Probing the Limits and Capabilities of Diffusion Models for the Anatomic Editing of Digital Twins

Karim Kadry,Shreya Gupta,Farhad R. Nezami,Elazer R. Edelman

2023-12-30

Abstract:Numerical simulations can model the physical processes that govern cardiovascular device deployment. When such simulations incorporate digital twins; computational models of patient-specific anatomy, they can expedite and de-risk the device design process. Nonetheless, the exclusive use of patient-specific data constrains the anatomic variability which can be precisely or fully explored. In this study, we investigate the capacity of Latent Diffusion Models (LDMs) to edit digital twins to create anatomic variants, which we term digital siblings. Digital twins and their corresponding siblings can serve as the basis for comparative simulations, enabling the study of how subtle anatomic variations impact the simulated deployment of cardiovascular devices, as well as the augmentation of virtual cohorts for device assessment. However, while diffusion models have been characterized in their ability to edit natural images, their capacity to anatomically edit digital twins has yet to be studied. Using a case example centered on 3D digital twins of cardiac anatomy, we implement various methods for generating digital siblings and characterize them through morphological and topological analyses. We specifically edit digital twins to introduce anatomic variation at different spatial scales and within localized regions, demonstrating the existence of bias towards common anatomic features. We further show that such anatomic bias can be leveraged for virtual cohort augmentation through selective editing, partially alleviating issues related to dataset imbalance and lack of diversity. Our experimental framework thus delineates the limits and capabilities of using latent diffusion models in synthesizing anatomic variation for in silico trials.

Computer Vision and Pattern Recognition,Image and Video Processing

What problem does this paper attempt to address?

This paper investigates the use of Latent Diffusion Models (LDMs) to anatomically edit Digital Twins of cardiovascular device deployments to create "Digital Siblings" with different anatomical features. Digital Twins are computer models based on patient-specific anatomical structures, used to accelerate and reduce the risk of medical device design. However, the use of patient-specific data alone limits the anatomical variability that can be explored.

In this study, the authors implemented a method to generate Digital Siblings through a three-dimensional digital twin case of the heart anatomy and characterized them through morphological and topological analysis. They edited the digital twins to introduce variability at different spatial scales and local regions, and found biases towards common anatomical features. These biases can be mitigated by selective editing to enhance the diversity of the virtual cohort, alleviating issues of data imbalance and lack of diversity.

The main contributions of this paper include:
1. Development of a latent diffusion model to generate myocardial label mappings and proposing an experimental framework to study how generation and editing techniques produce scale and region-specific digital twin variants.
2. Demonstrating that the latent diffusion model can introduce topological errors during generation and editing, with the number of such errors influenced by the editing method and seed characteristics.
3. Identifying biases in the generation process due to imbalanced datasets, which also affect scale and region-specific editing.
4. Showing how this editing method can enhance the diversity of the virtual cohort by exploring low-frequency anatomical distributions within the training set boundaries through scale-specific variations and adding morphology beyond the anatomical distribution through region-specific variations.

The study also covers existing work related to virtual anatomy generation, such as principal component analysis, variational autoencoders, generative adversarial networks, and applications of diffusion models in human anatomy, as well as methods for evaluating the quality of generated anatomy. The paper provides a new perspective by experimentally analyzing model performance, editing methods, and the morphological and topological characteristics of the generated anatomy, offering insights into the use of latent diffusion models in in vitro experiments.