Cardiac MRI Based Left Atrial Radiomics for Assessing Persistent Atrial Fibrillation and Catheter Ablation Outcomes

30 Pages Posted: 4 Aug 2022

See all articles by Soojung Park

Soojung Park

Korea University - Korea University Anam Hospital

Yongwon Cho

Korea University - Department of Radiology

Guan Yang

Imperial College London - National Heart and Lung Institute

kyungsang Kim

Harvard University - Massachusetts General Hospital

Yu-Whan Oh

Korea University - Korea University Anam Hospital

Yun Gi Kim

Korea University - Korea University Anam Hospital

Jaemin Shim

Korea University - Division of Cardiology

Jong-Il Choi

Korea University - Department of Internal Medicine

Young-Hoon Kim

Korea University Medical Center

Sung Ho Hwang

Korea University - Korea University Anam Hospital

Abstract

Background: Cardiac MRI can potentially evaluate the left atrium associated with atrial fibrillation (AF). Catheter ablation of persistent AF (PeAF) has a risk of recurrent AF.

Objectives: To investigate cardiac MRI-based left atrial radiomics (LAR) for PeAF and recurrent AF after catheter ablation.

Methods: This retrospective study analyzed late gadolinium enhancement (LGE) MRI from 331 consecutive patients who underwent cardiac MRI and catheter ablation of AF sequentially from January 2018 to December 2019. For training, 215 patients (mean age, 59 {plus minus} 10 years; 154 men) were evaluated. The LAR model based on the radiomics features of LGE-MRI was designed to classify AF types.

Results: 116 patients (mean age, 59 {plus minus} 9 years; 79 men; 51 patients with PeAF) were evaluated for validation. For diagnosing PeAF, the LAR model showed an area under the curve of 0.82 (95% confidence interval [CI]: 0.74 - 0.89), a sensitivity of 69%, a specificity of 82%, a positive predictive value of 74%, a negative predictive value of 77%, and an accuracy of 75%. After catheter ablation, 56 (48%) of 116 patients experienced recurrent AF with an average follow-up of 17.8 {plus minus} 5.4 months (range: 3 to 24 months). Cox regression analysis revealed that PeAF determined with the LAR model was independently associated with recurrent AF (hazard ratio: 2.9; 95% CI: 1.72 - 5.13; P < 0.001).

Conclusions: The LAR model of LGE-MRI identified the left atrium related to PeAF. The LAR model classifying AF types predicted recurrent AF in patients who have undergone catheter ablation of AF.

Keywords: atrial fibrillation, magnetic resonance imaging, left atrium, catheter ablation, deep learning

Suggested Citation

Park, Soojung and Cho, Yongwon and Yang, Guan and Kim, kyungsang and Oh, Yu-Whan and Kim, Yun Gi and Shim, Jaemin and Choi, Jong-Il and Kim, Young-Hoon and Hwang, Sung Ho, Cardiac MRI Based Left Atrial Radiomics for Assessing Persistent Atrial Fibrillation and Catheter Ablation Outcomes. Available at SSRN: https://ssrn.com/abstract=4181158 or http://dx.doi.org/10.2139/ssrn.4181158

Soojung Park

Korea University - Korea University Anam Hospital ( email )

1 Anam-dong 5 ka
Seoul, 136-701
Korea, Republic of (South Korea)

Yongwon Cho

Korea University - Department of Radiology ( email )

Guan Yang

Imperial College London - National Heart and Lung Institute ( email )

Kyungsang Kim

Harvard University - Massachusetts General Hospital ( email )

Yu-Whan Oh

Korea University - Korea University Anam Hospital ( email )

1 Anam-dong 5 ka
Seoul, 136-701
Korea, Republic of (South Korea)

Yun Gi Kim

Korea University - Korea University Anam Hospital ( email )

1 Anam-dong 5 ka
Seoul, 136-701
Korea, Republic of (South Korea)

Jaemin Shim

Korea University - Division of Cardiology ( email )

Seoul
Korea, Republic of (South Korea)

Jong-Il Choi

Korea University - Department of Internal Medicine ( email )

Young-Hoon Kim

Korea University Medical Center ( email )

Korea, Republic of (South Korea)

Sung Ho Hwang (Contact Author)

Korea University - Korea University Anam Hospital ( email )

1 Anam-dong 5 ka
Seoul, 136-701
Korea, Republic of (South Korea)

Do you have negative results from your research you’d like to share?

Paper statistics

Downloads
31
Abstract Views
142
PlumX Metrics