How to Analyze Electrograms

Abstract

To improve a known method for acquiring optimization in the method for analyzing an electrogram through a control system.SOLUTION: A method for analyzing an electrogram, where the electrogram has been recorded via a catheter inserted into a human body, causes a control system to: detect activation candidates in a first channel (K a), where a plurality of different potential sequences of activation candidates can be defined along the time dimension of the first channel (K a); assign costs comprising one or a plurality of independent cost components to the potential sequences; select a sequence of dominant activations from the potential sequences of the first channel (K a); and select the sequence of dominant activations that fulfills an optimization criterion based on the costs of the potential sequences. The control system assigns cost components to activation candidates, and the costs of the potential sequences comprise the cost components of the activation candidates defining the respective potential sequences.SELECTED DRAWING: Figure 2

Description

[Technical field]

[0001] The invention relates to a method for analysing electrograms according to the general part of claim 1 and to a control system according to claim 14 arranged to implement the proposed method.

[0002] This method is particularly related to atrial fibrillation and atrial flutter. Electrically, atrial fibrillation is the unregulated activation of atrial muscle cells. During atrial fibrillation, the atria contributes minimally to the function of the heart. Thus, atrial fibrillation reduces cardiac output, but does not pose any imminent danger. However, when chronic, atrial fibrillation is associated with increased morbidity and mortality.

[0003] During atrial fibrillation, competing activation rhythms may occur in the atria. Some of these rhythms may be stable, while others may appear and disappear rapidly. Various methods of assessing the severity of atrial fibrillation and evaluating treatment options include detecting rhythms in one or both atria by detecting sequences that represent rhythms. However, there may be more competing rhythms in certain areas of the atria, where atrial fibrillation is more chaotic and complex. Thus, measurements of electrodes placed in the atria will include various rhythms of local cellular depolarization and currents conducted via electrical conduction that are not based on cellular depolarization. By detecting local activations and deriving one or more dominant rhythms, atrial fibrillation can be better understood, for example, during a 3D mapping procedure. However, detecting which activations are local and which rhythms lead to these activations is difficult, especially during atrial fibrillation.

[0004] The known prior art (WO 2015 / 149153) on which the present invention is based relates to a method according to the general part of claim 1.

[0005] WO 2015 / 149153 proposes an annotation method in which in a first calculation step activation candidates are identified and a global estimate of the average activation interval (hereinafter "jump length") is calculated. Then in a second calculation step a sequence of activation candidates is identified, such that the activation candidates minimize the global difference between the individual activation intervals and the estimate of the global activation interval. This prior art achieves this by means of a cost function that assigns a cost to the jumps between the activation candidates. In particular, a cost is assigned to every possible jump between the activation candidates and a path through time with the minimum cost is identified. This global minimization can be achieved by computational methods for shortest path search known from computational graph theory. An example of a suitable shortest path search is the Djikstra algorithm.

[0006] This known algorithm provides a useful basis for annotating the dominant rhythm in one channel of an electrogram, here an intracardiac electrogram. Nevertheless, there is still room for improving this algorithm in order to improve the detection of local rhythms.

[0007] The present invention is based on the task of improving the known methods in such a way that further optimization is obtained with regard to the challenges listed above.

[0008] The above mentioned problem is solved by the features of the characterizing part of claim 1.

[0009] A main realization of the invention is that further information can be used to determine which activation candidates form physiologically valid sequences. This further information is available in the form of properties of the activation candidates themselves. It is now proposed to convert these properties into costs of the activation candidates. Exemplarily, if a cost function is used to find the desired sequence, the path through time then depends on the cost of each activation candidate that is part of the sequence instead of or in addition to the costs for the jumps between the activation candidates.

[0010] Several characteristics have been identified that can be used for such costs. For example, the morphology of the activation candidates of a rhythm may be more similar than that of the activation candidates of a different rhythm. In addition, cross-channel similarities between activation candidates, especially their timing, can be used. Local activations move through channels more slowly than far-field potentials, which often appear essentially simultaneously in some or all channels. In competition with this, a local activation sequence that does not appear in more than one channel or appears in only a few channels may be a regionally restricted activation sequence, whereas a local activation sequence that is seen in more channels or in more separated channels may be the dominant activation rhythm sought. The question of which rhythm to seek may also depend on the clinical application, which can also be coded into the cost.

[0011] Therefore, in order to identify sequences based on the properties of the activation candidates, it is proposed to assign cost components to the activation candidates.

