In addition to the diagnosis of AF in a 12-lead resting electrocardiogram (ECG), signal processing approaches on ECGs (Fig. 1) are increasingly used to characterize atrial conduction properties, which have been shown to have some predictive value for ablation outcome and AF recurrences after cardioversion. Artificial intelligence-enabled ECG algorithms can detect AF with a high accuracy using 12-lead or single-lead ECGs and could even identify the presence of AF from a single ECG recorded during sinus rhythm. Although most of these artificial intelligence models have not yet reached clinical routine, they could have an important implication in AF screening [8].

Electrocardiographic imaging (ECGi) is a new technique to study conduction properties by non-invasively estimating the conduction on the heart surface from a body-surface potential map and the patient’s heart and torso anatomy [9]. ECGi has mostly been used to study ventricular arrhythmias, but is increasingly also studied in AF.

During invasive clinical electrophysiological ablation studies, the current standard mapping approach involves three-dimensional (3D) electroanatomic voltage mapping to identify areas of low voltage or regions with increased electrogram fractionation [7]. Low-voltage areas (typically < 0.5 mV during sinus rhythm) in the atrium have been associated with endocardial scar and/or structural defects, although the threshold can vary with rhythm change. Several multielectrode mapping catheters are currently in use for detailed high density electroanatomic mapping. These catheters have 16–64 electrodes and different interelectrode spacing.

Advances in late gadolinium-enhanced magnetic resonance imaging (LGE-MRI) allow for noninvasive assessment of structural alterations in atrial cardiomyopathy. In LGE-MRI, a gadolinium-based contrast agent accumulates in areas of increased extracellular space. However, it is currently not possible to differentiate between LGE areas caused by replacement fibrosis, necrosis, inflammation, edema, or interference with adjacent hyperenhanced structures, such as epicardial fat. Additionally, different LGE-MRI postprocessing methods lead to discrepancies regarding extent and regional distribution of atrial cardiomyopathy, as well as the correlation to low-voltage assessed in endocardial voltage mapping [10]. The contrasting results may be related to technical challenges with LGE-MRI, including spatial resolution, motion artifact, irregular heart rates, and the quantitation of LGE, which can be algorithm-dependent. In a recent randomized controlled trial, LGE-MRI-guided fibrosis ablation plus pulmonary vein isolation, compared with pulmonary vein isolation catheter ablation only, resulted in no significant difference in atrial arrhythmia recurrence [11]. In addition to the assessment of atrial remodeling, the evaluation of ablation scar has also been approached by LGE-MRI (Fig. 2). A better integration of LGE-MRI information into the AF ablation pathway (e.g., interventional cardiac MRI) may result in better outcomes in the future [12].

In addition to fibrosis and ablation scar, epicardial adipose tissue, which has been shown to contribute to atrial remodeling processes and AF progression by its paracrine effects, leading to increased fibrosis or direct fatty infiltration in the contiguous atrial myocardium, can also be quantified by computed tomography (Fig. 3a, b) or MRI [13]. Epicardial adipose tissue volume has been associated with increased risk of developing AF, AF persistence, and postablation recurrences independent of other measures of adiposity [14]. Interestingly, electroanatomic mapping has confirmed more pronounced changes with larger low-voltage areas in the posterior and inferior left atrium that are adjacent to the posteriorly located increased epicardial adipose tissue seen on MRI [14]. These findings may support a targeted ablation strategy, such as posterior left atrial isolation, and the early initiation of lifestyle and pharmacological interventions to reduce bodyweight.

In addition to structural alternations, functional impairment of atrial deformation properties also represents an important component of the progressive atrial remodelling and AF substrate. During ventricular systole, left atrial strain derived by speckle-tracking echocardiography reflects left atrial expansibility and stiffness (Fig. 3c). In patients with paroxysmal AF, left atrial strain is related to individual and combined comorbidities [15].

Indications for atrial cardiomyopathy could potentially be detected in community settings based on quantification of circulating cardiovascular biomolecules (Table 1) or digital biomarkers derived from wearable devices (e.g., photoplethysmography, ECG, and heart sound signals [16]). Genome-wide association studies have also identified many common genetic AF susceptibility variants throughout the genome. Polygenic risk scores derived from these can be used to successfully predict a person’s risk of developing AF [17].