This study was accomplished using a constitutive transgenic mouse model, α-MHC-ChR2, at the age of 17 weeks and older, whereby ChR2 expression was proven by biomolecular protocols and restricted exclusively to cardiac tissue. In the experiment, a retrograde ex vivo perfusion after Langendorff was used. For this method, as described elsewhere (Richter et al., 2016), the murine heart was explanted, afterwards cannulated via the aorta and perfused with Tyrode's solution (130 mM NaCl, 4 mM KCl, 1 mM MgCl₂, 24 mM NaHCO₃, 1.8 mM CaCl₂, 1.2 mM KH2PO₄, 5.6 mM glucose, and 1% albumin/BSA; aerated with carbogen (5% CO₂ and 95% O₂)). A constant flow rate of (2.63 ± 0.58) ml/min was applied, whereby normal heart rate varied from 4 to 6 Hz and arrhythmic frequencies were observed between 15 and 30 Hz.

For experimental induction of arrhythmic patterns, normal perfusion was switched to low-K+ Tyrode's solution (130 mM NaCl, 2 mM KCl, 1 mM MgCl2, 24 mM NaHCO3, 1.8 mM CaCl2, 1.2 mM KH2PO4, 5.6 mM glucose, and 1 % albumin/BSA were aerated with carbogen (95 % oxygen and 5 % CO 2)) with 100 μM Pinacidil as described by Bruegmann et al. (2016). In contrast to other publications, Richter et al. (2016), Quiñonez Uribe et al. (2018), and Sasse et al. (2019) the initiation of arrhythmia in this work was induced with rapid optical pacing as also shown in Diaz-Maue et al. (2021). Sustained arrhythmias were achieved when remaining for t_term = 5 s, control experiments which confirm this criterion are further detailed in section 3.2. Should the induced arrhythmia terminate within t_term, it was classified as self-terminated. In addition, if an arrhythmia failed to be terminated at the fifth photostimulation attempt, then backup was used. The backup consisted of activating all μLED arrays at the same time at the highest experimentally used light intensity resulting in global illumination.

A custom-built Multi-Lead ECG device was designed to simultaneously record cardiac activity from different regions of the heart. Based on the Wilson Hypothesis by Wilson et al. (1934, 1946), a Wilson Central Terminal (WCT) was built by connecting three surface Ag/AgCl electrodes V1, V2, and V3 to a common point. The potential difference between these three electrodes and the WCT are referred to in this work as uni-polar lead measurements. During the experiments, the three electrodes were positioned at an angle of 120° as shown in Figures 1A,B. The signals acquired were digitally converted and afterward analyzed.

Motivated by the work of Schwaerzle et al. (2014) and Diaz-Maue et al. (2018, 2021), a set of custom-built micro-LED (μLED) arrays were designed. Briefly, blue LEDs (XZBBR155W5MAV, SunLED, USA) with a central wavelength λ_blue = 470 nm were reflow soldered onto a printed circuit board (PCB). The material used for the PCB was a prepreg laminate (104 ML, LPKF, Garbsen, Germany) with a thickness of 200 μm. The blue LEDs were arranged in a 3 × 3 array with a horizontal distance of 1.14 mm as shown in Figure 1D, yielding to an active illumination area of 4.76 mm². The conductive tracks on the PCB were isolated embedding the complete array into a layer of Polydimethylsiloxane (PDMS, Sylgard 184, Dow Corning, USA). PDMS was chosen because of its excellent biocompatibility, long-term stability, and electrical isolation properties (Hamelink, 1992). The optical characterization of the μLED arrays was performed with a PM100D optical power meter (Thorlabs, Germany) and the S120VC photodiode power sensor (Thorlabs, Germany). In this study, three μLED arrays (μLED1, μLED2, and μLED3) were positioned as well as the ECG electrodes in a 120° angle encircling the heart (cf. Figure 1B). The current flowing through the resulting 27 μLEDs was controlled via the lab computer using a custom-made μLED-Driver. Control and analysis were realized with a custom-software written in Python (Python Software Foundation, USA).

