / LAA model in coronal, sagittal, and axial planes, with structural MRI highlighting the placement of the actuators were measure (data not shown). Cine MRI images (data not shown) further illustrate the dynamic wall motion between reservoir and pump phases, with the LA area changing significantly during systole and diastole, (data not shown). Wall displacement driven by input actuation pressures of 10, 12, and 15 psi resulted in LA long -axis cross-sectional area variations between 15 and 23 cm2, replicating physiologically relevant atrial reservoir and pump phases24. These results demonstrate that the soft robotic myocardial substitute can faithfully replicate the physiological wall motion of the LA and LAA, as well as the dynamic 66

[0603] 45802538.1 functionality of the mitral and aortic valves. The system’s compatibility with clinical imaging modalities, such as MRI, echocardiography, and endoscopy, further highlights its potential as a platform for studying cardiac wall and valve mechanics in both healthy and pathological states.

[0604] Intra-atrial flow velocity measurement via phase-contrast MRI

[0605] Conventional rigid benchtop simulators lack the dynamic wall motion required to replicate physiological intra-atrial flow energetics, limiting their ability to provide meaningful insights. In contrast, the soft robotic LA / LAA model, with its MRI compatibility and high anatomical and functional precision, allows for the visualization and quantification of flow patterns and velocities within the LA and LAA under sinus rhythm using a 3T MRI system (Figure 25). Phase-contrast 2D flow imaging sequences were utilized to measure intra-atrial velocities and analyze flow patterns generated by the soft robotic actuators (data not shown). The flow streamlines show that fluid flow velocity was the highest (35-40 cm / s) at the right pulmonary vein junction and the blood flow was also slower (10-15 cm / s) at the left pulmonary vein junction, a pattern also seen in human patients in literature25. During the reservoir phase, peak flow velocities were lower (15-25 cm / s) and uniformly distributed, while the conduit and pump phases exhibited higher peak velocities (30-40 cm / s) driven by LV suction and atrial contraction (atrial kick), mirroring physiological behavior26. Velocity-time graphs (Figure 19A) show dynamic velocity changes within the LA and LAA throughout the cardiac cycle, with peak velocities in the range of 30-40 cm / s, consistent with literature (25-45 cm / s; e.g., 26.8 ± 5.5 cm / s in controls, 30-50 cm / s in MRI flow studies)27. Mean velocity measurements across actuation pressures (10, 12, and 15 psi) for the LA (Figure 19B) and LAA (Figure 19C) further confirmed the system’s ability to generate physiologically relevant flow velocities. At 15 psi, the mean velocities reached approximately 18 cm / s in the LA and 24 cm / s in the LAA, consistent with reported values in atrial (patho)physiology (15-25 cm / s in LA, 20-30 cm / s in LAA)26’28’29. Patient studies of paroxysmal AF report depressed mean velocities (10-13 cm / s in LA, <20 cm / s in LAA). This study demonstrates that the integration of soft robotics into the model facilitates the accurate reproduction of clinically relevant intra-atrial and appendage flow velocities, which are important for replicating flow behaviors, including stagnation, observed in pathological conditions such as atrial fibrillation.

[0606] Replicating pathological hemodynamics of atrial arrhythmias

[0607] This study demonstrates the ability of the soft robotic LA / LAA model to replicate pathophysiological hemodynamics characteristic of atrial arrhythmias, such as atrial flutter or fibrillation, including reduced atrial contractility and its associated impact on ventricular filling and cardiac outflow (CO). By modulating the actuation input pressure on the soft robotic actuators and varying actuation rates, the model effectively mimics the lack or reduction of atrial contractility 67

[0608] 45802538.1 observed during atrial arrhythmias and provides insights into the resulting hemodynamic alterations. As shown in Figure 19D, the patient-specific LA and LAA morphology associated with AF was recreated, featuring an enlarged atrium and diminished myocardial contractility, mimicked by incorporating only two soft robotic actuators. Figure 19E demonstrates the replication of mitral flow velocities and left atrial pressures during atrial arrhythmias (e.g., atrial flutter), with a noticeable absence of strong atrial contraction (atrial kick). The pressure-volume loop (Figure 19F) further mimics the impact of atrial arrhythmias on LA hemodynamics. Under sinus rhythm, the PV loop exhibits characteristic figure-eight a- and v-loops, corresponding to active atrial contraction (a-wave) and passive atrial filling (v-wave). In contrast, during atrial flutter, the PV loop is reduced to a v-loop alone, indicating the absence of active atrial contribution.

[0609] By varying heart rates from 40 to 150 bpm (with a 1:1 atrio-ventricular coupling), the model captured the hemodynamic changes associated with slow ventricular response (S VR), controlled ventricular response (CVR), and rapid ventricular response (RVR) during atrial flutter or sinus tachycardia scenarios. Figure 19G demonstrates that as heart rate increased, left atrial pressures (LAP), left ventricular pressures (LVP), and CO were significantly altered. During SVR (40 bpm), LAP and CO remained within physiological ranges due to sufficient ventricular filling time, but LVP showed a pronounced relaxation peak, highlighting impaired LV relaxation dynamics. Under CVR (60-80 bpm), normal atrial contraction maintained efficient LAP, LVP, and CO. However, during RVR (150 bpm), the reduced ventricular filling time led to diminished CO, increased diastolic LVP, and abnormal LAP dynamics, closely mimicking the hemodynamic effects of atrial flutter and rapid sinus tachycardia. The benchtop simulator was constrained to regular activation sequences, precluding reproduction of beat-to-beat variability intrinsic to AF. The effects of AF (rate irregularity) on systemic hemodynamics have been explored by a complementary lumped parameter model as explained in the data not shown. Specifically, AF simulations revealed elevated atrial pressures and volumes and reduced ventricular filling compared with sinus rhythm, despite similar average rates.