[0012] Also preferably herein, the electrogram is an intracardiac electrogram. The term "intracardial" is understood in a broad sense and relates to measurements from the inside, within and in the immediate vicinity of the human heart, e.g. in the pulmonary veins. Electrograms may alternatively have been recorded by any other invasive technique. In particular epicardial measurements are included.

[0013] In particular, it is proposed that the control system assigns cost components to the activation candidates, such that the cost of a potential sequence has a cost component of the activation candidate that defines the respective potential sequence.

[0014] In a preferred embodiment as defined in claim 2, the jumps between activation candidates are also assigned a cost component, which allows a consistent optimization for jump lengths when looking for periodic sequences and at the same time taking into account the properties of the activation candidates.

[0015] In a preferred embodiment according to claim 3, some cost components depend on the data from the second channel. In particular, the cost components assigned to the activation candidates may depend on the temporal distance between these activation candidates and the activation candidates from another channel. By defining the costs in this way, consistency between the electrodes is optimized. Such consistency is particularly advantageous in situations where there are competing activation rhythms.

[0016] According to claim 4, the cost component dependent on the second channel can have a time-dependent cost component and / or a time-independent cost component. A time-dependent cost component is a cost component that depends on the timing of the activation candidate, such that, all else being equal, a hypothetical change in the timestamp of the activation candidate will affect the cost component. If only the timestamp of the activation candidate is changed, the time-independent cost component will not be changed.

[0017] Without using a mapping system, the control system may not have any data on the relative positions of the electrodes and their respective channels. By using the order of the electrodes and channels, it is still possible to derive which electrodes are physically close to each other (claim 5). For example, a rhythm that is traceable through channels I-III but not through channels IV and V is highly unlikely to reappear in channel VI.

[0018] Claim 6 relates to a preferred embodiment of a method for detecting rhythms by combining unipolar and bipolar signals using a cost component dependent on unipolar channels as a cost component for activation candidates in bipolar channels. The idea is based on the fact that cell depolarization is always negative, and therefore the downward deflection of a unipolar electrode can be advantageously included in the cost component to detect local activation.

[0019] To detect far-field potentials and reduce their influence on the outcome of the analysis, the cost component may rely on data from coronary sinus electrodes and / or surface ECG electrodes (Claim 7). Coronary sinus electrodes can be used to detect actual beats of the heart and increase the cost component of activation candidates immediately after beats that are likely to be far-field interference.

[0020] The embodiment as claimed in claim 8 relates to basing the cost component for the activation candidates on their morphology. According to claim 9, in particular morphological similarities between the activation candidates are included in the cost component, for example by basing the cost component of a jump between the activation candidates on the morphological similarity of two activation candidates connected by a jump.

[0021] Claim 10 relates to the possibility of static and / or dynamic cost components. Static cost components are sequence independent, whereas dynamic cost components vary with the sequence. For example, the cost component of a jump can depend on the average jump length of the other jumps in this sequence, thus optimizing the overall similarity of the jump lengths in the sequence, rather than the similarity of the jump lengths to a predefined average jump length.

[0022] According to claims 11 and 12, the channels may be analyzed repeatedly and the processing of a channel may depend on previously processed channels. Possibly, more than one iteration may be performed for at least one channel.

[0023] In one embodiment according to claim 13, the analysis is performed periodically, in particular in real time, and the control system periodically analyzes the channels, for example every second. Since the dominant rhythm may change, preferably the detected sequence may be retroactively changed by new analysis cycles. To reduce the probability of detecting two different rhythms and of frequent changes in which one is detected, the results of previous analyses may be included in the cost. By including the past analyses in the cost, an artificial hurdle is created for changes between the detected rhythms.

[0024] Another teaching of equal importance, as set forth in claim 14, relates to a control system configured to implement the proposed method, the control system being configured to receive and / or measure electrograms.

[0025] All explanations given in relation to the proposed method are fully applicable.

[0026] In the following, an embodiment of the invention is explained with reference to the drawings. [Brief description of the drawings]

[0027] [Figure 1] FIG. 1 illustrates finding sequences through analysis of cost components. [Diagram 2] FIG. 4 illustrates the iterative derivation of cost components for multiple channels according to one embodiment. [Diagram 3] FIG. 13 illustrates alternative channel-dependent and time-dependent derivation of cost components.