The proof-of-working of the system was achieved by conducting an experiment measuring pacing thresholds (refer to Supplementary Figure 1). As it can be seen in Figure 1C, no electrical artifacts were induced in any uni-polar lead during photostimulation. This confirms that the PDMS layer provided adequate isolation for the electrical tracks on the PCB. Figure 1C shows as well, that the pacing signal of every μLED array is measured in the opposite electrode, i.e., when μLED1 illuminates the heart, the electrical activity is recorded in V2. In this scenario, since all three electrodes are forming the WCT terminal (zero voltage point), the change of potential that is caused by the expanding electrical wave initiated by μLED1 is measured by V1 and V3, these two contribute to the WCT equally due to their proximity, as a consequence, the maximum change of potential is then recorded by electrode V2.

Inspired by the theory of excitable media, three different illumination patterns, namely row, column, and array protocols, were designed to explore the most effective way of breaking or blocking the arrhythmia-generating spiral waves. The motivation for the row protocol was to drive the spiral wave(s) upward toward the atria, so that the uncoordinated electrical activity collides with non-excitable tissue, getting thereby terminated. As for the column and array protocol, it was intended to evoke a conduction block to terminate the arrhythmia by first exciting a defined area of tissue, consequently leading to a refractory period and therefore causing the spiral wave to be absorbed or blocked at the boundary. The column protocol illuminates only a small area of the heart at once, thus putting several successive segments into the non-excitable state. On the contrary, the array protocol illuminates larger areas at once thereby limiting the number of non-excitable segments to three. In this study, these three approaches were compared to determine which is more effective in terminating arrhythmia.

Accordingly, these three experimental protocols were applied in order to determine illumination patterns that optimize defibrillation success in terms of defibrillation energy and frequency of illumination. Furthermore, it was investigated whether the applied protocols were especially suitable for certain arrhythmia types or arrhythmia frequencies.

In the row protocol, the bottom three μLEDs of all three arrays were activated simultaneously, so that the illumination formed a ring around the heart. Afterwards, the two rows above followed one after the other with one pulse each. In this study, five repetitions of the mentioned sequence were used to deliver a total of k = 15 pulses per attempt. During the column protocol, first the very left column of μLED1 emitted one pulse. Subsequently, the middle and right column of μLED1 followed, afterwards μLED2 and μLED3 continued with the same pattern, respectively. In this case, the protocol was repeated only two times, consequently the light pulses circulated two times around the surface of the heart in a counterclockwise direction resulting in k = 18 pulses. The last protocol tested was the array protocol. Hereby, the entire μLED1 was activated to deliver five pulses to the heart. μLED2 and μLED3 followed, emitting five pulses each as well. In this way, the heart was circled only one time with a total number of k = 15 pulses. All illumination patterns are illustrated in Figure 1D.

In order to investigate the effect of light intensity, three different radiant fluxes, Φ = (1.7 ± 0.1) mW, (2.4 ± 0.1) mW and (3.3 ± 0.1) mW, were chosen for all three illumination patterns. Furthermore, five different pulse frequencies f_stim = 18, 20, 22, 24, and 26 Hz were tested to assess whether the chosen stimulation frequency f_stim influences the success of termination. The pulse duration of each pulse was kept constant at d_stim = 20 ms since this value proved to be particularly successful in earlier experiments (Diaz-Maue et al., 2021). The energy applied for every arrhythmia termination attempt Q was calculated as the product of the number of pulses k, the radiant flux Φ, and the pulse duration d_stim, yielding values between Q_min = (0.51 ± 0.03) mJ and Q_max = (1.19 ± 0.03) mJ (cf. Supplementary Figure 2).

Finally, a positive control group was introduced with N_m = 4 mice. Defibrillation attempts were performed with global illumination using three high-power LEDs arranged as described in Quiñonez Uribe et al. (2018). Here, the light intensity was LI= 1 mW/mm² which corresponds to a radiant flux of Φ = (58 ± 5) mW. Similarly to the three proposed protocols, photostimulation was performed with k = 15 pulses, pulse width d_stim = 20 ms, and frequencies of f_stim = 18, 20, 22, 24, and 26 Hz. This resulted in an energy of photostimulation of Q = (52.2 ± 4.5) mJ.

In cardiac tissue, ventricular tachyarrhythmia is subdivided into tachycardia (VT) and fibrillation (VF). In order to correctly assess the successes or failures of photodefibrillation attempts, it is necessary to classify the individual arrhythmia into different categories depending on their complexity.

