Analyzing Long-Term Electrocardiography Recordings to Detect Arrhythmias in Mice

Cardiac arrhythmias are common, affecting millions of patients worldwide¹. Ageing populations show a growing incidence and thus a major public health burden resulting from cardiac arrhythmias and their morbidity and mortality². Current treatment strategies are limited and often associated with significant side effects and remain ineffective in many patients^3,4,5,6. Novel and innovative therapeutic strategies that causally target arrhythmia mechanisms are urgently needed. To study the complex pathophysiology of arrhythmias, suitable animal models are necessary; mice have been proven to be an ideal model species to evaluate the genetic impact on arrhythmias, to investigate fundamental molecular and cellular mechanisms, and to identify potential therapeutic targets^7,8,9. Continuous ECG recording is a well-established concept in the clinical routine of arrhythmia detection¹⁰.

Implantable telemetry devices are among the most powerful tools available to study electrophysiology in mice as they allow continuous recording of the ECG (a common approach is to implant the leads in a lead-II position) over a period of several months in freely moving, awake mice^11,12. However, due to the huge number of data points (up to more than 1 million QRS complexes per day) and limited knowledge of murine standard values, the analysis of telemetry data remains challenging. Commonly available telemetry transmitters for mice last up to 3 months, leading to the recording of up to 100 million QRS complexes. This means that pragmatic analysis protocols are much needed to reduce the time spent with each individual dataset and will allow researchers to handle and interpret this huge amount of data. To obtain a clean ECG signal upon recording, transmitter implantation needs to be optimal-the lead positions should be as far apart as possible to allow higher signal amplitudes.

The interested reader may be referred to a protocol by McCauley et al.¹² for more information. Further, to minimize noise, cages and transmitters must be placed in a silent environment not prone to any disturbance, such as a ventilated cabinet with controlled environmental factors (temperature, light, and humidity). During the experimental period, lead positioning must be checked regularly to avoid loss of signal due to lead perforation or wound healing issues. Physiologically, there is a circadian alteration in ECG parameters in rodents as in humans, generating the need for a standardized approach to obtaining baseline ECG parameters from a continuous recording. Rather than calculating mean values of ECG parameters over a long period, analysis of a resting ECG similar to that in humans should be performed to obtain basic parameters such as resting heart rate, P wave duration, PR interval, QRS duration, or QT/QTc interval. In humans, a resting ECG is recorded over 10 s, at a normal heart rate of 50-100/min. This ECG includes 8 to 17 QRS complexes. An analysis of 20 consecutive QRS complexes is recommended in the mouse as "resting ECG equivalent". Because of the above-mentioned circadian alteration, a simple approach is to analyze two resting ECGs per day, one at daytime and one at night time. Depending on the light on/off cycle in the animal facility, suitable times are selected (e.g., 12 AM/PM), and basic parameters are obtained.

Next, a heart rate plot over time is used to detect relevant tachy- and bradycardia, with consecutive manual exploration of these episodes to get a first impression. This heart rate plot then leads to the important parameters of maximum and minimum heart rate over the recorded period as well as heart rate variability over time. After that, the dataset is analyzed for arrhythmias. This article describes a step-by-step approach to obtain these baseline ECG data from long-term telemetry recordings of awake mice over a recording period of up to three months. Further, it describes how to detect arrhythmias using the software, Ponemah version 6.42, with its analysis modules, ECG Pro and Data Insights, developed by Data Sciences International (DSI). This version is compatible with both Windows 7 (SP1, 64 bit) and Windows 10 (64 bit).

