University of Cambridge > > Isaac Newton Institute Seminar Series > From Clinic to Computation: Understanding Emergent Dynamics in AF Using Cardiac Models, AI and UQ

From Clinic to Computation: Understanding Emergent Dynamics in AF Using Cardiac Models, AI and UQ

Add to your list(s) Download to your calendar using vCal

If you have a question about this talk, please contact nobody.

FHTW02 - Fickle Heart: The intersection of UQ, AI and Digital Twins

Introduction: Cardiac digital twins present a multiscale physics and physiology constrained framework for studying atrial fibrillation (AF). RENEWAL -AF identified novel indices of AF tissue level dynamics that correlate with AF recurrence, termination, and persistence. However, it is unclear how tissue properties may impact these measures, which is typically also computationally intensive to explore. Methods: This pilot study addresses this using Gaussian Process Emulators (GPEs) to perform rapid parameter analysis. Simulations of AF (4×4cm 2D grids) using the Courtemanche and Luo-Rudy models were analysed (n=200 each), focusing on ionic conductance (Courtemance: GNa, GK1 , GCaL; Luo-Rudy: GNa, GK1 , GK1bar, GSi) and tissue conductivities (longitudinal, transverse). Parameters were sampled using a Latin hypercube design. Using virtual catheters simulating the HD-grid catheter, we measured: i) correlation length (ξ); ii) rate of spiral wave formation (λf); iii) rate of spiral destruction (λd) – indicators of atrial electrical desynchrony. Simulated ξ, λf and λd values were used to train GPEs implemented through GPErks. GPE performance was assessed using R2 and MSE ,and used for sensitivity analysis to quantify the importance of each model parameter. Results: No significant difference was found between mean simulated values calculated in-silico using the virtual catheter (ξ=27.15 (95%CI:22.69,31.61), λf=6.28 (95%CI:6.10,6.42)),  λf=3.81 (95%CI:3.38,4.24),  and mean clinical HD-grid measurements (ξ=34.63 (95%CI:27.78,41.49), λf =6.54 (95%CI:5.41,8.61)), λd=3.67 (95%CI:2.61,4.72), (all P>0.05). For the Courtemanche model, GPEs returned an R2=0.62 for ξ, compared to 0.66 for λf and 0.64 for λd. For the Luo-Rudy model, ξ R2=0.78, λf R2 = 0.69, and λd R2 =0.67. Sobol variance identified longitudinal tissue conductivity and GSi as most influential on ξ, which was also observed in addition to GK1 for λf and λd. Conclusion: Despite scope to optimise and improve GPE performance further, this pilot demonstrates GPEs’ potential to efficiently map bespoke tissue scale AF metrics to model parameters, potentially further supporting cardiac digital twins for clinical/research use.

This talk is part of the Isaac Newton Institute Seminar Series series.

Tell a friend about this talk:

This talk is included in these lists:

Note that ex-directory lists are not shown.


© 2006-2024, University of Cambridge. Contact Us | Help and Documentation | Privacy and Publicity