The obtained ECG recordings were post-processed after the experimental series to classify the arrhythmia type. For this purpose, a segment of every uni-polar lead measurement was extracted 1 s before every photostimulation attempt. Three different procedures were applied to determine the complexity of an arrhythmia. First, the evaluation of the morphology of the ECG signal according to the updated Lambeth convention (Curtis et al., 2013) was considered. Second, a sine fit was applied to the ECG recordings to identify the dominant frequency of the arrhythmia. The sine fit was implemented in the mathematical language Python by means of a curve fit that uses non-linear least squares to fit a sine function to the data. Amplitude, frequency, and phase were thereby fitted simultaneously. The sine fit represents an automated version of the periodicity analysis, as similarly performed by Skanes et al. (1998) in Langendorff perfused sheep hearts. Third, the Fast Fourier Transform (FFT) in combination with a Lorentzian fit was employed. The resulting spectrum of the FFT was used to estimate the dominant frequency of the arrhythmia as well as to investigate the variety of frequencies involved, as also described by Skanes et al. (1998). Based on the results of these methods, all arrhythmia were classified into three categories, namely monomorphic ventricular tachycardia (mVT), polymorphic ventricular tachycardia (pVT), and ventricular fibrillation (VF), with spatiotemporal complexity increasing from mVT to pVT and VF (Curtis et al., 2013).

Figure 2 shows illustrative examples of the three types of arrhythmia. The ECG representative recordings shown here were all acquired by electrode V3. Clearly, it is shown that the ECG turns into a more irregular pattern the more complex the arrhythmia becomes.

Accordingly, the sine fit for the rather regular mVT is almost in perfect agreement with the ECG signal, apart from slight variations in the amplitude (cf. Figure 2A). Furthermore, the increasing complexity of the arrhythmia leads to the occurrence of moderate deviations of the fit to the ECG signal in pVT (cf. Figure 2B) and to major deviations in VF (cf. Figure 2C). In order to quantify the observed deviations, the sine fit algorithm computes the normalized difference between the fitted signal and the original signal, in addition to the optimal parameters for amplitude and frequency. This value is then subtracted from 1 and the self-defined indicator “goodness of fit” (gof) is obtained. When gof is close to 1, the fit is more precise. Thereupon, the following intervals were defined, gof > 0.6 for mVT, 0.2 ≤ gof ≤ 0.6 for pVT, and finally, gof < 0.2 for VF.

Moreover, an FFT was performed to evaluate the ECG signal in the frequency domain. Since the FFT returns a spectrum of frequencies that can be detected in the original signal, a Lorentzian fit was applied to quantify the relevance of the different frequencies. The values for the center of the fit and the full width half max (fwhm) gave information about the dominant frequency and the width of the fit indicating if there was only one dominant frequency or if several frequencies contributed substantially. Accordingly, if 1 < fwhm < 10, the arrhythmia was classified as pVT, as mVT if fwhm ≤ 1, and as VF if fwhm ≥ 10.

In the case of the mVT in Figure 2A, the Lorentzian curves fitted to the FFT of the ECG signals of V1, V2, and V3 were very narrow and coincide for all three ECG signals. Figure 2B corresponds to a pVT, in this example, the Lorentzian curves were broader and slightly different for each electrode. Furthermore, components of different frequencies could be observed alongside the main peak. Notably, the Lorentzian fits varied substantially for the VF shown in Figure 2C. It is therefore clear that the three electrodes V1, V2, and V3 recorded ECG signals that contain distinct frequency information, which is reflected in the large width of the curves.

The described analysis was performed for all three uni-polar lead recordings. Should one of the three mentioned methods suggest different arrhythmia types, the whole arrhythmia was classified according to the most complex type.