[0610] The influence of LA contractility on ventricular filling and cardiac outflow was quantified by systematically varying the actuation input pressure of the LA soft robotic actuators. Figure 19H shows that increasing input pressure led to greater mitral flow peak A velocities, higher LAP peak a values, and enhanced CO, with strong correlations (r > 0.94) between these parameters and LA actuation pressure. This finding highlights the importance of atrial contractility in maintaining efficient ventricular filling and systemic outflow. The effect of LAA contractility on LAA ejection fraction was assessed by varying the LAA actuation input pressure. As shown in Figure 191, higher actuation pressures correlated with increased LAA ejection fraction, transitioning from impaired 68

[0611] 45802538.1 emptying observed in AF to healthy ejection levels characteristic of sinus rhythm (SR). The ability to replicate this pathophysiological feature emphasizes the utility of the model in studying LAA dynamics and potential thrombus formation in conditions such as AF30-32. Color Doppler echocardiography (Figure 19J) provided a real-time visualization of simulated LAA flow under both healthy and pathological conditions. In healthy SR (60 bpm), ordered flow through the LAA ostium was observed, whereas atrial flutter (150 bpm) demonstrated reduced flow across the LAA ostium, a known contributor to thrombus formation and embolic risk.

[0612] Device sizing and placement for LAA occlusion in patient-specific models

[0613] To demonstrate testing and adjusting device sizing and placement for LAAO in patientspecific anatomies using our simulator, occlusion devices were deployed into the LAA to evaluate their performance and assess the degree of closure. For this patient-specific model, two WATCHMAN FLX™ devices, sized 24 mm and 27 mm, were tested by an operator guided by echocardiography (Figure 20A). Real-time echocardiography and endoscopy imaging allowed precise placement of the devices. The WATCHMAN device functioned within the pulsatile environment of the soft robotic LA model (data not shown). The level of occlusion achieved by the devices was assessed using 2D color Doppler imaging (Figure 20B). The 24 mm device showed incomplete occlusion, evidenced by the presence of color jets behind the device, indicating residual flow through the LAA. In contrast, the 27 mm device demonstrated effective occlusion, with no visible color jets, confirming more complete closure of the LAA. This highlights the importance of proper device sizing in achieving occlusion and validates the ability of the model to assess device performance in real-time.

[0614] To further confirm appropriate device placement, a phase contrast MRI was performed, providing detailed visualization of the implanted device within the LAA (Figure 20C). The gray shading in the MRI images represents the implanted device, clearly visible within the LAA. Postimplantation analysis showed the absence of the LAA in the reconstructed fluid volume, indicating successful closure with the 27 mm device. This study highlights the utility of the patient-specific soft robotic simulator to guide device placement in complex anatomies, providing operators with a platform to practice and adjust deployment strategies in realistic clinical scenarios.

[0615] Lumped parameter modeling of LAA occlusion effects on LA hemodynamics

[0616] To understand how left atrial appendage occlusion affects left atrial filling and emptying, a lumped parameter model was developed. The model simplifies the cardiovascular system into a set of ordinary differential equations that represent pres sure- volume relationships, flow dynamics, and vascular resistances. Figure 21A illustrates the electrical analog of the LPM, which models the systemic and pulmonary circulations as a network of resistances and capacitances, allowing for the 69

[0617] 45802538.1 simulation of LAAO effects on LA hemodynamics. Figure 21B compares the mean left atrial pressure (mLAP) before and after LAAO, showing that the model predicted an increase in mLAP from 17.1 mmHg to 19.5 mmHg post-LAAO. This reflects the added load on the LA due to reduced reservoir function of the LAA, consistent with clinical observations of mild elevations in LA pressure following occlusion33. The mitral flow waveform revealed a slight increase in peak velocities post-LAAO, potentially suggesting an enhanced atrial contribution to ventricular filling (Figure 21C). The PV flow waveform, however, displayed a subtle increase in backward flow (negative velocities), indicating slight retrograde flow into the pulmonary veins during atrial contraction post-LAAO.

[0618] To validate the LPM, measurements from a porcine in vivo model were compared to the simulated results (Figure 21D). The in vivo data showed a similar increase in mLAP from 15.8 ± 1.3 mmHg before LAAO to 17.3 ± 0.9 mmHg after LAAO, closely aligning with the LPM predictions. The waveforms for LAP also showed comparable patterns, confirming the model's accuracy in replicating LA hemodynamics and studying the nuanced effects of LAAO in a closed-loop cardiovascular system. This increase in mLAP has also been reported in clinical literature in human patients33,34.

[0619] Finite element analysis of atrial fibrillation impact on LA mechanics

[0620] The impact of AF on left atrial wall mechanics and pressure- volume hemodynamics was investigated using a dynamic finite element analysis model based on the SIMULIA Living Heart Project. The model integrates realistic electrical, structural, and fluid flow physics to simulate dynamic response of the heart, allowing the study of coupled electromechanical behavior where electrical excitation drives mechanical contraction. Following the electro-mechanical simulation of sinus rhythm, the model was modified to represent AF (as described in the Methods section) and analyzed over three sequential irregular cardiac cycles.