1. Prearrangements

1. Start Ponemah 6.42 software, and confirm the username and serial number of the software license on the following screen by clicking on Continue.
2. Load the experiment containing the ECG of interest
   1. If Ponemah is started for the first time, note that the Ponemah Get Started dialog opens, offering three options: 1) Create Experiment, 2) Load Experiment, 3) Import Experiment.
      1. Select Load Experiment to open a file. Once the Browse for Folder dialog opens, select the experiment file with the extension (".PnmExp"), and load the file by clicking on Open.
      2. To open a data set recorded in Ponemah 5.x or Dataquest ART, use the Import Experiment function.
         NOTE: If the software is reopened, the last experiment is loaded automatically within the main window for further review. In the menu under Experiment, the same three options as in the Ponemah Get Started dialog are offered: 1) Create Experiment, 2) Open Experiment, 3) Import Experiment.
3. Click on Actions / Start Review from the toolbar, and go to the Load Review Data dialog box, which provides an overview of all the mice subjects and the respective signals recorded within the loaded experiment (Figure 1A).
   1. Select the recording referring to the mouse that will be analyzed by clicking on the checkbox next to the mouse number in the left panel Subjects.
   2. Select the checkbox next to ECG in the middle panel Signal Types.
   3. Determine the duration of the signal that will be analyzed with the extreme right panel Time Range. Observe the following three options: Entire Experiment, which will load all the ECG data from the selected mouse; Parser Segments, which will only load data contained within Parser Segments added during a previous review session; Time Range, which allows the loading of a specific time range either by entering a specific start and end date or by entering a time duration.
   4. To save the selection, use the Loading Definitions dialog in the upper left corner, which also allows loading of previously saved selections.
      NOTE: The size of the selected data will be indicated by either a green or red bar based on the file size in the upper right corner under Data Size. Currently, the software allows loading up to 3 GB of data for Review; 3 GB data can be equivalent to a continuous 24 h recording of 3-4 days.
   5. Click on OK to load the selected data set into Review.
4. After clicking OK, observe that the Ponemah Review window opens along with several separate windows. Although the Events and Parameters windows are opened and shown by default, manually select other necessary windows based on graphs of interest under the Graphs/Graph Setup toolbar.
   NOTE: If Events and Parameters do not open by default, they can be activated by Window/Parameters and Window/Events.
   1. Make note of the Graph Setup dialog, which allows setting up to 16 graphical windows providing both raw data (e.g., ECG signals) and derived parameters (e.g., XY loop) (Figure 1B).
   2. Select the Enable Page checkbox to show the ECG tracing. In the list below, choose the line including the desired mouse (under Subject) and data type (under Presentation) by clicking on the respective checkbox on the left. Use the following settings: Type, Primary; Label, up to 11 characters displayed in the title bar of the window; Time, 0:00:00:01 indicating seconds as the unit used.
      1. Enter the appropriate information in the Label, Unit, Low, and High text boxes.
         ​NOTE: Enable two more pages, Heart Rate Trend and Template, which are helpful for the analysis of basic ECG parameters and for arrhythmia detection.
   3. In the Heart Rate Trend page, activate another graph page and define as a trend to plot the heart rate (HR) over time. Use the following settings to plot the HR for the entire data that are loaded in Review Type, Trend; Input, ECG; Presentation, HR; Label, HR Trend; Unit, bpm; Low: 50; High: 1000.
      ​NOTE: Templates are ECG cycles with accurately placed marks that can be used as representative ECG cycles for pattern recognition analysis. They allow the selection of a small number of representative cycles and the matching of these templates to the entire ECG, thereby annotating all other cycles accordingly.
      1. To use the template function, create a Template Library (a file in which the Templates are stored) for each subject. Do so by selecting the Template Setup/Template library option (Figure 2A).
      2. Select New… from the dropdown menu under Template library to create a new Template Library.
         ​NOTE: There are a few more options in the dropdown menu: No Binding disassociates any previous configured Template Library from the Subject. Browse associates an existing Template Library that was configured during a previous Review session.
      3. Next, configure a Template graph, select Setup/Experiment Setup/Graph Setup, and select a Page to use as a Template graph page. Check the Enable Page check box, select Template for the Type, and ensure that Input reflects the users Subject/Channel selection. Type the appropriate information in the Label, Unit, Low, and High text boxes, and click on the OK button to display a graphic window for each graphic page that was configured under Graphic Setup as shown in Figure 2B.
         NOTE: A graph setup page for Template settings will appear as shown in Figure 2B. According to the page selected in the graph setup dialog, the title bars of the windows are labeled from pages 1 - 16, based on the number of pages enabled (examples for pages 1, 2, 3 are shown in Figure 3A, Figure 3B, and Figure 3C, respectively).
5. Make some important adjustments in the ECG tracing window (Figure 3A).
