We are excited to announce the release of the first publication from the TARGET project. The paper, titled "AI-based Derivation of Atrial Fibrillation Phenotypes in the General and Critical Care Populations," presents a new AI-based methodology that utilizes generative topographic mapping (GTM) to uncover and characterize distinct clinically relevant phenotypes of atrial fibrillation (AF) across diverse patient populations.

This milestone was achieved through the dedicated work of the project's coordinator, Sandra Ortega-Martorell, and researchers Ryan A. A. Bellfield, Ivan Olier, Robyn Lotto, Ian Jones, and Ellen A. Dawson from Liverpool John Moores University, alongside Gregory Y. H. Lip from University of Liverpool and Anil M. Tuladhar from Radboud University Medical Centre.

About the Paper

AF is the most common heart arrhythmia globally, associated with increased mortality and morbidity risks. Traditional clinical risk scores have limitations in predicting AF and its complications due to the complexity and heterogeneity of AF patients. This study, made possible through funding from TARGET and the DECIPHER project (LJMU QR-PSF), proposes an innovative AI-based methodology aimed at addressing these challenges by classifying patients into distinct, clinically meaningful phenotypes.

Using generative topographic mapping, a probabilistic machine learning method, and Ward’s minimum variance method, the research successfully identified distinct clinical phenotypes of AF in both general and critical care populations, drawn from two large databases: UK-Biobank and MIMIC-IV. This approach can enhance the accuracy of patient stratification, making it possible to tailor prevention and treatment strategies more effectively for each phenotype.

Read the full paper at the following link to learn more about the methodology used and the final results!