Obtaining information about the frequency content of a signal is a usual task of the FFT (Chorro et al., 2006; Nash et al., 2006; Masse et al., 2007; Umapathy et al., 2010; Caldwell et al., 2012) or of the Short-Time Fourier Transform STFT (Tseng and Tseng, 2020; Coult et al., 2021). However, both require stationary signals (Clayton and Murray, 1993; Mansier et al., 1996; Seely and Macklem, 2004) and are therefore only suited for providing information about a signal as a whole. In the case of cardiac arrhythmia, the obtained signals contain short-term transients which cannot be completely described in the frequency domain alone and also require temporal resolution. Accordingly, to do a comprehensive evaluation of the transitory events in the ECG recordings, especially during photodefibrillation, the continuous wavelet transform (CWT) was chosen to deliver information in the combined time-frequency domain as also described by Torrence and Compo (1998), Abbate et al. (2002). The CWT signal processing was done using MATLAB (Mathworks, USA). First, all photodefibrillation attempts were localized and the recorded signal was shortened to 3 s before and 3 s after photostimulation, where the points t_o and t_f were defined, respectively, as the time for the first and last light pulse given. Afterwards, the CWT was calculated using a Morse wavelet as mother wavelet and the corresponding magnitude scalograms were plotted. Second, in order to obtain the frequencies which correspond to the largest scales present in the CWT, the maximum scale a_max(t_s(i)) for every sampling point t_s(i) was computed, then all scales in the scalogram matrix a which were larger than 0.8·a_max(t_s(i)) were extracted. Following this criteria, the frequencies corresponding to a > 0.8·a_max were obtained yielding an array with several values for every t_s(i). The mean value of every array was then calculated yielding to fc¯(ts(i)) which represents now the instantaneous frequency for every point at time t_s. Furthermore, the dominant frequencies for every time range of interest, namely, before fc_bef, during fc_dur, and after fc_aft photodefibrillation were calculated as the mean value of fc¯(ts(i)) during the required intervals, defined as before t₀ − 3 ≤ t_sbef ≤ t₀, during t₀ < t_sdur ≤ t_f, and after t_f < t_saft ≤ t_f + 3. The same method was applied for every ECG recording acquired, generating the values fc_bef(1−3), fc_dur(1−3), and fc_aft(1−3).

Performing the arrhythmia classification as mentioned in section 2.4 for the total number of recorded arrhythmia N_arr = 458 from N_m = 11 mice, it was found that the majority of events, namely 55 % were classified as pVTs. mVT and VF events made up 24% and 21%, respectively. Regarding the three illumination patterns, N-arr=143±16 arrhythmias were acquired per pattern.

In order to define successful termination, it was examined whether an arrhythmia terminated simultaneously with the end of illumination or whether it continued beyond the point t_f. In case of the latter, the time until the ECG signal returned to the base line was denoted as transient time. About half of all recorded arrhythmia exhibited transient times ranging between 50 and 2200 ms. Since 90 % of all arrhythmia with transients were terminated within 1300 ms after the end of illumination, it was defined that transient times which were less than or equal to 1300 ms correspond to successful termination whereas transient times larger than 1300 ms were classified as non-successful.

When considering the success rates of photodefibrillation regarding the three different illumination patterns without further distinguishing the parameters frequency of stimulation f_stim or applied radiant flux Φ, it became apparent that for mVTs, the column protocol worked most efficiently, since it was able to terminate (96.4 ± 1.7) % of all mVT independently of the selected parameters (see Figure 3A). The row and array protocol both showed a lower success rate around 78 % for mVT. In comparison, no such clear trend emerged for the pVT events. All protocols showed similar success rates for this arrhythmia type, on average (63.6 ± 4.5) %. Moreover, the column illumination pattern offered the lowest success rate of (37 ± 9) % for VF events. Regarding the row protocol, it is remarkable that it terminated VF events more reliably than pVT even though VF represent the more complex arrhythmia type.

In summary, these results indicate that different illumination patterns could be suited for different arrhythmia types. Figure 3A shows that the row protocol is overall most efficient for VF, whereas the column protocol should be chosen when attempting to terminate mVT. During pVT, all illumination patterns worked comparably well. In contrast, the control group employing global illumination terminated 100 % of N_arr = 75 independently of their type. In order to confirm the significance of the presented results, statistical tests were made using a one-tailed z-test for proportions as shown in Figure 3A.

Unfortunately, the sample size for VF events proved to be too limited to provide meaningful results once a further distinction according to the stimulation parameters was to be made. This is a consequence of the circumstance that it was not possible to determine the arrhythmia type during the experiments, leading to a randomized distribution of mVT, pVT, and VF. For further evaluation, mVT and pVT were combined into the group of VT, because this supports a sufficient sample size of arrhythmia when analyzing the success of termination based on different parameters. Considering the individual illumination patterns, the mean number of VT for evaluation per protocol type lies at N-VT=116±14.

In order to analyze the dependence on the applied radiant flux Φ, the termination success was investigated for two fixed stimulation frequencies f_stim= 20 Hz as well as f_stim = 22 Hz and varying radiant flux considering only VT events. Figure 3B top and bottom display the success rates of the row and column protocol exceeding those of the array protocol. It is important to notice that the column illumination pattern obtained the highest success rates with the lowest applied radiant flux Φ = (1.7 ± 0.1) mW and therefore the lowest energy of stimulation Q= (0.61 ± 0.04) mJ for f_stim = 20 Hz. Conversely, the row illumination pattern delivered better results with increasing radiant flux and stimulation energy in both cases. Both the column and the row protocol reached a 100 % success rate for a certain parameter combination thereby providing a reliable mechanism for arrhythmia termination.