   1. Adjust the Y-axis representing the ECG amplitude by double-clicking within the ECG tracing window to select Scaling. Here, select Autoscale or adjust manually by using High Axis Value and Low Axis Value.
   2. To adjust the X-axis representing the time, click on the respective toolbar icons: Zoom In to expand the time span (i.e., fewer QRS complexes are shown), Zoom Out to compress the time span (i.e., more QRS complexes are shown).
   3. To show DT (Delta Time) and RT (Real Time) in the lower left corner, left-click on the ECG tracing with the cursor (a vertical black line) to position and see real-time information at the cursor location under RT.
   4. As DT shows a time interval of the user's choice, right-click on the window, to both position the cursor and to select Reset Delta Time within the dialog that appears. Left-click to another position within the ECG tracing to measure the time interval between the selected time intervals shown as Delta Time (DT).
6. Ensure that each segment of the tracing (P, Q, R, T wave) is recognized and correctly annotated for the ECG analysis. To achieve this, define and analyze Attributes by a right click within the ECG window, and click on the Analyze/Attributes option.
   NOTE: The ECG Analysis Attributes dialog opens as shown in Figure 4A. At the top of this dialog, several options (QRS, PT, Advanced, Noise, Marks, Notes, Precision) allow for the adjustment to various settings (explained below).
   1. Click on the QRS tab to adjust R and QS identification.
      1. QRS Detection Threshold: Apply the entered percentage to the largest derivative peak illustrated within the waveform window.
         NOTE: Define an optimal value to eliminate undersensing (i.e., some R waves may be not detected) and oversensing of peaks (i.e., other peaks, such as T waves, may be misinterpreted as R waves). The threshold (region highlighted in pink in Figure 4A) should intersect with the derivative of the ECG. Ideally, the attribute values, which help to identify QRS complexes and to distinguish between clear cycles and noise events, should be maintained at constant (or almost constant) levels between all recordings from one project to allow comparability over different animals per project. After establishing optimal values, maintain the attribute settings for the entire recording.
      2. Min R Deflection: Ensure that the R amplitude change (based on minimum/maximum signal values and not isoelectric levels) exceeds this value before annotating it as an R wave.
         NOTE: Min R Deflection should be ideally higher than noise and lower than the expected deflection of R wave. A low value may result in noise sensing and therefore oversensing, a high value may result in undersensing.
      3. Maximum Heart Rate: Ensure that the value entered here is higher than the maximum heart rate expected.
         NOTE: A low value may result in undersensing, a high value may result oversensing as noisy cycles have a greater chance of getting marked as R waves.
      4. Minimum Heart Rate: Ensure that the value entered here is close to the lowest heart rate expected.
         ​NOTE: Adjust heart rate limits for each recording individually depending on the signal amplitude and the degree of noise. Researchers must be aware that a wide range of heart rates may result in failure to detect arrhythmias; a narrow range of heart rate, however, may result in extreme oversensing (e.g., thousands of episodes identified as "tachycardia", which no longer allow a meaningful analysis).
      5. Adjust Peak Bias to detect positive and negative R waves.
         NOTE: A positive Peak Bias favors detection of positive R waves; a negative Peak Bias favors detection of negative R waves.
      6. Intra Cardiac: Use this setting in cases where the P wave rapidly changes and when its derivative may exceed the derivative of the R wave resulting in false annotation of the P wave as an R wave.
      7. Baseline Recovery Threshold: Set this value, which represents a "blanking period" around the R wave, to prevent the software from searching for Q or S waves as small artefacts might otherwise result in false annotation of Q or S waves.
         ​NOTE: For example, a value of 0 will result in searching for Q/S waves from the peak of the R wave, a value of 70 will result in searching for Q/S waves only after 70% recovery of the R wave height.
   2. Click on the PT tab for settings for the detection of P and T waves.
      1. Max QT interval: Adjust this interval to define the interval at which a detected T wave will be accepted.
      2. T window from S: Adjust this setting to define the search interval for a T wave starting from S wave to the right.
      3. T window from R: Adjust this setting to define the search interval for a T wave starting from R wave to the left.
      4. P Window from R: Adjust this setting to define the search interval for a P wave starting from R wave to the left.
      5. T Direction: Set Both as default to search for both positive and negative T waves as this setting defines if only positive, only negative, or both positive/negative T waves are searched.
      6. P Direction: Set Both as default to search for both positive and negative P waves as this setting defines if only positive, only negative, or both positive/negative P waves are searched.
      7. P Placement: Adjust this setting to shift the P mark towards (high value) or away (low value) from the peak of the P wave.
      8. T Placement: Adjust this setting to shift the T mark towards (high value) or away (low value) from the peak of the P wave.