[0621] Predicted LA wall stress during its function as a conduit, pump, and reservoir is presented in Figure 21E for both SR and AF. Relatively uniform stress distributions were observed during SR, reflecting healthy tissue mechanics and functional coordination. In contrast, AF disrupted these coordinated mechanics, leading to irregular and elevated stress distributions. Results demonstrate a marked increase in LA wall stress during AF, specifically, mean LA wall stress over a cardiac cycle increased from 3.17 kPa in SR to 19.54 kPa in AF. The loss of mechanical coordination during AF created stress concentration zones, which are known to act as sites of microtrauma or remodeling These high-stress regions can predispose the atrium to structural abnormalities such as dilation, wall thinning, and fibrotic tissue deposition, potentially contributing to the progression of AF3 37.

[0622] 70

[0623] 45802538.1 In addition to wall stress, LA pressure and volume data were analyzed to evaluate hemodynamic differences between SR and AF and their correlation with wall stress (Figure 21F).

[0624] The analysis revealed that AF significantly increased mean LA pressure, rising from 7.04 mmHg in SR to 14.49 mmHg in AF (Figure 21 G). The pressure-volume curve demonstrated the characteristic figure-eight shape in SR, while the a-loop was absent in AF. Elevated wall stress and stiffness were associated with reduced compliance, leading to a steeper pressure-volume curve in SR compared to AF (Figure 21G). The shape of the pressure- volume loop generated by the FEA model was further validated by pressure- volume measurements obtained from a healthy live porcine in SR (Figure 21H). By integrating pressure- volume analysis with wall stress data, the model highlights the interplay between mechanical and hemodynamic dysfunction in AF. The FEA model incorporated simplified fluid cavities to impose physiological pressure- volume loading on the atrial wall. No full computational fluid dynamic (CFD) simulations of blood flow were performed; rather, the cavity elements served as boundary conditions to couple chamber pressures with structural deformation.

[0625] Comparison against human imaging and porcine hemodynamic data

[0626] The soft robotic LA and LAA model was validated against clinical imaging data from human patients and hemodynamic data from a porcine model, demonstrating its accuracy in replicating physiological and pathological atrial mechanics and hemodynamics. Figure 22A compares the LAA wall motion in human patients (Patient 4 and Patient 7) with that of the simulator during atrial diastole and systole. The echocardiographic images show a close match in LAA contraction dynamics, with the simulator accurately replicating the inward contraction of the LAA ostium observed in human imaging. Quantification of LAA ostium contraction percentages (Figure 22B) further confirms this accuracy, with the simulator achieving a contraction range of 0-79.5%, covering the range of the variability observed across human patients. Figure 22C provide direct visualization of the LAA ostium motion during atrial systole and diastole using endoscopic imaging (movie not shown). The simulator captures the cyclic contraction and relaxation of the ostium, closely mimicking the anatomical behavior observed in humans (e.g., Patient 1), demonstrating the simulator's capability to recreate intricate LAA geometrical changes during the cardiac cycle.

[0627] Figure 22D compares LA area changes during diastole and systole across porcine, human, and simulator data. The echocardiographic images highlight similar LA contraction and relaxation patterns across all three cases, with the simulator showing comparable inward wall motion during systole. Quantitative analysis of the LA area change (percentage reduction from diastole to systole)

[0628] 71

[0629] 45802538.1 reveals no significant difference among the porcine (37.4 ± 3.8%), human (39.2 ± 4.1%), and simulator (36.8 ± 4.7%), indicating the model's ability to replicate atrial function with high fidelity.

[0630] The replication of LA hemodynamics was validated by comparing LAP waveforms from porcine in vivo measurements and the simulator (Figure 22E). Both datasets exhibit similar characteristic waveforms, reflecting the pump, reservoir, and conduit phases of the atrial cycle. The peak LAP values of the simulator closely matched those observed in the porcine model, with the simulator maintaining physiologically relevant pressure ranges.

[0631] The present study introduces a multimodal framework that integrates experimental and computational approaches to replicate left atrial function with high fidelity in sinus rhythm, atrial fibrillation, and atrial flutter for testing stroke prevention strategies. Combining a soft robotic benchtop simulator, lumped parameter modeling, and finite element analysis, the framework overcomes the limitations of traditional methods such as reliance on animal models and cadaveric specimens. This model suite recreates physiological hemodynamics and motion, with the potential to improve LAAO outcomes by (i) facilitating device testing and development in a representative and physiologically relevant model, (ii) allowing patient-specific procedural planning and simulation, and (iii) serving as a tool to better understand left atrial hemodynamics and postprocedural physiological changes following LAAO.

[0632] The soft robotic benchtop simulator replicates the anatomy, hemodynamics, and biomechanics of the left atrium and left ventricle of the heart. The use of 3D-printed LA geometries, coupled with soft robotic actuators, allows precise recreation of contractile motion, including the atrial kick, which is often neglected in existing models. Besides 3D printing, the models of the LA and LA A can potentially also be created using simultaneous utilization of additive manufacturing and traditional casting / molding techniques using elastomers (e.g., silicone) or Polyvinyl Alcohol (PVA) cryogel, to accurately replicate tissue mechanical properties and surface finish, for adaptability to user needs. Integration of a soft robotic left ventricle further enhances the system by overcoming the limitations of traditional pulsatile pumps, which exhibit paradoxical behavior when attached to passive ventricular models. The soft robotic left ventricle was validated in a hybrid synthetic-biological configuration with a swine circulatory system. In this setup, the soft robotic LV replaced native ventricular contraction to sustain systemic circulation, integrating with the native vasculature. This validation highlights the robustness of the mock flow loop and its ability to recreate clinically relevant hemodynamics in a dynamic and reproducible manner for benchtop testing. Importantly, the echocardiography and MRI-compatible design of the simulator allow detailed structural imaging and measurement of intra-atrial flow dynamics, further enhancing its utility in studying complex conditions such as AF. Additionally, the use of an internal 72

[0633] 45802538.1 imaging system with a clear blood-mimicking fluid allows for direct visualization of various internal structures in a way that surpasses current in vivo methods. While intracardiac echocardiography (ICE) is the closest available option for real-time cardiac imaging in patients, direct endoscopic visualization is not feasible outside of open cardiac surgery, where surgical manipulation alters hemodynamics. In contrast, this model provides unobstructed, high-fidelity visualization of intra-cardiac structures and flow dynamics under controlled conditions, making it a superior alternative to in vivo studies for device testing and mechanistic research.