Supplementary Figure 3 displays the success rates for each illumination pattern in dependence on the dominant frequency of the arrhythmia f_arr 1 s before illumination considering all VT events. In this case, the dominant frequency f_arr was determined as the weighted average of the results of the sine fit, the FFT, and the Lorentz fit to the FFT spectrum. Moreover the distribution of f_arr can be found in Supplementary Figure 4. The presented results show that the success rate of the row protocol declined with increasing f_arr. Conversely, the column and array protocol showed no such trend. Overall, the array protocol reached the highest success rate only in one case, namely for f_arr=21 Hz. In all the other cases, the success rates of either the row or column protocol exceeded that of the array illumination pattern.

Additionally, experimental validation regarding the possibility of self-termination was performed by inducing arrhythmia with the μLED arrays in place, but without applying photostimulation. Given that the arrhythmia was sustained for at least 5 s, it persisted for (30.97 ± 2.50) s for all arrhythmia N_arr = 53 evaluated. Subsequently, the arrhythmia was terminated using global illumination to prevent permanent damage to the heart due to the ongoing chaotic behavior. Thus, it can be concluded that the induced arrhythmia have a self-termination time above (30.97 ± 2.50) s once they have persisted for more than t_term=5 s. Since the stimulation protocols take a maximum of 15 s upon completion, the termination of the chaotic cardiac rhythm can clearly be attributed to the stimulation by the μLED arrays.

Optical based stimulation does not induce electrical artifacts due to saturation of the amplifier in the ECG recordings, therefore it was possible to analyze time series of the three acquired uni-polar leads before, during, and immediately after all photodefibrillation attempts. Since the dynamic transients of complex arrhythmia could be better characterized with the CWT, the results in this section always use the frequencies fc calculated with the method described in section 2.5. Additionally, all types of arrhythmia were included without distinguishing between them during evaluation. In particular, it could be observed that the local dominant frequencies fc_dur notably changed during photodefibrillation independently of the frequency used for photostimulation f_stim. Trials in which fc_bef remained constant could not be found, refer to Supplementary Figure 5.

Table 1 shows that during illumination all three tested patterns provoked a change of the mean frequency Δ¯fc calculated for the three recorded ECG, where Δfc = fc_dur−fc_bef. This effect is more prominent in the row illumination pattern where 76 % of all arrhythmia displayed a decrease in dominant frequency during photostimulation. Moreover, all three illumination patterns showed to perform better when Δfc was decreased. In other words, when fc_dur decelerates compared to fc_bef, the probability of termination success increases, as it is presented in columns 5 and 6 of Table 1. To support this result, the success rates for deceleration and acceleration measured in different uni-polar leads around the heart were extracted and are shown in Figure 3C. The success rates here were organized in four groups, where the first group dec.₁ contains the attempts in which deceleration could be observed only in one lead, whereas acceleration was detected in the other two leads. In a similar way, the second group dec.₂ contains the tries in which two leads got decelerated, thus one was accelerated. Finally, the last two groups dec.₃ and acc.₃ show the termination rates when all three uni-polar leads measured deceleration, respectively, acceleration. The most efficient approach for terminating arrhythmia is given when the deceleration of fc_dur is achieved in all three monitored areas of the heart, here the array pattern reached the best efficiency. With exception of the row protocol, the termination rate of the group acc.₃ exhibits rather low values.

To further analyze this behavior, Figure 3D shows the scatter plots with marginal smoothed histograms of the conducted experiments. The smoothed histograms on the right show that there is an accumulation of successful cases when the frequency of the arrhythmia during photostimulation fc_dur drops below 20 Hz approximately. Directly comparing the three tested patterns, it can be observed that the distribution of terminated arrhythmia in the column protocol is broader than for row and array patterns, demonstrating that the column protocol is more robust against the arrhythmia measured in this work. Furthermore, the right side histograms suggest that the row protocol is more successful when it is able to reduce fc_dur to around 15 Hz, the same claim holds for the array protocol. Considering the upper histograms, it can be observed that column and array protocol perform better for slow arrhythmias.