      9. Alternate End of T: Adjust this setting to search for an alternative T wave beyond the first potential T wave. Enter a lower value to select the first T wave and a higher value to select the alternative T wave.
      10. Peak Sensitivity: Adjust this parameter to eliminate small peaks when identifying P and T waves. Use this in conjunction with Peak Identification.
          NOTE: A value of 0 defines maximum sensitivity; a value of 100 defines minimal sensitivity. The minimal Peak Sensitivity value depends on the quality of the signal. If the level of noise is low and/or the P and T waves are clearly distinguishable, these waves are well triggered, even when the Peak Sensitivity is 100. Generally, the Peak Sensitivity and Peak Identification do not need adjustment unless the signal is noisy, and the analysis algorithm is encountering issues with the detection of P and T waves. If so, the best results are achieved by adjusting the parameter in steps of 25.
      11. Peak Identification: Use this parameter in conjunction with Peak Sensitivity to define the threshold for the identification of P and T waves. Lower up to 0 Peak Sensitivity if small P/T waves are not identified. If P/T waves are not identified even when Peak Sensitivity is set to 0, then lower Peak Identification, adjust in steps of 25.
      12. High ST Segment: Use this attribute if the T wave is very close to the QRS complex resulting in a high ST segment.
          NOTE: As mice lack a distinct ST segment, with a T wave occurring directly after the QRS complex, this setting should not be used in mice.
   3. Click on the Advanced Attributes tab to set low/high pass filters, to define the J point to determine ST elevation/depression (not useful in mice), to set correction factors for QT measurement, and to define arrhythmic QRS complexes by the height of the R wave and the duration of the QRS complex.
      ​NOTE: Use the default settings predefined within this tab. If the signal is affected, e.g., by electromagnetic interference, adjust the filter settings here, which may help to improve signal quality. Definition of "arrhythmic QRS complexes" does not improve the accuracy to detect premature ventricular capture beats over the method suggested here (each PVC will also result in a pause and is therefore detected by this approach). The other settings are only relevant to very specific research questions and are therefore not described in detail here.
   4. Use the Noise Tab to adjust attributes to identify noise.
      1. Click on the checkbox Enable Noise Detection to identify noise, and set Bad Data Marks.
      2. Click on the checkbox Enable Dropout Detection to set Bad Data Marks around data defined as dropout based on the maximum/minimum signal value. Adjust Min Good Data Time, which defines the time between two dropout segments also considered as dropout even if the signal is good.
      3. Adjust the Bad Data Threshold to define the level of noise above which the ECG signal cannot be properly analyzed.
         NOTE: This noisy segment of data will be included between Bad Data Marks and will not be analyzed. No ECG-derived parameters will be reported for these segments of "bad data".
      4. Specify the Min Noise Heart Rate below which heart rates are considered as noise.
   5. Use the Marks tab to turn on and off validation marks.
      ​NOTE: It is recommended to always turn on Mark Cycle Numbers, which will add a continuous number to each R wave identified. This will help to navigate through the ECG recording.
   6. Use the Notes tab to enter notes that will appear in the experimental log file.
   7. Use the Precision tab to define the precision at which parameters are reported.
   8. Set attributes and click on Recalculate to see the effects of the adjustments made in the Waveform window as a preview.
   9. If (in an ideal situation) all ECG waves are correctly annotated, click on OK to confirm the attributes settings, which opens the Effects and Scope of Changes dialog. To analyze the ECG, click on the checkboxes Reanalyze the channel and The entire channel and confirm by clicking on OK.
7. Depending on the input settings in the Attributes dialog, make a note of the validation marks that are displayed in the ECG tracing. Go through the recording manually, and check if validation marks as well as bad data marks are set correctly. Use Data Insights for checking the R marks and ECG Pro for checking P and T marks.
   1. If many marks are incorrect, modify Attributes and reanalyze the recording.
      NOTE: Specific settings can be applied to specific segments of data when the ECG morphology is different from the rest of the recording. The Ponemah software manual provides standard values for ECG Analysis Attributes for different species under Ponemah Software Manual/ Analysis Modules / Electrocardiogram / Attributes Dialog. To start with, these values can be used and then adjusted manually, until enough or (in an ideal situation) all ECG waves are marked.
   2. Perform manual clean-up if only a few marks are incorrect. Move each validation mark (except for R wave marks) to the correct position by left-clicking, holding, and moving the respective mark. Right-click within the ECG recording to add additional validation marks or mark arrhythmic R waves. Right-click on an incorrectly set mark to delete this mark.
8. Click on Actions/Logging Rate (or press F8) to set the Logging Rate, which defines how often derived data is logged to the Derived Parameter List View or plotted to graphs that use the derived parameters. For analysis of basic ECG Parameters and Arrythmia, use Epoch 1 as the standard setting, which sets logging rate to each cycle.
   ​NOTE: The Logging Rate can be augmented at any time during Acquisition or Review.