[0028] The figures show here also one embodiment of the proposed method for analyzing an electrogram 1, which is preferably an intracardiac electrogram. In FIG. 1 it is generally shown how a cost function is used to determine a sequence 2 from a number of nodes and edges between the nodes. In FIGS. 2 and 3 an exemplary part of an intracardiac electrogram 1 is shown quite diagrammatically. The illustrated intracardiac electrogram 1 has a coronary sinus channel M, which is only very diagrammatically drawn at the top, and eight bipolar atrial channels L, purely by way of example. The corresponding intracardiac electrogram 1 may have been recorded during a mapping procedure. The intracardiac electrogram 1 has preferably been recorded during atrial fibrillation.

[0029] The analysis is performed by a control system, e.g. a general purpose computer, a mapping system, or a dedicated hardware device including a processor. The proposed method is therefore a computer-implemented method. Any of the steps described herein may therefore be performed by a control system, which may be adapted accordingly.

[0030] The electrogram 1 is recorded via a catheter inserted into the human body. The catheter has a number of electrodes. At least one channel K of the electrogram 1 is recorded by, and preferably between, the electrodes. The channel K may be unipolar, recorded by one electrode of the catheter, or bipolar, recorded between two electrodes of the catheter.

[0031] The proposed method generally relates to detecting rhythms in the form of sequences 2 in electrograms 1, in particular in quasi-real-time during surgery. The control system can display the results of the analysis to a user, in particular a surgeon. This information may be used by the user for diagnostic purposes and / or to devise a treatment plan.

[0032] An electrode placed in the atrium always measures the local electrical depolarization of the cells and the potential that reaches the electrode via electrical conduction, also known as the far-field potential or far-field interference. During normal cardiac activity in a healthy atrium, the local activation has one rhythm that corresponds to the atrial beating. During atrial fibrillation, different sites may have different rhythms, and one site may even produce competing rhythms. Furthermore, random activation and activation from adjacent tissues will be visible in the channel K of a particular electrode. To better understand the complex composite image of atrial fibrillation, it is desirable to annotate one or a few rhythms for a particular channel K, for example to later create a map of the electrical activation of the atrium.

[0033] As part of the proposed method, the control system detects activation candidates 3 in a first channel K_a. The terms "first" and "second" are not limiting and are used only to aid in the understanding of the description. Any channel K may be a first channel K_a and any other channel K may be a second channel K_b, not yet described. The activation candidates 3 are portions of the signal of the respective channel K that have a peak with certain characteristics or are otherwise selected as being capable of representing a local depolarization. Several such activation candidates 3 are shown in the intracardiac potential diagram 1 of Figs. 2 and 3.

[0034] A number of different potential sequences 4 of activation candidates 3 can be defined along the time dimension 5 of the first channel K_a. Essentially any combination of at least 3, preferably at least 5, more preferably at least 10 activation candidates 3 that are subgroups of all activation candidates 3 can form a potential sequence 4. It should be understood that the number of potential sequences 4 is greater than 1 and that sequence 2, which is a subgroup of activation candidates 3, does not have all activation candidates 3 of channel K.

[0035] The control system assigns a cost to the potential sequence 4, which has one or more independent cost components. Generally speaking, the cost of a potential sequence 4 is constituted by the cost components of the potential sequence 4, in particular the cost components of the elements of the potential sequence 4. For example, a sequence 2 with several activation candidates 3 distributed over a time dimension 5 can have cost components assigned to the jumps 6 between the activation candidates 3, each jump 6 being the time distance between an activation candidate 3 and the next activation candidate 3 selected from the potential sequence 4. The cost component assigned to the jump 6 can simply be the length of the jump 6. Exemplarily further, the cost can be the sum of the cost components. However, it should be noted that the length of the jump 6 alone is not a valid cost function, since in this case the sum of the costs would simply be the time difference between the first activation candidate 3 and the last activation candidate 3.

[0036] The control system selects a dominant activation sequence 2 from the potential sequences 4 of the first channel K_a, for example by selecting the potential sequence 4 with the lowest cost. More generally, the control system selects a dominant activation sequence 2 that satisfies an optimization criterion based on the cost of the potential sequences 4. In particular, optimization criteria that are mathematically equivalent or similar to finding a minimum cost are possible.

[0037] Also preferably here, the first channel K_a is a bipolar channel K and / or an atrial channel L and / or is not a coronary sinus channel M.

