Success rates of simulated multi-pulse defibrillation protocols are sensitive to application timing with individual, protocol-specific optimal timings

Figures

• 

  FIGURE 1. The contractions of the heart coincide with the excitations of its muscle cells, which are visible in their transmembrane voltages (image colours). Excited cells propagate contraction signals to any neighbours sufficiently close to rest, sustaining the fibrillation present in this simulated Fenton-Karma model of a thin sheet of cardiac tissue. The resulting fibrillation patterns are chaotic and change shape rapidly, which is also evident in the model’s dominant time-scale of only ∼ 130 ms (or ∼ 7.7 Hz).

• 

  FIGURE 2. We estimate the success rate of a given defibrillation protocol is in three stages: We define a current profile to represent the average current induced by the electric field of its electric field (left), apply this profile to a fibrillation episode at randomly generated induction sites of varying conductivity (centre), observe the outcome, and then repeat this process over multiple induction-sites distributions as well as fibrillation episodes (right) to count the ratio of successful defibrillations among them.

• 

  FIGURE 3. Each defibrillation protocol is represented by the average current it induces in fibrillating cardiac tissue over time. Our three general protocol types of choice each feature pulses of uniform strength and length, but differ in the lengths of their inter-pulse periods (red): Widening for ADP, uniform for LEAP.

• 

  FIGURE 4. We measure the timing sensitivity of a given protocol type and pulse count in two stages (rows): First (top), we establish five representatives of known success rates when applied to 100 fibrillation episodes without any timing, ensured through appropriate uniform pulse strengths. Then (bottom), we apply these representatives to copies of the same episode at various timings and record their new success rates.

• 

  FIGURE 5. Each protocol type (rows and cols., bold names) and pulse count (cols., pairs) is represented by five protocols of known success rates (y-axes, coloured dots) when applied to 100 fibrillation episodes without any timing considerations. We ensure this by selecting appropriate values for their uniform pulse strengths (x-axes, coloured dots), which we determine by sampling each protocol type and pulse count over pulse strengths ranging from 0 to 5 a.u. (x-axes) and recording the corresponding averaged success rates (y-axes) shown as dose-response curves (grey). The non-monotonic LEAP 1.0 and 1.5 dose-response curves shown here are in line with previous findings on dose-response curves of multi-pulse protocols, where it was found that their curves may assume more complex shapes (Lilienkamp et al., 2022a).

• 

  FIGURE 6. All protocols show timing sensitivity in their defibrillation success rates (y-axes), measured by applying each of them to copies of the same fibrillation episode at various timings (x-axes). This holds across protocol types (cols., bold names), pulse counts (cols., pairs), and pulse strengths (rows). We choose the pulse strengths such that a given protocol reaches a specific target success-rate (rows, colours and horizontal lines) when applied to 100 different episodes without any timing considerations.

• 

  FIGURE 7. Of all protocol types/pulse counts (matrix rows and cols.), only the five- and ten-pulse LEAP 0.5 protocols (matrices, row 5 and col. 4) allow for the mutual inference (bold, green circles) of optimal defibrillation timings across all target success-rates (matrices). For every other potential pair, there is at least one target success-rate (matrix) showing an insufficient timing-correlation ( < 0.5, thin circles) between them. We measure these correlations through Spearman’s rank, as it makes no assumption beyond monotonicity (e.g., linearity) about the relationship between the timing data.