2. Analysis of basic ECG parameters

NOTE: In addition to validation/bad data marks, the software also automatically measures and calculates a large variety of derived parameters which are then reported in the Derived Parameter List.

1. Click on Subject Setup/Channel Details to select any of the derived parameters.
   NOTE: In the Derived Parameter List, each parameter is linked to the number of the respective QRS complex.
   1. Double-click on a row in the Parameter table to display the corresponding ECG cycles in the center of the primary ECG graphic window and easily find and visualize the morphology of the ECG cycles that correspond to the derived parameters in the selected raw data.
      NOTE: It is possible to synchronize in both directions: from the table to the graphic and also from the graphic to the table. When the logging rate is 1 Epoch, the synchronization is done for each individual cycle. This is easy to check from the cycle number (NUM) in the Parameters table and in the graphic. Especially in long recordings, this synchronization feature between the tables and the graphics is very useful.
2. To account for the circadian alteration in ECG parameters, rather than calculating mean values of ECG parameters over a long period, analyze a resting ECG similar to that in humans to obtain basic ECG parameters such as resting heart rate, P wave duration, PR interval, QRS duration, or QT/QTc interval. Analyze 20 consecutive QRS complexes in the mouse as "resting ECG equivalent".
   NOTE: In humans, a resting ECG is recorded over 10 s at a normal heart rate of 50-100 /min. This ECG includes 8 to 17 QRS complexes.
   1. As mice follow a circadian rhythm, analyze two resting ECGs per day, one at day time and one at night time to control for circadian effects. Select suitable times depending on the light on/off cycle in the animal facility, e.g., 12 AM/PM.
   2. Select a section of the ECG with good signal quality and stable heart rate in the HR Trend graph within a defined reasonable timeframe around this time point (e.g., ±30 min).
   3. Confirm the accuracy of the validation marks or adjust manually in 20 consecutive QRS complexes. Add missing validation marks.
   4. For further calculations and visualizations, mark the lines containing the values of these 20 consecutive QRS complexes in the Derived Parameter List, and copy to a spreadsheet or statistics software.

3. Arrhythmia detection using pattern recognition (ECG PRO module)

NOTE: Ponemah's ECG PRO module uses selected QRS complexes as templates for further analysis. The ECG patterns of the templates are compared to all QRS complexes within the recording to calculate the percentage of similarity ("match") and to recognize arrhythmias (e.g., atrial or ventricular premature capture beats). The number of QRS complexes needed to be marked depends on the variability of the QRS-amplitude within the recording. In certain cases, selecting and marking one QRS complex gives a similarity of 80 percent with the respective recording, marking the majority of QRS cycles. However, this is an ideal case and during analysis, the number of QRS complexes that need to be marked as templates is usually higher.

1. Mark QRS complexes as templates until at least a match of 80 percent or higher is achieved. Furthermore, use template matching to mark P, Q, S, and T waves if these are not or inadequately recognized after attributes settings (section 1.7).
   NOTE: R marks must be identified for cycles prior to analyzing with ECG PRO. This requires that either the R marks are preserved from acquisition or the attribute-based analysis has to be executed prior to performing ECG PRO analysis. The other marks (P, Q, S, and T) need not be present for ECG PRO analysis.
2. After completing Template setup (as described in 1.4.4), select a desired ECG wave (with marked R). If necessary, adjust the Validation Marks to accurately reflect the appropriate positions of the ECG Marks of interest. Right-click on the cycle in the Display Panel in the ECG Tracing window, select Add Cycle and Analyze [Single Template], and make a note of the cycle that appears in the Template window.
   NOTE: An Autoscale may need to be performed for both the X- and Y-axes to see the full Cycle. ECG Marks may be moved within the Template graph page.
3. Right-click on the Display Panel of the Template window, and select Add Cycle and Analyze (Single Template) to launch the Template Analysis dialog shown in Figure 4B. Select the desired Template Match Region to which all other ECG cycles will be compared. If needed, change the advanced settings for the desired Match Region.
   NOTE: Multiple Match Regions may be selected depending on the desired output from the analysis (the Derived Parameters of interest).
4. Select a Data Range on which to perform the analysis.
   NOTE: The Data Range allows the reanalysis of the data visible in the graph, the data from the left edge of the visible region from the primary graph forward to the end of the loaded data set, the data within the Parser Segments, or the entire channel.
5. Select the type of Cycles to Analyze.
   1. Select All to compare the Template Library to All cycles with a valid R mark.
   2. Select Unmatched to skip previously matched cycles and compare the Template Library to only the unmatched cycles.
      NOTE: This is useful when adding additional Templates to the Template Library for greater match coverage, as the processing time is shorter.
6. Select the desired Match Method. When selecting multiple Match Regions and Whole Cycle, use the Template that, on average, matches the cycle best to place the marks. When Region is used, for the best match for each Match Region, place the marks from different Templates.
7. Select OK to execute the analysis.
   NOTE: Additional Template Cycles can be added to the Template Library, and the Template analysis can be re-run until the desired Dialog Match % is achieved. Doing this readjusts the waves in all the cycles that match the template.
8. Save Template Libraries through Templates/Save when the Review Session is closed.
9. To detect arrhythmia using template match, tag templates that have morphology different from that of the physiological waves after doing the template match (as described in section 3.1.) by right-clicking and selecting Add Template Tag, and select a type of cycle (e.g., atrial ectopic, ventricular ectopic). Analyze these Tags using Data Insights.

4. Arrhythmia detection: a simplified manual approach using Data Insights

NOTE: For arrhythmia analysis, a correct annotation of P and R waves is necessary. However, even if clear P waves are visible within the ECG tracing, these P waves are sometimes not adequately identified even after adjusting the Attribute settings. As R waves are usually adequately recognized and annotated, a practical approach for further arrhythmia analysis using Data Insights is proposed below. For a general overview on arrhythmia detection using Data Insights and its predefined species-specific searches, the interested reader may be referred to Mehendale et al.¹³.