[0038] It is proposed that the control system assigns cost components to the activation candidates 3, where the cost of a potential sequence 4 has a cost component of the activation candidate 3 that defines each potential sequence 4. A general example of how a sequence 2 can be selected is shown in FIG. 1. Seven activation candidates 3 are shown in FIG. 1, which are shown as nodes, and several possible jumps 6 between the nodes. To make the diagram easier to understand, not all possible jumps 6 are shown. Jumps 6 that are above a predefined length and / or below a predefined length may be culled to reduce the number of potential sequences 4. Each of the activation candidates 3 has an assigned cost component "c", the subscript being the number of the activation candidate 3. As an optional addition, FIG. 1 also shows cost components assigned to the jumps 6 between the nodes, marked with "j", with a subscript indicating which node the jump 6 leads from to which node. For example, the selected sequence 2, shown by the solid line, will have a cost consisting of the sum of the cost components of the activation candidates 3 having numbers 1, 2, 4, 5 and 7, and the cost components of the jumps 6 between nodes 1 and 2, 2 and 4, 4 and 5, and 5 and 7.

[0039] While it is possible to use any mathematical equivalent of assigning cost components to nodes, for example by making the cost component of jump 6 dependent on its start node and target node, and subsequently assigning cost components to activation candidates 3, thereby hiding the cost components of the nodes in the cost component of jump 6, it should still be understood that the cost components of activation candidates 3, regardless of their mathematical expression, depend on the respective activation candidate 3 and can be distinguished from the cost components that depend on jump 6, in particular on the jump length CL, except that the next jump CL does not depend on the activation candidate 3 itself. Here, and preferably, the cost components of activation candidates 3 do not depend on the timing of activation candidate 3 in channel K and / or on the position of activation candidate 3 in sequence 2.

[0040] Preferably, the cost components of the activation candidates 3 are time-invariant within a channel K. Thus, for example, if three activation candidates 3 with different morphologies are placed at 0 ms, 100 ms and 200 ms in a hypothetical channel K, if all cost components are time-invariant, the cost of sequence 2 will be the same if the order of the activation candidates 3 is changed by swapping the positions of two of the activation candidates 3, all else being equal. Of course, it is difficult to realize such a swap in a real signal.

[0041] It should also be noted that the costs, including the cost components mentioned herein, are non-zero and non-infinite, or prohibitively large and therefore effectively infinite. Such infinite or zero costs may be present in addition to the costs mentioned above, and in different iterations of the method, various portions of the costs may be zero or infinite, but the costs mentioned here and relevant herein are always those that, by design, may realistically lead to the selection of each element of sequence 2.

[0042] In Fig. 2 and Fig. 3, a preferred embodiment of an algorithm for selecting a sequence 2 from several activation candidates 3 is shown. In the following, this embodiment is used as an example to explain further advantageous embodiments of the invention. In Fig. 2, as a first step, a first channel K_a, here the top atrial channel L, is selected and processed first (top arrow on the right). The control system detects activation candidates 3 (marked with a circle) in the first channel K_a and optionally assigns each activation candidate 3 a position in time, for example based on the position of the main peak of the activation candidate 3. The detection of activation candidates 3 can be performed using known algorithms, for example based on peak detection, morphology detection or another technique. Moving further down in Fig. 2, the control system assigns a cost component c to the activation candidates 3.

[0043] According to one embodiment, it is proposed that the control system assigns a cost component to the jumps 6 between the activation candidates 3. This embodiment can also be seen in FIG. 2 by the cost component j. The cost of a potential sequence 4 then has a cost component of the jumps 6 between the activation candidates 3 that define the respective potential sequence 4. The component assigned to the jumps 6 may depend on the difference in length of the jumps 6 and on an estimate of the average jump length CL (also called cycle length). The average jump length CL may be the expected jump length CL for the expected dominant rhythm. Known methods for estimating the average jump length CL include the calculation of the dominant frequency and other methods, for example using the Fourier transform. In FIG. 2 it is shown that the average jump length CL for the first channel K_a can be estimated from all atrial channels L. In general, the average jump length CL for a channel K can be estimated from the same channel K, or from several channels K, or from another channel K, as described for the second channel K_b.

[0044] In one embodiment, potential sequences 4 are restricted by defining start and / or end zones for sequences 2 and considering only sequences 2 that have first and / or last activation candidates 3 within the respective zones.

[0045] With certain information it is already possible to select sequence 2 from potential sequences 4. It is also possible to process all channels K in this way and thus analyse the electrogram 1.

