Relationships Between Information Theoretic Measures (Entropy) in Human, Animal Experimental, and Computer-Simulated Atrial Fibrillation

D Dharmaprani, A McGavigan, D Chapman, R Kutieleh, S Thanigaimani, K Pope, P Kuklik, A Ganesan

Research output: Contribution to conferenceAbstractpeer-review


Background: It is postulated that high entropy (En) correlates with rotors. Although many En algorithms are available, little exists on the relationships between them, and their correlation to dominant frequency (DF). We examine relationships between Shannon, approximate and sample entropy (ShEn/ApEn/SampEn) and DF in human (H), sheep (S), and computer-simulated (SIM) AF.

Methods: Sixty-four-electrode basket bi-atria sustained AF recordings (H: 5 min; S: 20 min) were acquired. Computer-simulated bipolar electrograms were generated using models of locally stable spiral waves (Luo–Rudy model), and complex AF scenarios of rotors surrounded by wave break-up (Tusscher–Panfilov model). ShEn, ApEn, SampEn, and DF were computed in Matlab. Relationships between measures were determined by (1) Pearson's correlation coefficient (CC) to examine linear relationships; and (2) mutual information (MI) to determine non–linear relationships.

Results: In humans and sheep, ShEn showed moderate correlation with ApEn/SampEn (H CC: 0.44 ± 0.04, MI –1.10 ± 0.02; S CC: 0.52 ± 0.07, MI 1.36 ± 0.23). In SIM of stable spirals with low meander, ApEn/SampEn did not well identify the rotor pivot, resulting in weak linear correlation to ShEn but moderate MI (CC 0.27 ± 0.12; MI 1.6 ± 0.03). Contrastingly, ShEn/ApEn/SampEn strongly correlated under greater rotor meandering. In complex AF SIM, En relationships broke down (moderately disorganised: CC 0.62 ± 2, MI 0.41 ± 0.01; highly disorganised: CC 0.37 ± 0.02, MI 0.05 ± 0.01). In all cases, DF weakly correlated with En.

Conclusion: Strong relationships exist between ApEn/SampEn/ShEn for stable rotors, but break down under complex wave propagation. Spiral-wave meander affects ApEn/SampEn's ability to identify rotor pivot. Weak linear/non–linear correlations between En and DF suggest measures observe different aspects of AF phenomena. These findings provide potentially useful insights for AF bipolar electrogram analysis.


Conference66th Cardiac Society of Australia and New Zealand Annual Scientific Meeting, the International Society for Heart Research Australasian Section Annual Scientific Meeting and the 12th Annual Australia and New Zealand Endovascular Therapies Meeting
Abbreviated titleCSANZ2018 ANZET18


  • atrial fibrillation
  • computer simulation
  • Entropy


Dive into the research topics of 'Relationships Between Information Theoretic Measures (Entropy) in Human, Animal Experimental, and Computer-Simulated Atrial Fibrillation'. Together they form a unique fingerprint.

Cite this