1. Open Data Insights by clicking on Experiment/Data Insights.
   1. Observe the Search panel at the top of the Data Insights dialog.
      NOTE: On the left of the panel, it shows which search rule is applied to which channel/subject and the number of hits using this search rule. In the middle, all the search rules are listed, and on the right, the specific definition of a selected search rule is displayed.
   2. Observe the Results panel displayed in the lower part of the Search panel.
      ​NOTE: For each search hit, the corresponding ECG section is shown (top) along with a table indicating the time within the recording and the results of each search parameter (middle).
   3. Observe the number of search hits displayed as a histogram in the bottom of the panel.
2. Given that the normal heart rate of a mouse is 500-724/min¹⁴, define a search rule bradycardia to detect bradycardia.
   1. Right-click within the search list, and select Create New Search to open the Search Entry dialog.
   2. Right-click within the white box, and select Add New Clause.
   3. Using the dropdown menus and text fields, define the search rule Bradycardia-single as Value(HR _cyc0) < 500. Click on OK to add this search rule to the list. Apply this search rule by clicking and dragging it to the channel of interest on the left.
      NOTE: The search rule Bradycardia-single identifies every individual RR interval that is longer than 120 ms (= less than 500/min.).
   4. As bradycardia requires more than one long RR interval, define an additional search rule Bradycardia as Series(Bradycardia-single, 1)>=20. Click on OK to add this search rule to the list. Apply this search rule by clicking and dragging it to the channel of interest on the left.
      NOTE: In the Results panel, each section within the ECG recording consisting of at least 20 QRS complexes with a heart rate less than 500/min. is displayed.
   5. To confirm bradycardia and to reject false results (e.g., due to R wave undersensing), review each result manually. Left-click on the waveform, and press STRG+R to reject the selected result, which will disappear from the list of results.
      NOTE: The rejected results are saved under Result/Rejects.
3. To detect tachycardia, define a search rule tachycardia.
   1. Right-click within the search list, and select Create New Search to open the Search Entry dialog.
   2. Right-click within the white box, and select Add New Clause.
   3. Using the dropdown menus and text fields, define the search rule Tachycardia-single as Value(HR _cyc0)>724. Click on OK to add this search rule to the list. Apply this search rule by clicking and dragging it to the channel of interest to the left.
      NOTE: The search rule Tachycardia-single identifies every individual RR interval that is shorter than 82 ms (= more than 724 /min).
   4. As tachycardia requires more than one short RR interval, define an additional search rule Tachycardia as Series(Tachycardia-single, 1)>=20. Click on OK to add this search rule to the list. Apply this search rule by clicking and dragging it to the channel of interest on the left.
      NOTE: The Results panel displays each section within the ECG recording consisting of at least 20 QRS complexes with a heart rate of more than 724/min.
   5. To confirm tachycardia and to reject false results (e.g., due to R wave oversensing), review each result manually. Left-click on the waveform and use the shortcut STRG+R to reject the selected result, which will disappear from the list of results.
4. To detect sinoatrial and atrioventricular blocks, define a search rule Pause.
   1. Right-click within the search list, and select Create New Search to open the Search Entry dialog.
   2. Right-click within the white panel, and select Add New Clause.
   3. Using the dropdown menus and text fields, define the search rule Pause as Value(RR-I_cyc0)>300. Click on OK to add this search rule to the list. Apply this search rule by clicking and dragging it to the channel of interest to the left.
      NOTE: The Results panel displays each section within the ECG recording with a pause of at least 300 ms.
   4. To confirm a pause, to decide if the pause is a sinoatrial or atrioventricular block, and to reject false results (e.g., due to R wave undersensing), review each result manually. Left-click on the waveform, and press STRG+R to reject the selected result, which will disappear from the list of results.
   5. To detect Ectopic Rhythm, run the template match to these rhythms first (e.g., ventricular ectopic), and then search for all the matched cycles to this template in Data Insights.
5. Right-click within the search list, and select Create New Search to open the Search Entry dialog.
   1. Right-click within the white box, and select Add New Clause.
   2. Click on Value using the dropdown menu, and select Template. On the right side, select the tag of the previously created template.
      NOTE: The Results panel displays each section within the ECG recording with the cycle matching the Template.
   3. To confirm the results and to reject false results (e.g., due to R wave undersensing), review each result manually. Left-click on the waveform and press STRG+R to reject a particular cycle, which will disappear from the list of results.
      ​NOTE: All search statements created can be imported and saved with suitable file names. All result tables can be saved and exported in spreadsheet/ASCII output format for further statistical analysis.

