University of Cambridge > > Isaac Newton Institute Seminar Series > A Generative Shape-Compositional Framework to Synthesise Populations of Virtual Heart Chimaeras

A Generative Shape-Compositional Framework to Synthesise Populations of Virtual Heart Chimaeras

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

In the realm of in-silico trials for medical devices, the creation of accurate and variable virtual organ populations is a cornerstone for success. Traditional methods often grapple with the challenge of incomplete or inconsistent anatomical data across individuals, stemming from varied imaging modalities and clinical pathways. Addressing this gap, we introduce a groundbreaking approach that leverages a generative model capable of synthesizing complete multipart anatomical structures, termed “Virtual Chimeras.” This innovative framework is designed to learn from unpaired datasets, where each substructure originates from different subjects with missing or partially overlapping anatomical information. At the heart of our method lies a dual-component framework that combines a part-aware generative shape model and a spatial composition network. The former captures the inherent variability of each organ structure within the population, while the latter adeptly assembles these structures into coherent multipart entities. A novel aspect of our approach is the employment of a self-supervised learning scheme, enabling the spatial composition network to effectively train on partially overlapping data with weak labels. Our validation on cardiac structure shapes derived from UK Biobank’s magnetic resonance images showcases the superior performance of our method over traditional PCA -based shape models. The Virtual Chimeras not only exhibit enhanced plausibility and variability but also demonstrate significant advancements in generalizability and specificity. This work paves the way for more accurate and reliable in-silico trials, offering a powerful tool for the development and testing of new medical devices and therapies. Join us as we delve into the intricacies of Virtual Chimeras, a pivotal step forward in the digital transformation of medical research. Co-authors: This is joint work with Haoran Dou, Seppo Virtanen, Nishant Ravikumar and Alejandro F. Frangi. The work reported in this talk is accepted and will appear in IEEE Transactions on Neural Networks and Learning Systems.

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