Personalized Topology-Informed Localization of Standard 12-Lead ECG Electrode Placement from Incomplete Cardiac MRIs for Efficient Cardiac Digital Twins

TL;DR

The paper addresses the challenge of localizing standard 12-lead ECG electrodes from incomplete cardiac MRI data to enable efficient cardiac digital twins. It proposes a topology-informed model (TIM) that converts electrode localization into a keypoint detection problem and reconstructs 3D torso geometry via a surface skeleton guided by keypoints, leveraging incomplete MRI contours. On 200 UK Biobank subjects, TIM achieves electrode localization errors of $ED_{electrode}=1.24\pm0.293$ cm and faster inference (~2 s) than conventional SSM-based methods (30–35 min), while also enabling accurate in-silico ECG simulations with $Corr_{Pearson}=0.989\pm0.013$ for key leads. These results demonstrate the feasibility of automatic electrode detection from routine MRI to create robust, patient-specific CDTs, with potential for real-time clinical and research applications, albeit with limitations regarding electrode-placement variability and generalizability to broader datasets.

Abstract

Cardiac digital twins (CDTs) offer personalized in-silico cardiac representations for the inference of multi-scale properties tied to cardiac mechanisms. The creation of CDTs requires precise information about the electrode position on the torso, especially for the personalized electrocardiogram (ECG) calibration. However, current studies commonly rely on additional acquisition of torso imaging and manual/semi-automatic methods for ECG electrode localization. In this study, we propose a novel and efficient topology-informed model to fully automatically extract personalized ECG standard electrode locations from 2D clinically standard cardiac MRIs. Specifically, we obtain the sparse torso contours from the cardiac MRIs and then localize the standard electrodes of 12-lead ECG from the contours. Cardiac MRIs aim at imaging of the heart instead of the torso, leading to incomplete torso geometry within the imaging. To tackle the missing topology, we incorporate the electrodes as a subset of the keypoints, which can be explicitly aligned with the 3D torso topology. The experimental results demonstrate that the proposed model outperforms the time-consuming conventional model projection-based method in terms of accuracy (Euclidean distance: $1.24 \pm 0.293$ cm vs. $1.48 \pm 0.362$ cm) and efficiency ($2$~s vs. $30$-$35$~min). We further demonstrate the effectiveness of using the detected electrodes for in-silico ECG simulation, highlighting their potential for creating accurate and efficient CDT models. The code is available at https://github.com/lileitech/12lead_ECG_electrode_localizer.

Figures (11)

• Figure 1: Illustration of the setup of the torso with electrodes (labeled with cyan-blue dots) and simulated 12-lead ECG.
• Figure 2: Visualization of extracted full and refined contours from cardiac MRIs and the 3D arrangement of the refined torso contours (only partially visualized here). Here, the acquired cardiac MRIs include scout images covering a larger range of the torso, typically used as localizers, and heart-focused slices in various views, including long-axis (LAX) views (2-, 3-, and 4-chamber) and a stack of short-axis (SAX) views.
• Figure 3: Topology-informed model (TIM) for 12-lead ECG electrode detection and torso geometry reconstruction from sparse and incomplete torso contours. Here, the electrodes are labeled via red dots within the predicted keypoints, which are then converted into surface skeleton to guide the 3D torso reconstruction. Note that the diagram takes 32 keypoints as an example. PCN: point completion model.
• Figure 4: Statistical shape model (SSM) based 3D torso mesh reconstruction and electrode localization from multi-view 2D cardiac MRIs.
• Figure 5: The standard placement of torso electrodes for 12-lead ECG.