Recording long-term ECGs results in huge data sets. The options for further analyses are manifold and depend on the individual research project. This protocol provides a description of some very basic readouts that can be used by most researchers, especially for screening experiments, e.g., when characterizing a transgenic mouse line or when investigating the effects of a specific treatment in a disease model. A previous project involved the study of a novel drug candidate to determine whether it possessed cardiotoxic effects by analyzing ECG parameters over time. Telemetry transmitters were implanted 20 days before treatment, and ECG recordings were started 10 days before treatment to allow sufficient wound healing and acclimation of the mouse. Before treatment, the ECG was studied every three days; within the first week after treatment, the ECG was studied every day, after which the ECG was analyzed every seven days until the end of recording three weeks after treatment.

This approach allowed the detection of periods of reduced heart rate, increased atrioventricular (PR interval) and ventricular (QRS duration) conduction, as well as altered repolarization (QTc interval) in mice treated with the new drug as shown in Figure 5. This first step served as a "screening" that allowed the identification of time periods within the recording that potentially contained arrhythmias. A more detailed examination of the ECG revealed sinus pauses causing reduced heart rate two days after treatment and various degrees of atrioventricular (AV) blocks causing reduced heart rate six days after treatment. The latter finding was further supported by the prolonged PR intervals at this time point. To obtain these ECG parameters, 20 QRS complexes should be analyzed per time point and may therefore not be able to detect paroxysmal arrhythmia episodes at other time points.

To address this issue, it is advisable to specifically search for bradycardia and tachycardia episodes as well as for pauses using the ECG Pro module followed by manual review of detected episodes. This approach allows the detection of all relevant arrhythmias and the determination of the specific type of arrhythmia within the whole recording. For example, a tachycardia episode was detected in this study, which was identified as an atrial fibrillation.

As previously demonstrated, this approach further allows the determination of the time course of arrhythmia occurrence, e.g., the time to first AV block after macrophage depletion¹⁴. Representative traces, as shown in Figure 6, are obtained as described above (Figure 6A: normal sinus rhythm; Figure 6B: sinus pause; Figure 6C: AV-block I°, Figure 6D: AV-block II° type Mobitz 1; Figure 6E: AV block II° type Mobitz 2; Figure 6F: AV block III°; Figure 6G: atrial fibrillation).

Figure 1: Loading and reviewing data in Ponemah. (A) Load Review Dat dialog providing an overview of all the mice and signals recorded within the loaded experiment. (B) Graph Setup Dialog to set graphical windows providing both raw data (e.g., ECG signals) and derived parameters. Please click here to view a larger version of this figure.

Figure 2: Template setup in Ponemah. (A) Template Setup window to configure and select a new or browse already configured Template Library. (B) Graph setup page for Template settings. Please click here to view a larger version of this figure.

Figure 3: ECG tracings. (A) Screenshot of the windows containing the ECG trace; (B) heart rate plot; and (C) Template window. Please click here to view a larger version of this figure.

Figure 4: Analysis of attributes of an ECG tracing. (A) An ECG Analysis Attributes dialog. At the top of this dialog, several tabs (QRS, PT, Advanced, Noise, Marks, Notes, Precision) allow the adjustment of various settings. The settings are presented in the middle part of the dialog. At the bottom of the dialog, the ECG tracing is shown in the waveform window. At the top of the waveform window, the ECG tracing is shown; at the bottom, the derivative of the ECG tracing, including a visualization of the setting thresholds above, is shown. In the example presented here, a QRS Detection Threshold of 40% is defined, which is indicated by the pink background at the bottom. (B) Template Analysis Dialog: Select the desired Template Match Region to which all other ECG cycles will be compared. In this example, the T Wave is selected as the Match Region for analysis with a Minimum Match of 85%. This means that if the T Region does not match with at least 85% confidence, the cycle will not be marked as a match. Please click here to view a larger version of this figure.

Figure 5: Basic ECG Parameters over time in a drug intervention cohort. Blue panel: night time, yellow panel: daytime. From left to right: Heart Rate, PR interval, QRS duration, QTc interval. Please click here to view a larger version of this figure.

Figure 6: Representative ECG traces. (A) Normal sinus rhythm, (B) sinus pause, (C) AV-block I°, (D) AV-block II° type Mobitz 1, (E) AV block II° type Mobitz 2, (F) AV block III°, (G) atrial fibrillation. Scale bars = 100 ms. Abbreviation: AV = atrioventricular. Please click here to view a larger version of this figure.

Figure 7: Analysis flowchart. Abbreviation: HR = heart rate. Please click here to view a larger version of this figure.