[0046] However, according to one embodiment, it is proposed that the cost of the potential sequences 4 of the first channel K_a, in particular the cost components assigned to the activation candidates 3, includes cost components that depend on data derived from the second channel K_b. In this embodiment, contrary to the example given above, the first analyzed channel K in FIG. 2 is the second channel K_b, and one of the other channels K is the first channel K_a. Alternatively, as explained below, the top channel K can be analyzed twice, in which case it does not matter which channel K is defined as the first channel K_a. In FIG. 2 it is shown that after analyzing the top channel K (here the second channel K_b), an average jump length CL estimate can be derived from the selected sequence 2 of the second channel K_b. This estimate is then used to find the sequence 2 in the first channel K_a by using data from another channel K, here the average jump length CL estimate.

[0047] In Fig. 3, another preferred embodiment of the analysis of the first channel K_a, here the channel K with the number II, is shown by assigning a cost component to the activation candidate 3 of the first channel K_a depending on the second channel K_b (the channel K with the number I, previously analyzed). Here, and preferably, the cost component assigned to the activation candidate 3 of the first channel K_a depends on the activation candidate 3 of the selected sequence 2 of the second channel K_b, previously analyzed. In Fig. 3, it is shown how the cost component of the activation candidate 3 can depend on the temporal distance to the activation candidate 3 of the second channel K_b. If a similar sequence 2 is found in another channel K, the probability that the sequence 2 actually represents the dominant rhythm increases.

[0048] Also preferably, here, the first channel K_a and the second channel K_b are channels K from the same catheter and / or bipolar channel K and / or electrode of the atrial channel L, and / or neither the first channel K_a nor the second channel K_b is the coronary sinus channel M.

[0049] According to one embodiment, it is proposed that the cost component dependent on the data derived from the second channel K_b has a time-independent cost component and / or a time-dependent cost component, which is the same independent of the timing of each element of the potential sequence 4 to which it is assigned. The above example with the average jump length CL is time-independent.

[0050] Generally speaking, the time-independent cost component preferably depends on the difference between the jump lengths CL of the first channel K_a and on an estimate of the average jump length CL of the second channel K_b.

[0051] The above example having a cost component of an activation candidate 3 is time-dependent because this cost component depends on the timing of the activation candidate 3 with respect to the activation candidates 3 of another channel K. Generally speaking, a time-dependent cost component preferably depends on the timing of the activation candidate 3 of a first channel K_a compared to the timing of the activation candidate 3 of a second channel K_b.

[0052] It is preferable that the second channel K_b is an adjacent channel K of the first channel K_a. The control system can use the order of the electrodes to derive which channel K is the adjacent channel K. Typically, the channel K or electrode number is selected based on the physical location of the electrode in the catheter. Therefore, an easy way to estimate the physical neighbors of the electrodes is to use the order of the provided channels K.

[0053] Another way of assigning cost components to potential sequences 4 is based on the unipolar signal of bipolar channel K. If the first channel K_a is a bipolar channel K, the cost of the potential sequence 4 of the first channel K_a, in particular the cost components assigned to the activation candidates 3, may have cost components that depend on the data derived from one or two unipolar channels K of the electrodes of the bipolar channel K, in particular on the data regarding the polarity of the electrodes. This embodiment is based on the recognition that cellular depolarization is always negative and thus a downward deflection of the electrodes indicates whether the activation is local or not.

[0054] According to another embodiment, it is proposed that the cost of the potential sequence 4 of the first channel K_a, in particular the cost components assigned to the activation candidates 3, have cost components that depend on data derived from the coronary sinus channel M and / or surface ECG electrodes. The coronary sinus electrodes can be used to estimate which activation candidates 3 are likely to be far-field interferences. Thus, the cost components that depend on data derived from the coronary sinus channel M and / or surface ECG electrodes can depend on data on the time distance between the activation candidates 3 in the first channel K_a and on the ventricular beats detected on the coronary sinus channel M and / or surface ECG electrodes. As an alternative to increasing these cost components, activation candidates 3 that are likely to be far-field interferences can be completely removed so that they are not selected as part of the sequence 2.

[0055] The cost components assigned to the activation candidates 3 may also have cost components that depend on the morphology of the respective activation candidates 3. The morphology may include descriptive features such as, for example, amplitude and / or dV / dt slope and / or RMS value and / or frequency content and / or correlation coefficient, or the cost may be derived from the output of an algorithm, such as, for example, a machine learning model that takes electrogram intervals as at least one of its inputs.