In order to illustrate the evolution of arrhythmia during photostimulation Figures 3E–G show an example of a termination event for each row, column, and array patterns with corresponding representative uni-polar lead recording, calculated CWT, and the representation of the local change of frequency on the heart fc_dur. In Figure 3E, it is shown that the frequency and morphology changes of the measured ECG are clearly noticeable during photodefibrillation. Dashed lines indicate the beginning and end of the light pulses. The frequency and amplitude changes can be seen more pronounced in Figure 3F. These magnitude scalograms show that all three row (left), column (center) and array (right) protocol, modify the initial frequency fc_bef to some extent. Figure 3G emphasizes that uni-polar lead recordings were able to register different cardiac dynamics in the different locations. Moreover, it could also be observed that the change of dominant frequency did not necessarily appear in all regions of the heart during photostimulation. This fact might indicate that the electrical activity of some cardiac tissue was not altered in a significant way. For instance, setting a change of frequency threshold to ΔfcT = ± 0.5 Hz for every uni-polar lead, it was found that the employed illumination patterns induced a variation in frequency below ΔfcT in only one ECG recording in 19.9 % of the cases for the row protocol, 23.1 % of the total cases in the column protocol, and 29.7 % in the array protocol.

The presented results also point out that the Multi-Lead recording of cardiac activity delivers more spatial information about the dynamics and temporal evolution of an arrhythmia, e.g., during transient events such as during photodefibrillation, than single lead recordings as commonly used.

The present study describes a possible dependence of the termination success on the development of the arrhythmia frequency during pacing. Though, at the present time, no conclusions can be drawn about the underlying excitation patterns. The complexity of arrhythmia and, consequently, the arrhythmia frequency may change spatially due to meandering spirals. To investigate this, the number of ECG electrodes used is not sufficient. For this, the electrode density needs to be increased in relation to the heart surface. Another method to overcome this challenge would be the usage of optical mapping, however, while it is true that it is widely accepted as the standard technique for the electrical spatial characterization of arrhythmia, it can also produce non-desired electrophysiological changes as shown by Kappadan et al. (2020). Other methods like combined optical mapping with motion tracking (Christoph et al., 2017, 2018) could also be helpful, but the motion compensation can only be done after the experiments during post-processing. Therefore, and with the motivation of opening the possibility to translate the presented results into in vivo trials or also bigger animal models, the authors decided to introduce the Multi-Lead ECG measurement as a local monitor of cardiac electrical activity. Further efforts are being invested in the real-time processing of the acquired ECG measurements. While optical mapping data requires much more computational power, ECG measurements can be directly analyzed during trials with lower processing resources such as microcontrollers or digital signal processors, facilitating the near to instantaneous analysis of measured data.

A somehow obvious limiting factor of the present study is the currently difficult translation of the described technique into clinically relevant in vivo applications. For this purpose, in addition to the above mentioned optimizations of the electrical measurement methods, anatomically adapted measurement and stimulation arrays are needed, the detailed study of which is the subject of current investigations.

Further limitations arise from the optical characteristics of the μLED used, such as the beam angle. This affects the distribution of the emitted light intensity on the cardiac surface and thus the propagation mode of the initiated excitation. Characterization of the propagation parameters would be highly important for understanding the termination processes and for establishing a feedback system based on them. Therefore, increasing ECG resolution and the study of μLED light distribution patterns in combination with optical visualization as a validation method is part of ongoing research.

With the intention of finding parameters that might predict the termination of an arrhythmia even before the photodefibrillation attempt is finished, the CWT processing method was introduced in this work as an alternative method to the FFT for determining the dominant frequency. It could successfully be shown that the CWT facilitates the characterization of arrhythmia frequency during short events of light stimulation. Further looking for prediction parameters, Yang et al. (2019) investigated whether the amplitude spectrum area (AMSA) and spectral energy could be considered indicative for the success of subsequent attempts at arrhythmia termination. As a result, they found that both parameters significantly increased in cases that lead to successful defibrillation. In contrast, if defibrillation did not succeed, no increase in AMSA or spectral energy could be observed. Likewise, in the present study, a change in the amplitude of the ECG recording of arrhythmia was noticed in some cases. Based on these two approaches (frequency and amplitude analysis) and the presented arrhythmia classification method, the design of a new feedback protocol is being pursued, with the goal of improving the photodefibrillation outcome by evaluating the mentioned parameters in real-time.