BACKGROUND: Machine learning is a promising approach to personalize atrial fibrillation management strategies for patients after catheter ablation. Prior atrial fibrillation ablation outcome prediction studies applied classical machine learning methods to hand-crafted clinical scores, and none have leveraged intracardiac electrograms or 12-lead surface electrocardiograms for outcome prediction. We hypothesized that (1) machine learning models trained on electrograms or ECG signals can perform better at predicting patient outcomes after atrial fibrillation ablation than existing clinical scores and (2) multimodal fusion of electrogram, ECG, and clinical features can further improve the prediction of patient outcomes. METHODS: Consecutive patients who underwent catheter ablation between 2015 and 2017 with panoramic left atrial electrogram before ablation and clinical follow-up for at least 1 year following ablation were included. Convolutional neural network and a novel multimodal fusion framework were developed for predicting 1-year atrial fibrillation recurrence after catheter ablation from electrogram, ECG signals, and clinical features. The models were trained and validated using 10-fold cross-validation on patient-level splits. RESULTS: One hundred fifty-six patients (64.5±10.5 years, 74% male, 42% paroxysmal) were analyzed. Using electrogram signals alone, the convolutional neural network achieved an area under the receiver operating characteristics curve of 0.731, outperforming the existing APPLE scores (area under the receiver operating characteristics curve=0.644) and CHA2DS2-VASc scores (area under the receiver operating characteristics curve=0.650). Similarly using 12-lead ECG alone, the convolutional neural network achieved an AUROC of 0.767. Combining electrogram, ECG, and clinical features, the fusion model achieved an AUROC of 0.859, outperforming single and dual modality models. CONCLUSIONS: Deep neural networks trained on electrogram or ECG signals improved the prediction of catheter ablation outcome compared with existing clinical scores, and fusion of electrogram, ECG, and clinical features further improved the prediction. This suggests the promise of using machine learning to help treatment planning for patients after catheter ablation.
AgraïmentsLa realització d'una tesi doctoral es per una part el treball científic més important en la carrera d'un jove investigador, però d'altra part també representa el proces vital que completa l'etapa d'aprenentage i dona pas a l'etapa adulta. Aquesta tesi ha suposat doncs no sols un proces de desenvolupament científic, sino també un desenvolupament personal modelat per les relacions humanes en aquest període involucrades. Per això, em cal agrair a totes aquelles persones que han fet posible el transcórrer d'aquest treball la seua participació.En primer lloc, he tingut el plaer de treballar amb persones de les que he aprés molt més del que els podré agrair: Edu, Conrad, Xavi, Santi, Jaume, Adolfo, Cristina, Ramón, Ismael, Toni, Álvaro, Paco, Javier, Lluis, Laia, Irene i molts més. També he d'agraïr a Pepe la seua aposta per un jove teleco que poc més que ganes de treballar podía oferir. Gràcies especials a Alejandro i a Jorge, el vostre treball ha sigut indispensable per a aquesta tesi, i heu sigut uns excelents companys tant en el treball com en l'oci. Espere que totes aquestes relacions pervisquen per molts anys.María i Andreu, rebeu el meu sincer agraïment, vostra ajuda i direcció ha sigut indispensable, i el vostre treball incansable. Sou responsables de que hui en dia el grup de investigació siga el que és, i a més heu conseguit convertir-lo en una gran família. També m'és indispensable agrair a Omer Berenfeld la ajuda que desde l'altra part del Atlantic m'ha proporcionat, així com a l'acolliment amb els braços oberts durant l'estada en el seu grup d'Ann Arbor. També he d'agrair de manera especial la ajuda del doctor Felipe Atienza en aquesta tesi.Per altra banda, i de manera infinitament més ampla, tinc que donar gràcies als meus pares, sense els quals no haguera pogut recorrer ni una fracció del camí que m'ha dut fins ací. A ma mare, de la que he heretat la que espere serà la meua professió en els pròxims passos d'aquest camí, la docència. I per extensió, a tots els mestres que m'han conduït fins ací i m'han permés vore el món des dels muscles d'un gegant.Per últim, a la persona més important de la meua vida, en la que compartisc aquest camí i fa que sols tinga sentit poder recorrer-lo amb ella. Aquest treball ha sigut gràcies a tu.iii
Electrocardiographic imaging (ECGI) can characterise cardiac pathologies such as atrial fibrillation (AF)through specific markers based on frequency or phase analysis. In this study, the effect of the geometry of patients' torso and atria in the ECGI resolution is studied.A realistic 3D atrial geometry was located on 30 patient torsos and ECGI signals were calculated for 30 different AF simulations in each torso. Dominant frequency (DF) and reentrant activity analysis were calculated for each scenario. Anatomical and geometrical measurements of each torso (30-80% of variability between patients) and atria were calculated and compared with the errors in the ECGI estimation versus the departing EGM maps.Results show evidences that big chest dimensions worsen the non-invasive calculation of AF markers (p<0.05). Also, higher number of visible electrodes from each atrial region improves ECGI characterization measured as lower DF deviations (0.64±0.26 Hz vs 0.72±0.27 Hz, p<0.05) and higher reentrant activity coincidence (10.1±12.2% vs 3.4±3.4%, p<0.05).Torso and atrial geometry affect the quality of the noninvasive reconstruction of AF markers such as DF or reentrant activity. Knowing the geometrical parameters that worsen non-invasive AF maps may help to measure each detected AF driver reliability.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
customersupport@researchsolutions.com
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.
Copyright © 2025 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.