[0056] In another embodiment, the clinical use of the analysis may be coded into the cost components, specifically the cost components assigned to activation candidate 3. The clinical use may be input by the user, specifically selected from a list.

[0057] The cost of a potential sequence 4 may further have a cost component that depends on a measure of similarity, especially morphological, between the activation candidates 3 of the selected sequence 2 of the first channel K_a and / or of the previously analyzed second channel K_b. Similar activation candidates 3 are more likely to form valid sequences 2. By using activation candidates 3 of sequences 2 that have already been determined to be valid, the number of bad morphologies in the comparison is reduced.

[0058] Another possibility is that the costs of the potential sequences 4 of the first channel K_a, in particular the cost components assigned to the activation candidates 3 and / or to the jumps 6, comprise static cost components independent of the potential sequences 4 and / or dynamic cost components dependent on the potential sequences 4. It is often computationally easier not to use cost components that depend on the sequence 2 itself, but in particular for morphological comparisons it can be advantageous to compare the morphology of the activation candidates 3 of the sequence 2 and then adapt the cost of the sequence 2.

[0059] Preferably, the dynamic cost components assigned to the jumps 6 depend on a measure of similarity of the jump lengths CL of the respective potential sequences 4, thereby rewarding jumps 6 of similar length. Additionally or alternatively, the dynamic cost components assigned to the activation candidates 3 may depend on a measure of similarity of the activation candidates 3 of the respective potential sequences 4.

[0060] Again, preferably, the analysis of channel K can be summarized as using global optimization rather than rigorous forward search. Here, too, a graph search algorithm, such as Dijkstra's, is preferably used to select sequences that satisfy the optimization criteria. In general, a global optimization algorithm is preferably used on the costs of a particular channel to select sequences that satisfy the optimization criteria.

[0061] Nevertheless, for computational efficiency, the control system may repeatedly analyze channel K, with the analysis of channel K depending on the analysis of a previously analyzed channel K. Preferably, the cost components of potential sequences 4 of channel K depend on the analysis of a previously analyzed channel K. More preferably, at least one channel K is analyzed at least sometimes, especially twice if a measure of accuracy of the analysis is not met, and the second analysis depends on the analysis of a channel K analyzed after the first analysis. This may be the case especially for the first analyzed channel K.

[0062] As mentioned above, in one embodiment it is proposed that the time-dependent cost component depends on the timing of the activation candidates 3 of the first channel K_a compared to the timing of the activation candidates 3 of the selected sequence 2 of the second channel K_b previously analysed.

[0063] Turning to clinical applications, the control system periodically receives an updated channel K, analyzes the updated channel K, and the updated channel K has previously analyzed parts and new parts, and the selection of activation candidates 3 in the previously analyzed parts changes from time to time with the analysis of the updated channel K. This embodiment is considered to be very rare. The analysis does not usually change the annotations made in the previous analysis round. However, the dominance rhythm in a particular region may change over time, and this change may have appeared in previous data, although not yet in time for the selected sequence 2. Retrospective relabeling of the data allows for a faster reaction and better detection of such changes.

[0064] Preferably, the cost component of a potential sequence 4 in the analysis of the update channel K has a cost component that depends on the selection of the activation candidates 3 in the previously analyzed part. By providing a kind of memory of previous labels in the cost component, frequent changes switching between two equally dominant rhythms are made less likely.

[0065] Another teaching of equal importance relates to a control system adapted to implement the method according to any one of claims 1 to 13, the control system being adapted to receive and / or measure an electrogram 1. The control system may comprise a processor and / or a memory. The control system may be a local unit with a processor, possibly a user interface etc. The control system may also comprise a cloud processor in one embodiment. Thus, the control system does not have to be limited to one device. The control system may comprise an interface to a catheter. The control system is adapted to implement the proposed method.

Claims

1. A method for analyzing, via a control system, an electrogram (1), the electrogram (1) being recorded via a catheter inserted into a human body, the catheter having a plurality of electrodes, at least one channel (K) of the electrogram (1) being recorded by the electrodes; Detecting an activation candidate (3) in a first channel (K_a) by the control system; A plurality of different potential sequences (4) of activation candidates (3) can be defined along a time dimension (5) of the first channel (K_a), and the control system assigns costs having one or more independent cost components to the potential sequences (4); A method for selecting, by the control system, a dominant activation sequence (2) from the potential sequences (4) of the first channel (K_a), and selecting, by the control system, the dominant activation sequence (2) that satisfies an optimization criterion based on the costs of the potential sequences (4), comprising: the control system assigns cost components to activation candidates (3), and the costs of the potential sequences (4) comprise the cost components of the activation candidates (3) that define each of the potential sequences (4).