[0634] The LPM and FEA components of the framework complement the experimental simulator by providing mechanistic insights into systemic and localized hemodynamic phenomena. The LPM allows predictive modeling of hemodynamic changes, such as atrial inflow, outflow, and pressures pre- and post-LAAO intervention. The FEA model adds granularity by quantifying biomechanical stress distributions and pressure- volume relationships, elucidating the structural and mechanical implications of atrial fibrillation. These computational tools bridge the gap between bench testing and clinical applicability, allowing comprehensive analyses that extend beyond experimental models alone. Although the results from the three models are presented independently in this study, the framework can be integrated into a hybrid experimental-computational configuration, allowing dynamic data exchange, cross-validation, and the use of one model’s outputs as boundary conditions for another. This interconnected approach could offer comprehensive insights, addressing multifaceted hypotheses and bridging system-level and localized analyses in the future. LPM model calibration and comparison were performed using the porcine dataset collected in this study, which provided close agreement between simulations and experiments.

[0635] The suite of benchtop and computational framework successfully simulated the reservoir, conduit, and pump modes of the LA and replicated the clinical structure-function parameters of pressure, volume, and flow. By adjusting circuit and control system parameters and by selectively activating-deactivating soft robotic actuators, the ability to simulate a wide range of atrial dysfunctions was demonstrated. These included clinically relevant abnormalities in contractility, compliance, LAA emptying fraction, arrhythmias, and LAA velocity, with the resulting effects captured in systemic hemodynamic parameters such as left ventricular filling and cardiac outflow. This capability makes the platform uniquely suited for studying not only normal physiological conditions but also pathological scenarios, offering insights into the systemic impacts of left atrial and left atrial appendage dysfunctions.

[0636] The ability to replicate device settings, systemic hemodynamics, and localized biomechanical stresses demonstrates the robustness of the entire framework, making it suitable for diverse applications, including intervention testing, clinical research, procedural planning, and education.

[0637] 73

[0638] 45802538.1 For instance, this flexible framework involving a suite of three independent models can allow users to adopt specific components to suit their unique needs. Device designers can leverage the benchtop platform to test early prototypes of their devices under realistic conditions. Clinical researchers can employ the LPM and FEA platforms to investigate specific research questions, such as understanding the impact of disease progression or interventions on systemic and localized hemodynamics. Clinicians can use the simulators for procedural planning, refining procedural techniques, surgical trainee practicing, and educating patients by demonstrating intervention dynamics on a beating heart model38.

[0639] While the simulator demonstrates strong potential for translational research, it is important to delineate its intended applications. In clinical practice, device selection is largely driven by anatomical considerations visible on high-resolution imaging, and the choice between commercially available LAAO device sizes (e.g., 24 mm vs 27 mm) can typically be accomplished with imaging guidance and operator expertise alone. The value of the present platform lies in part in areas where conventional imaging may be insufficient, such as in borderline or highly complex anatomies, or in the early stages of device prototyping and development. In such contexts, direct deployment of a device into a patient-derived anatomy within the simulator can reveal subtleties of performance, including peri-device leakage or incomplete seal, that may not be apparent from imaging alone. Thus, the simulator provides unique advantages for device innovation, training in technically challenging anatomies, and mechanistic studies of device-tissue-flow interactions under physiologically realistic conditions. The present study introduces a proof-of-concept framework. To illustrate potential impact, a case example includes the following: a patient with a challenging chicken-wing LAA morphology, where device choice is less straightforward. In such a scenario, the simulator could test multiple occluder designs in the patient-derived anatomy, revealing risks such as peri-device leak or embolization under flow, insights not easily obtained from imaging alone. The visual comparisons of device deployments serve as a preliminary demonstration of this concept. While clinical imaging is generally sufficient for device sizing, the platform adds value by allowing functional testing of deployment and stability.

[0640] The ability to simulate patient-specific anatomy and hemodynamics can allow clinicians and trainees to practice device deployment and evaluate potential outcomes before performing actual procedures. This capability is particularly relevant for LAAO, a highly complex intervention requiring precise device placement and a deep understanding of patient-specific anatomy and flow dynamics. Clinical developers can leverage this beating heart simulator to train physicians in rehearsing complex clinical device deployments, for precision and confidence in real-world procedures39. Furthermore, the capability to generate datasets representing both healthy and 74

[0641] 45802538.1 diseased states paves the way for this framework to be adopted in regulatory settings, potentially accelerating device development in the future12. Most of the above described validation is qualitative or semi-quantitative. For instance, the simulator’s LAA ostial contraction spanned 0-80%, whereas patient values were 20-50%. This wider range reflects extreme boundary conditions that were tested (e.g., atrial standstill or exaggerated contraction), while under physiologic sinus rhythm the simulator produced -40% contraction, within the clinical range. Similarly, pressure and flow patterns aligned with patient trends, though with broader variability due to the simulator’s exquisite tunability. Quantitative benchmarks against literature further support physiological realism (e.g., LAA emptying velocities of -50 cm / s in sinus rhythm and -15 cm / s in AF, consistent with reported ranges). The platform therefore is valuable for capturing relative trends, supporting device development, and serving as a training and research tool.