The surface ECG is the primary diagnostic tool for patients suffering from heart rhythm disorders, providing insights into many electrophysiological phenomena. Nevertheless, sufficient analysis of cardiac surface ECG pathologies requires knowledge and definition of normal physiologic parameters. Many years of epidemiological research have led to broad consent on what is physiologic in humans and thus enabled physicians worldwide to clearly distinguish the pathologic. However, the analysis of surface ECG data is a major challenge in murine models; distinguishing between physiological and pathological ECG results can be difficult due to incomplete understanding and definition of basic ECG parameters^15,16. In 1968, Goldbarg et al. were the first to describe ECG in healthy mice¹⁷. Besides showing heart rates and basic ECG patterns, such as PR interval and QRS duration, they described major differences between anesthetized and awake animals and differences between various anesthetics and different murine breeds, which was later confirmed by other groups^16,17.

These early data emphasize why interpretation of murine ECG data is delicate and complicated. With growing interest in murine models for arrhythmia research in the past decades, more research has been focused on mouse electrophysiology and has generated evidence on the patterns of activation and repolarization in the mouse heart. The interested reader may be referred to a recent article by Boukens et al. for a detailed review of the murine ECG and its underlying currents¹⁵. Kaese et al. provided an overview on murine ECG standard values and major differences between human and murine ECG traces¹⁸. The first major difference is heart rate: healthy awake mice have a heart rate of 550-725 beats per minute, PR intervals of 30-56 ms, a QRS duration of 9-30 ms, and a repolarization phase that is very distinct from that observed in humans¹⁴. Further, the murine ECG regularly shows the occurrence of J-waves and a small and less distinctive T-wave, making analysis of the ST-segment and QT interval difficult^18,19. Overall, murine models have become the most widely used model organism for cardiovascular research, including arrhythmias⁸.

Taking into consideration the above described interspecies differences that very likely also influence arrhythmogenesis, these models can provide valuable insights. The analysis of basic ECG parameters, such as heart rate and duration of different intervals, can be reliably done using software such as Ponemah, LabChart, or ECGAuto among many others with their respective analysis algorithms. Examples for data display are shown in Figure 5. Arrhythmia detection, however, is far more delicate, and there are no widely established approaches for murine long-term ECG analysis for arrhythmias. Different approaches have been used to overcome the technical and methodological difficulties associated with arrhythmia detection of long-term ECG recordings in mice. These approaches range from only using short recordings for the manual analysis for arrhythmias²⁰ to simple considerations accepting inaccuracy as described by Thireau et al.²¹. These researchers performed heart rate variability analysis by simply excluding all sections of their recording with R-R intervals not contained in the range of the mean R-R interval ± 2 standard deviations to exclude all arrhythmias, ectopic beats, and artefacts without any manual review. This is the reason for this semimanual approach using Ponemah and its consecutive analysis modules, ECG Pro and Data Insights. This software solution can be used to analyze a vast range of physiologic signals, ranging from ECG in large mammals to blood pressure or temperature data in very small species.

The software comes with many resources on how to analyze different types of data. Nevertheless, although working quite well with ECG signals from larger animals, the low signal amplitude and therefore, high noise of signals derived from species, such as living and awake mice, can lead to a number of difficulties using a common approach to analysis. Noise will often mask P or T waves and thus disable the use of most of the predefined search rules within Data Insights. Care must be taken to define optimal values of the QRS detection threshold and to keep the attribute values used to identify QRS complexes and distinguish between clear cycles and noise events. A high percentage of the QRS detection threshold may result in undersensing (i.e., some R waves may be not detected), whereas a low percentage may result in oversensing (i.e., other peaks, such as T waves, may be misinterpreted as R waves). Further, specific questions in arrhythmia research in mice are understandably not the main topic of the materials provided by DSI, and finding specific information can be difficult. Within this protocol, a simple and pragmatic approach is used to define different arrhythmias extrapolating established human definitions.

For example, in human long-term ECG data, a pause longer than 3 s is considered significant²². This results in a human heart rate of 20/min., representing a third of the minimum physiologic heart rate of 60/min. As described by Kaese et al.¹⁸, the murine minimum physiologic heart rate equals 550/min., making 200/min. approximately a third of that rate. According to the human definition, pauses of more than 0.3 s can be assumed to be significant in mice. Further, it is a simple and pragmatic approach to describe differences in baseline parameters as relative changes to the respective control. This takes into consideration the differences between individual mouse lines and is an elegant way to identify the probable pathologic without relying on (often lacking) established normal values. This simple approach, summarized in Figure 7, is suitable for all groups studying cardiac arrhythmias in murine models using implantable telemetry devices. It leads to the evaluation of general ECG parameters as well as data on heart rate over time and the detection of a wide variety of arrhythmias. Therefore, this article attempts to provide a step-by-step approach for ECG and arrhythmia analysis and adds significantly to the guidance and manuals that have already been published.