2. 2. The method of claim 1, further comprising the step of: assigning cost components to jumps (6) between activation candidates (3) by the control system, and the cost of the potential sequences (4) comprises the cost components of the jumps (6) between the activation candidates (3) that define each of the potential sequences (4).

3. 2. The method of claim 1, wherein the cost of the potential sequence (4) of the first channel (K_a) has a cost component that depends on data derived from a second channel (K_b), and the first channel (K_a) and the second channel (K_b) are channels (K) from the same catheter and / or bipolar channel (K) and / or atrial channel (L) electrodes, and / or neither the first channel (K_a) nor the second channel (K_b) is a coronary sinus channel (M).

4. the cost components dependent on data derived from the second channel (K_b) have time-independent and / or time-dependent cost components, 4. The method according to claim 3, characterized in that

5. 4. The method according to claim 3, characterized in that said second channel (K_b) is an adjacent channel (K) of said first channel (K_a).

6. 2. The method of claim 1, wherein the first channel (K_a) is a bipolar channel (K) and the cost of the potential sequence (4) of the first channel (K_a) has a cost component that depends on data derived from one or two unipolar channels (K) of the electrodes of the bipolar channel (K).

7. 2. The method of claim 1, wherein the cost of the potential sequences (4) of the first channel (K_a) has a cost component that depends on data derived from a coronary sinus channel (M) and / or surface ECG electrodes.

8. 2. The method of claim 1, wherein the cost components assigned to activation candidates (3) have a cost component that depends on the morphology of the respective activation candidate (3).

9. 2. The method of claim 1, wherein the cost of the potential sequence (4) has a cost component that depends on a measure of similarity between the activation candidates (3) of the first channel (K_a).

10. 2. The method of claim 1, wherein the cost of the potential sequence (4) of the first channel (K_a) has a static cost component that is independent of the potential sequence (4) and / or a dynamic cost component that is dependent on the potential sequence (4).

11. 2. The method of claim 1, further comprising the step of repeatedly analyzing a channel (K) by the control system, the analysis of a channel (K) depending on the analysis of a previously analyzed channel (K), and the cost components of the potential sequences (4) of a channel (K) depending on the analysis of a previously analyzed channel (K).

12. 5. The method of claim 4, characterized in that the time-dependent cost component depends on the timing of activation candidates (3) of the first channel (K_a) compared to the timing of activation candidates (3) of a selected sequence (2) of a previously analyzed second channel (K_b).

13. 2. The method of claim 1, further comprising the steps of: periodically receiving an updated channel (K) by the control system; analyzing the updated channel (K), the updated channel (K) having a previously analyzed portion and a new portion; and the selection of activation candidates (3) in the previously analyzed portion changing from time to time with the analysis of the updated channel (K).

14. 14. A control system configured to carry out the method according to any one of claims 1 to 13, said control system being configured to receive and / or measure said electrograms (1).

15. The method of claim 2, wherein the cost component assigned to a jump (6) depends on the difference between the lengths of the jumps (6) and an estimate of the average jump length (CL).

16. The time-independent cost component depends on the difference between the jump lengths (CL) of the first channel (K_a) and on an estimate of the average jump length (CL) of the second channel (K_b) and / or 5. The method of claim 4, wherein the time-dependent cost component depends on the timing of activation candidates (3) of the first channel (K_a) compared to the timing of activation candidates (3) of the second channel (K_b).

17. The method described in claim 10, wherein the dynamic cost components assigned to jumps (6) depend on a measure of similarity of the jump lengths (CL) of each of the potential sequences (4) and / or the dynamic cost components assigned to activation candidates (3) depend on a measure of similarity of the activation candidates (3) of each of the potential sequences (4).

18. The method of claim 11, further comprising analyzing at least one channel (K) at least from time to time, twice, the second said analysis being dependent on an analysis of a channel (K) analyzed after the first said analysis.

19. The method described in claim 13, wherein the cost component of the potential sequence (4) in the analysis of the update channel (K) has a cost component that depends on the selection of an activation candidate (3) in the previously analyzed portion.

Images