[0642] The present study highlights the utility of the simulator for preclinical testing and mechanistic understanding in atrial fibrillation. However, this workflow can be equally adaptable for other LA-focused interventions40'41. For example, heart failure with preserved ejection fraction (HFpEF) represents a complex clinical syndrome characterized by symptoms of heart failure, a preserved ejection fraction, and evidence of diastolic dysfunction. Unlike heart failure with reduced ejection fraction (HFrEF), where several evidence-based therapies have shown significant benefits, effective treatments for HFpEF remain limited, emphasizing a unmet medical need. One emerging approach involves designing devices to reduce diastolic pressure in the left ventricle by strategically alleviating preload from the left atrium41-43. These devices create a shunt that allows blood to flow directly from the left atrium to the aorta, bypassing the left ventricle. The efficacy and safety of such devices involves rigorous preclinical testing to adjust their design and functionality. Herein, lies another potential role of the present benchtop beating left heart simulator, designed to replicate the biomechanics and hemodynamics of the left atrium. This simulator’s ability to mimic patientspecific left atrial anatomy and function, along with the entire left-sided circulation, can allow a thorough evaluation of how shunt devices interact with various anatomical configurations and disease states. By recreating the precise anatomical and physiological conditions of the left atrium, including the hemodynamic challenges of HFpEF, the simulator serves as a tool for assessing device performance under highly controlled yet clinically relevant conditions.

[0643] Notably, all intracardiac LAAO devices require transseptal puncture for deployment, a step that is conventionally left unclosed due to its tendency to seal spontaneously. However, in some cases, persistent interatrial shunting occurs, which is usually inconsequential but occasionally problematic. Understanding how transseptal punctures behave in different hemodynamic states

[0644] 75

[0645] 45802538.1 could be an area for further study using this framework, particularly in the context of postprocedural atrial remodeling and long-term interatrial shunt persistence.

[0646] Beyond simulating LA hemodynamics and motion, the simulator also supports comprehensive hemodynamic modeling of the entire left-sided cardiac circulation. The mechanical valves integrated into the simulator can be replaced with bioprosthetic valve designs, allowing it to replicate pathological conditions such as mitral and aortic valve diseases44. This capability extends to testing valve replacement scenarios. For example, the mechanical mitral valve in the model, which precisely replicates mitral inflow during diastole and occlusion during systole, can be substituted with bioprosthetic valves to investigate various mitral valve disease states and repair or replacement interventions45. This adaptability makes the soft robotic simulator a versatile platform for studying valve interventions and their effects on hemodynamics, offering significant potential to advance device innovation and therapeutic strategies for complex cardiac conditions.

[0647] Stress modeling using the computational FEA model could aid in staging AF and refining patient selection for rhythm control. With rising AF ablation efficacy and expanding indications, there is an increasing need to predict who will benefit most. While broad rules exist, e.g., massive bi-atrial enlargement predicts poor rhythm control success, many patients fall into a clinical gray zone. If stress concentration patterns from computational modeling could help stratify patients for ablation, this could provide a valuable predictive tool. Integrating biomechanical stress markers with electroanatomic and clinical data may enhance AF management. While these insights highlight intriguing avenues for exploring atrial mechanics in arrhythmogenesis, the translation of such models into real-time rhythm-control strategies remains speculative and faces substantial practical hurdles. Personalizing a full electromechanical model to an individual patient remains highly challenging and time-intensive, requiring days to weeks for segmentation, meshing, and simulation. Consequently, even state-of-the-art personalized atrial models are currently limited to retrospective analyses and offline planning rather than intra-procedural use. Non-invasive imaging approaches, such as echocardiographic speckle-tracking to quantify left atrial reservoir and booster strain, already provide practical, patient-specific metrics of atrial function and fibrosis burden, and can often serve as accessible surrogates for model-derived parameters. The added value of finite element simulations lies in their ability to map spatially heterogeneous stress and strain distributions across the atrial wall, potentially revealing regional stress concentrations or remodeling substrates that global strain indices may not capture. Thus, the current role of such models as research and hypothesis-generation tools, aimed at exploring mechanistic links between atrial biomechanics and arrhythmogenesis. The present current FE model lacks many aspects

[0648] 76

[0649] 45802538.1 needed for ablation guidance (c.g., high-resolution fibrosis mapping, accurate rcpolarization dynamics, long-term simulation of inducibility)46,47.

