BOATMAP

Paper: https://www.sciencedirect.com/science/article/pii/S0010482524012861

@article{meisenzahl_2024,
title = {BOATMAP: Bayesian Optimization Active Targeting for Monomorphic Arrhythmia Pace-mapping},
journal = {Computers in Biology and Medicine},
volume = {182},
pages = {109201},
year = {2024},
issn = {0010-4825},
doi = {https://doi.org/10.1016/j.compbiomed.2024.109201},
author = {Casey Meisenzahl and Karli Gillette and Anton J. Prassl and Gernot Plank and John L. Sapp and Linwei Wang},
}

Component overview

bayes3d.py

Uses a GP to predict the expected similarity between the target ecg and a prospective ecg in a different location on the heart. Has an acquisition function to suggset where to collect the next ecg from.

guide.py

Iteratively fits our surrogate model in bayes3d.py and follows the suggested locations to collect ecgs from. The process stops when an ecg collected reaches a similarity threshold that is sufficient.

similarity_metrics.py

Models the similarity between a target ecg and a newly collected ecg.

termination_criteria.py

Predicate to determine if the collected ecg is the predicted location of the target ecg.

util.py

Extracts the coordinates in the provided heart geometry that are viable for pace mapping.

Install

python -m venv ./venv
source ./venv/bin/activate
pip install -r requirements.txt

My python version is 3.9.6

Run

python exmaple.py

Usage

There are two functions that need to be defined in example.py.

1. A function to fetch the heart geometry. A dense tetrahedron volume mesh, where each row in torso_pts is a xyz locaion of a vertex, torso_uvc is the uvc coordinates assocated with each vertex, and each row in biv_elem describes a tetrahedron which holds indices into the torso_pts and the last colum describes the tissue status (scar, grey zone, healthy), also biv_elem is only defined on the left and right ventricle.

def heart_geometry():
    XYZ_COL = ['x', 'y', 'z']
    UVC_COL = ['apicobasal', 'transmural', 'rotational', 'transventricular']
    TETRA_COL = ['a', 'b', 'c', 'd']
    torso_pts = pd.read_csv(DATA_PATH / "geometry/torso.1000um.retag.pts", skiprows=1, names=XYZ_COL, sep=" ")/1000
    torso_uvc = pd.read_csv(DATA_PATH / "geometry/torso.1000um.retag.hpts", skiprows=1, names=UVC_COL, sep=" ")
    biv_elem = pd.read_csv(DATA_PATH / "geometry/biv.elem", skiprows=1, names=TETRA_COL+['tag'], sep=" ")
    return torso_pts, torso_uvc, biv_elem

2. A function to simulate an ecg at an index in the torso_pts file. The current code assumes the ecgs are precomputed and reads them from the file.

def simulate_ecg_fn(index: int):
    p = DATA_PATH / f"ecg/{index:09}.json"
    j = json.loads(p.read_text())
    # provided ECG must be in this format
    validate(j, schema=schema.ecg)
    return j