# Generating patient-specific computational models with point cloud data from human atrial electrophysiology studies

## TL;DR

Researchers created patient-specific heart models using point cloud data to simulate and study atrial fibrillation, showing promising results for future medical applications.

## Contribution

A novel pipeline for generating patient-specific atrial models using point cloud data without image segmentation is introduced.

## Key findings

- Simulated electrograms matched measured ones in local activation time and cross-correlation peak for some patients.
- Increased fibrosis density in models led to more multiphasic electrogram morphologies.
- The method produced models with morphological features consistent with other fibrosis electrophysiology studies.

## Abstract

One in five patients diagnosed with atrial fibrillation die within a year after being diagnosed according to a prospective cohort study done in the United Kingdom, making this disease the focus of many scientific studies. One approach to studying this disease has been computational models since they have demonstrated powerful capabilities in understanding and analyzing the biochemical processes underlying atrial fibrillation. To create accurate patient models for studying cardiovascular diseases, we developed a pipeline for generating patient-specific models of the left atrial posterior wall using point cloud data without image segmentation. Our goal was to evaluate the performance of these models by comparing simulated electrograms to those obtained from atrial fibrillation patients. We created models for two different paroxysmal atrial fibrillation patients under healthy tissue conditions. To validate our model, we compared simulated and measured electrograms using various metrics. Some electrograms matched well in terms of local activation time and cross-correlation peak, whereas others showed significant differences in amplitude and duration. Additionally, we explored the impact of modeling fibrotic tissue on electrogram morphology by creating four persistent atrial fibrillation patient models with varying fibrosis densities and types. Simulations indicated that increased modeled fibrosis density led to more multiphasic electrogram morphologies, with little impact from fibrosis type. The fibrosis simulations also had morphological characteristics seen in other fibrosis electrophysiology modeling studies like deflections patterns and amplitudes, strengthening the reasoning behind using this type of model generation methodology. Our findings suggest that point cloud data is sufficient for creating accurate left atrial posterior wall models, which can simulate electrograms comparable to measured waveforms. This method could be useful for patient-specific studies, potential specialized ablation procedures, and arrhythmia research.