[0650] Additional iterations could incorporate feedback-driven adjustments to simulate disease progression and chronic remodeling. Importantly, because the actuation is software-driven, additional versions of the simulator could incorporate randomized beat intervals to more faithfully reproduce AF hemodynamics. The models could benefit from integration with machine learning algorithms to increase predictive capabilities. The benchtop and computational models could serve as data generators for training machine learning algorithms to predict outcomes such as hemodynamics or device performance. In turn, machine learning could speed up or even replace complex simulations by creating a surrogate of the finite element model through deep learning that runs in milliseconds49. While 2D phase contrast MRI and echocardiography confirm the capability of model for flow and motion measurement, integrating PIV and 4D flow MRI in future could enhance 3D hemodynamic assessment and vortex analysis. While the system can replicate acute hemodynamic states, modeling long-term tissue remodeling would require incorporating biologically responsive materials or computational frameworks capable of simulating growth and adaptation. Nevertheless, actuator control could in principle be programmed to represent certain dilated or remodeled states, or even tuned to introduce low-frequency adjustments over time, providing a simplified means of approximating aspects of remodeling without reproducing the underlying biology. Additionally, although the present study focused on device evaluation and hemodynamic characterization, additional forms could facilitate pharmacological testing or thrombus modeling. For example, in vitro assays with blood or blood analogs or thrombogenic substrates could allow investigation of clot initiation and growth under different flow conditions, while in silico coupling with thrombogenesis models could provide dynamic predictions of thrombus formation and dissolution50. For patient-specific treatment design, accounting for segmentation uncertainty (e.g., bracketing best- and worst-case geometries) will be important. By contrast, applications in device development, comparative testing, and training are less sensitive to such variability.

[0651] This multimodal framework represents a transformative approach to studying LA function and interventions, offering unparalleled fidelity and versatility. A model suite capable of recapitulating biomechanics and measuring pressure and flow can serve as a valuable tool for investigating the impact of LA function or dysfunction on hemodynamic conditions that increase the risk of thrombus formation and stroke. By integrating experimental and computational techniques, it addresses longstanding gaps in cardiovascular research, advancing the fields of biomechanics, device development, and personalized medicine. Its applications in device testing,

[0652] 77

[0653] 45802538.1 clinical research, procedural planning, and training highlight its potential to improve patient outcomes and drive innovation in AF management.

[0654] References:

[0655] 1. Al-Saady, N„ Heart 82, 547-554 (1999).

[0656] 2. Madden, J. L. Journal of the American Medical Association 140, 769-772 (1949).

[0657] 3. Chugh, S. S., et al. Circulation 129, 837-847 (2014).

[0658] 4. Bisbal, F., Journal of the American College of Cardiology 75, 222-232 (2020).

[0659] 5. Blackshear, J. L. & Odell, J. A. The Annals of thoracic surgery 61, 755-759 (1996).

[0660] 6. Maarse, M., et al. JAMA neurology 81, 1150-1158 (2024).

[0661] 7. Alkhouli, M., et al. Journal of the American College of Cardiology 81, 1063-1075 (2023).

[0662] 8. Emmert, M. Y., et al. Journal of Cardio thoracic Surgery 15, 1-12 (2020).

[0663] 9. Yu, F„ et al. Basic to Translational Science 9, 971-981 (2024).

[0664] 10. Aycock, K. I., et al. Frontiers in Medicine 11, 1433372 (2024).

[0665] 11. Pathmanathan, P., et al. PLOS Computational Biology 20, e012289 (2024).

[0666] 12. Ahmed, K. B. R., Annals of Biomedical Engineering 51, 6-9 (2023).

[0667] 13. Ciobotaru, V., et al. Left atrial appendage occlusion simulation based on three-dimensional printing: new insights into outcome and technique. EuroIntervention 14, 176-184 (2018).

[0668] 14. Robinson, S. S., et al. Nature Biomedical Engineering 2, 8-16 (2018).

[0669] 15. Duenas-Pamplona, J., Sierra-Pallares, J., Garcia, J., Castro, F. & Munoz-Paniagua, J.

[0670] Boundary-condition analysis of an idealized left atrium model. Annals of Biomedical Engineering 49, 1507-1520 (2021).

[0671] 16. Mendez, K„ et al. Cardiovascular Engineering and Technology, 1-17 (2025).

[0672] 17. Cresti, A., et al. EuroIntervention 15, e225-e230 (2019).

[0673] 18. Goette, A., et al. Journal of Arrhythmia 32, 247-278 (2016).

[0674] 19. Mahajan, R., HeartRhythm Case Reports 6, 169-173 (2020).

[0675] 20. Park, C„ Device 2(2024).

[0676] 21. Park, C., Advanced Functional Materials 32, 2206734 (2022).

[0677] 22. ANKENEY, J. L., Circulation Research 4, 95-99 (1956).

[0678] 23. Laniado, S., et al. Circulation 51, 104-113 (1975).

[0679] 24. Ro§ca, M„ Heart 97, 1982-1989 (2011).

[0680] 25. Matsumoto, M., et al. Magn Reson Med Sci 19, 290-293 (2020).

[0681] 26. Evin, M„ et al. Left atrium MRI 4D-flow in atrial fibrillation: association with la function, in 2015 Computing in Cardiology Conference (CinC) 5-8 (IEEE, 2015).

[0682] 27. Sekine, T., et al. Magnetic Resonance in Medical Sciences 21, 293-308 (2022).

[0683] 78

[0684] 45802538.1 28. Demirkiran, A., et al. Scientific reports 11, 5965 (2021).

[0685] 29. Markl, M., et al. Journal of Cardiovascular Magnetic Resonance 17, 43 (2015).

[0686] 30. Wu, M.-Y., et al. Scientific Reports 9, 17864 (2019).

[0687] 31. Li, Y.-H., et al.. International Journal of Cardiology 68, 39-45 (1999).

[0688] 32. Akosah, K. O., CHEST 107, 690-696 (1995).

[0689] 33. Alarouri, H. S., et al. Heart Rhythm (2024).

[0690] 34. Tomotsugu Tabata, M„ et al. The American journal of cardiology 81, 327-332 (1998).

[0691] 35. Sohns, C. & Marrouche, N. F. European heart journal 41, 1123-1131 (2020).

[0692] 36. Lee, J.-H., et al. Frontiers in Physiology 12, 686507 (2021).

