Cardiac Digital Twins at Scale from MRI: Open Tools and Representative Models from ~55000 UK Biobank Participants

TL;DR

An automatic open-source pipeline for creating patient-specific left and right ventricular meshes from cardiovascular magnetic resonance images is described, its application to a large cohort of ~55k participants from UK Biobank, and the construction of the most comprehensive cohort of adult heart models to date is described.

Abstract

A cardiac digital twin is a virtual replica of a patient's heart for screening, diagnosis, prognosis, risk assessment, and treatment planning of cardiovascular diseases. This requires an anatomically accurate patient-specific 3D structural representation of the heart, suitable for electro-mechanical simulations or study of disease mechanisms. However, generation of cardiac digital twins at scale is demanding and there are no public repositories of models across demographic groups. We describe an automatic open-source pipeline for creating patient-specific left and right ventricular meshes from cardiovascular magnetic resonance images, its application to a large cohort of ~55000 participants from UK Biobank, and the construction of the most comprehensive cohort of adult heart models to date, comprising 1423 representative meshes across sex (male, female), body mass index (range: 16 - 42 kg/m$^2$) and age (range: 49 - 80 years). Our code is available at https://github.com/cdttk/biv-volumetric-meshing/tree/plos2025 , and pre-trained networks, representative volumetric meshes with fibers and UVCs will be made available soon.

Figures (6)

• Figure 1: Overview. An overview of the proposed pipeline. Reproduced by kind permission of UK Biobank ©.
• Figure 2: Picking the ES frame. The LV volume transients (LV voxel count on each time frame divided by the maximum LV voxel count across time frames) of six example participants for all the views separately, and their sum across the views. For the SAX view, only the five mid-slices were used for the calculation. The ES frame was picked as the time frame that minimizes the sum across all views.
• Figure 3: Contour and landmark extraction from segmentations. Illustration of contours and landmarks automatically extracted from the ED frame of a randomly selected participant. Contours: Red: LV endocardium, Green: LV epicardium, Brown: RV septum, Blue: RV free wall. Landmarks: Yellow: Points on the mitral valve, Pink: Points on the tricuspid valve, Teal: Points on the aorta-LV intersection, Purple: Apex. Reproduced by kind permission of UK Biobank ©.
• Figure 4: Demographic bins and associations with derived phenotypes. (A) The number of female and male participants in this cohort in specific age and BMI bins. The bins that contain more than three participants are marked with a * and a representative mesh is created for each of these bins. The colour bar represents the absolute value of the number of participants. (B),(C) Associations of derived phenotypes of all representative hearts with sex, age and BMI. The derived phenotypes LVEDV, RVEDV and LV mass are plotted as kernel density plots along with linear-regression lines for the whole cohort (black), for female (blue) and for male (red) (n=1423).
• Figure 5: Examples of representative meshes for different demographic bins.