[0693] 37. Nattel, S., Circulation: Arrhythmia and Electrophysiology 1, 62-73 (2008).

[0694] 38. Wang, D. D., et al. Cardiovascular Imaging 14, 41-60 (2021).

[0695] 39. Dimitriadis, K., et al. Canadian Journal of Cardiology (2024).

[0696] 40. Inoue, K. & Smiseth, O. A. Journal of Cardiology (2024).

[0697] 41. Gordon, J. S., et al. Artificial organs 46, 2109-2117 (2022).

[0698] 42. Kado, Y., et al. The International Journal of Artificial Organs 44, 465-470 (2021). 43. Miyagi, C., Heart failure reviews 26, 749-762 (2021).

[0699] 44. Morray, B. H. Journal of the American College of Cardiology 77, 71-79 (2021).

[0700] 45. Harky, A., et al. Progress in Cardiovascular Diseases 67, 98-104 (2021).

[0701] 46. Boyle, P. M., et al. Nature Biomedical Engineering 3, 870-879 (2019).

[0702] 47. Gonzalo, A., et al. The Journal of Physiology 602, 6789-6812 (2024).

[0703] 48. Ozturk, C., et al. Advanced Science 12, 2404755 (2025).

[0704] 49. Alber, M., et al. npj Digital Medicine 2, 115 (2019).

[0705] 50. Guerrero-Hurtado, M., Computer Methods and Prog in Biomedicine 267, 108761 (2025).

[0706] 51. Singh, M., et al. Science Translational Medicine 16, eadk2936 (2024).

[0707] 52. Singh, M„ et al. Nature Cardiovascular Research, 1-17 (2023).

[0708] 53. Hu, L., et al. Nature Biomedical Engineering 7, 110-123 (2023).

[0709] 54. Roche, E. T., et al. Science Translational Medicine 9, eaaf3925 (2017).

[0710] 55. Doshi, D. & Burkhoff, D. Journal of Cardiac Failure 22, 303-311 (2016).

[0711] 56. Moscato, F„ Med Eng Phys 30, 1149-1158 (2008).

[0712] 57. Colacino, F. M., Medical Engineering & Physics 29, 829-839 (2007).

[0713] 58. Morley, D„ et al. The J of Thoracic and Cardiovascular Surgery 133, 21-28. e24 (2007).

[0714] 59. Guyton, A. C. Physiological Reviews 35, 123-129 (1955).

[0715] 60. Baillargeon, B, et al., European Journal of Mechanics-A / Solids 48, 38-47 (2014).

[0716] 79

[0717] 45802538.1 61. Levine, S., et al. Dassault systemes’ living heart project, in Modelling Congenital Heart Disease: Engineering a Patient-specific Therapy 245-259 (Springer, 2022).

[0718] 62. Holzapfel, G. A. Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences 367, 3445-3475 (2009).

[0719] 80

[0720] 45802538.1

Claims

CLAIMSWe claim:

1. A left atrium simulator for a patient comprising:a left atrium structure formed of a first polymer; andone or more soft robotic actuators,wherein the left atrium structure comprises a receiving portion for attaching to a left atrial appendage attachment,wherein each of the one or more soft robotic actuators is attached to and conforms to a region of the left atrium structure.

2. The left atrium simulator of claim 1, further comprising the left atrial appendage attachment formed of a second polymer, wherein the left atrial appendage attachment is attached to the receiving portion.

3. The left atrium simulator of claim 1 or 2, comprising two or more soft robotic actuators, wherein a first soft robotic actuator is attached to and conforms to the anterior region of the left atrium structure, and wherein a second soft robotic actuator is attached to and conforms to the posterior-inferior region of the left atrium structure.

4. The left atrium simulator of claim 3, comprising a third soft robotic actuator that is attached to and conforms to the circumference of the receiving portion for attaching the left atrial appendage attachment thereto.

5. The left atrium simulator of claim 3 or 4, comprising a fourth soft robotic actuator that is attached to and conforms to the circumference of the roof-anterior region of the left atrium structure.

6. The left atrium simulator of claim 1 or 2, comprising four soft robotic actuators, wherein:a first soft robotic actuator is attached to and conforms to the anterior region of the left atrium structure,a second soft robotic actuator is attached to and conforms to the posterior-inferior region of the left atrium structure,a third soft robotic actuator that is attached to and conforms to the circumference of the receiving portion for attaching the left atrial appendage attachment thereto, anda fourth soft robotic actuator that is attached to and conforms to the circumference of the roof-anterior region of the left atrium structure.

7. The left atrium simulator of any one of claims 3-6, the first and second soft robotic actuators are attached to each other and optionally form a unitary piece that surrounds the circumference of the anterior and posterior-inferior regions.81458025388. The left atrium simulator of any one of claims 2-7, wherein the first and second polymers are the same and each of the first and second polymers is polysiloxane.

9. The left atrium left atrium simulator of any one of claims 1-8, wherein each of the one or more soft robotic actuators comprises a bladder, a tubing, and a mesh.

10. A kit for assembling a left atrium simulator for a patient, comprising:a left atrium structure formed of a first polymer;one or more left atrial appendage attachments; andone or more soft robotic actuators,wherein the left atrium structure comprises a receiving portion,wherein each of the one or more left atrial appendage attachments comprises an attachment end configured to be received by the receiving portion, andwherein each of the one or more left atrial appendage attachments has a different shape.

11. The kit of claim 10, wherein the one or more left atrial appendage attachments have a shape selected from the group consisting of cactus, chicken wing, windsock, and cauliflower.

12. The kit of claim 10 or 11, wherein each of the one or more soft robotic actuators has a dimension and shape configured to attach to a region of the left atrium structure.

13. A left heart simulator for a patient comprising:the left atrium simulator of any one of claims 1-9; anda left ventricle simulator,wherein the left ventricle simulator comprises:a left ventricle structure; andone or more soft robotic left ventricle actuators,wherein each of the one or more soft robotic actuators is attached to and conforms to a region of the left ventricle structure,wherein the left atrium simulator is attached to the left ventricle simulator.

14. The left heart simulator of claim 13, comprising a first soft robotic left ventricle actuator that is attached to and conforms to the circumference at the base of the left ventricle structure, a second soft robotic left ventricle actuator that is attached to and conforms to the circumference of the middle section of the left ventricle structure, and a third soft robotic left ventricle actuator that is attached to and conforms to the circumference of the apical section of the left ventricle structure.

15. The left heart simulator of claim 13 or 14, further comprising a fourth, a fifth, and a sixth soft robotic left ventricle actuator, wherein each of the fourth, fifth, and sixth soft robotic left ventricle actuators is attached to and conforms to the surface of the left ventricle structure along a helical axis at about a 60-degree angle from the basal plane of the left ventricle structure.8245802538.

116. The left heart simulator of any one of claims 13-15, wherein the left ventricle structure is formed of a third polymer.

17. A system comprising:the left heart simulator of any one of claims 13-16; anda mock circulatory flow loop,wherein the left heart simulator is in fluid connection with the mock circulatory flow loop.

18. The system of claim 17, further comprising a mechanical mitral valve (MV) at the interface of the left atrium simulator and the left ventricle simulator; and a mechanical aortic valve (AoV) at the interface of the left ventricle simulator and a conduit of the mock circulatory flow loop.

19. The system of claim 17 or 18, wherein the mock circulatory flow loop comprises one or more pressure sensors and / or one or more flow sensors.

20. The system of any one of claims 17-19, wherein the mock circulatory flow loop comprises an electro-pneumatic control unit.

21. A method of using the left atrium simulator of any one of claims 1-9 for testing a medical device, comprising:(i) implanting the medical device at a desired location in the left atrium structure;(ii) optionally selecting a left atrial appendage attachment having desired shape and attaching the selected left atrial appendage attachment to the receiving portion of the left atrium structure;(iii) connecting the left atrium simulator to a mock circulatory flow loop to form a system; and(iv) running the system.

22. A method of using the left heart simulator of any one of claims 13-16 for testing a medical device, comprising:(i) implanting the medical device at a desired location in the left atrium structure and / or the left ventricle structure;(ii) optionally selecting a left atrial appendage attachment having desired shape and attaching the selected left atrial appendage attachment to the receiving portion of the left atrium structure;(iii) connecting the left heart simulator to a mock circulatory flow loop to form a system; and(iv) running the system.

23. A method of using the system of any one of claims 17-20 for testing a medical device, comprising:8345802538.1(i) implanting the medical device at a desired location in the left atrium structure and / or the left ventricle structure;(ii) optionally selecting a left atrial appendage attachment having desired shape and attaching the selected left atrial appendage attachment to the receiving portion of the left atrium structure; and(iii) running the system.

24. A method of using the left atrium simulator of any one of claims 1-9 for pre-procedure testing of a medical device for a patient, comprising:(i) imaging the patient’s heart, particularly the patient’s left atrial appendage;(ii) selecting a left atrial appendage attachment having a shape that generally corresponds with the shape of the image of the patient’s left atrial appendage;(iii) implanting the medical device at a desired location in the left atrium structure;(iv) attaching the selected left atrial appendage attachment to the receiving portion of the left atrium structure;(v) connecting the left atrium simulator to a mock circulatory flow loop to form a system; and(vi) running the system.

25. A method of using the left heart simulator of any one of claims 13-16 for pre-procedure testing of a medical device for a patient, comprising:(i) imaging the patient’s heart, particularly the patient’s left atrial appendage;(ii) selecting a left atrial appendage attachment having a shape that generally corresponds with the shape of the image of the patient’s left atrial appendage;(iii) implanting the medical device at a desired location in the left atrium structure and / or the left ventricle;(iv) attaching the selected left atrial appendage attachment to the receiving portion of the left atrium structure;(v) connecting the left heart simulator to a mock circulatory flow loop to form a system; and (vi) running the system.

26. A method of using the system of any one of claims 17-20 for pre-procedure testing of a medical device for a patient, comprising:(i) imaging the patient’s heart, particularly the patient’s left atrial appendage;(ii) selecting a left atrial appendage attachment having a shape that generally corresponds with the shape of the image of the patient’s left atrial appendage;8445802538.1(iii) implanting the medical device at a desired location in the left atrium structure and / or the left ventricle;(iv) attaching the selected left atrial appendage attachment to the receiving portion of the left atrium structure; and(v) running the system.

27. The method of any one of claims 21-26, wherein the system running step comprises (a) starting flow of a fluid through the system; (b) tuning one or more operation parameters to match desired physiological conditions; and (c) measuring one or more system parameters.

28. The method of claim 27, wherein the one or more operation parameters comprise flow pressure and / or flow resistance.

29. The method of claim 27 or 28, wherein the one or more system parameters comprise fluid flow, pressure contraction, position of the medical device, and / or leakage around the medical device.

30. The method of any one of claims 24-29, wherein the procedure is atrial ablation, left atrial appendage occlusion, left atrial appendage ligation, or mitral valve procedure.8545802538.1

Images