This article goes through the steps/process of performing arrhythmia analysis through the Data Insights tool. A DSI webinar was also done on how to efficiently process your ECG data with Data Insights and can be found by clicking the following link: Advancements in Research Virtual Discussion Data Insights
1. Determine definitions of the arrhythmias in study protocol
- Will you look at all PAC findings or only when P is buried in the preceding T wave
- QRS widening for V Ectopic, will this be determined by a given msec or by looking at the morphology
2. Attribute Analysis
- Increase MIN HR
- Noise will want to be used, but with an understanding that it may remove arrhythmias in high noise level areas
- For more information on Attribute Analysis, please see the following article: Video Tutorial - Ponemah ECG Analysis for v5.3x and v6.x
3. Use Data Insights searches to check your attribute analysis (QC step)
- ar_PAC (works really well to find places where R may be marked on T wave or P wave
- use “dv” searches (dv=data validation) to ensure proper placement of R waves
- HR_Missed Beats
- HR Limits (sort by HR value to look at the extreme ends)
- HR change
4. Run ECG PRO Template Analysis
- Template Analysis - P, Q, and S waves at 80%
- For more information on Template Analysis, please see the following article: Video: ECG PRO Template Analysis
Use Data Insights (Unmatched clean search) with Template analysis
- Use DI search: an_Unmatched Cleans
- Distribution of the unmatched clean (are there any areas not being represented)
- Are there arrhythmias present in the unmatched clean, dv_HR Missed Beats, dv_HR Limits, dv_HR change?
- Can remove noisy waveforms by adding bad data mark OR adjusting the noise setting in attribute analysis
5. Get any unmatched P, Q, and S waves
- Delete marks from all unmatched cycles
- Analyze with just Q and S at 80% (for V. Ectopics) for unmatched cycles to get any missing Q or S waves
- Analyze with just P and Q and/or P and S (for PR interval, 1st deg AV HB) for unmatched cycles to get any missing P waves
6. Search using the DI searches beginning with “ar_” (ar=arrhythmia)
For more information on using Data Insights for Arrhythmia Detection, please see the following article: Video Tutorial - Using Data Insights for Arrhythmia Detection
6.1 Review AB Block 1st
- Check to see if P is placed on Ts and fix
6.2 Review V Ectopics
- Reject and/or Add BDMs
6.3 Review PAC
- Added 3 clause: not ar_Junctional nor V Ectopic (otherwise it can be listed in both) and PCtcyc0 = 0
- Sort by %Decrease and start at the maximum numbers
6.4 Review Junctional Beats (includes premature and escape beats)
- The junctional search relies on the logging rate averages, therefore set the logging rate appropriately. Test datasets that have been used by the Data Services team determined a logging rate of 30 seconds worked effectively with their specific datasets.
- Note P, Q, and S wave match must be performed in order for PCt to be equal to zero instead of NaN for ar_junctional search
- If PACs are found it may be listed in Junctional beat by default unless clause in PAC search prevents this; if clause is added in PAC to not be junctional, if it meets the criteria of PAC’s RR interval changes and it’s rejected, then it will go to the PAC search clause
7. Isolated P waves
Change the template Specific Attribute for the templates to search for isolated P waves by turning on “Search Window (ms)” and “Offset (ms)”.
Re-do Template Analysis for P and Q waves (will only find iso P waves). If there are 2 or more isolated P waves without corresponding QRS, only 1 of the isolated P waves will be found.
- Sort by PR Value and look at the low end and move to the longer PR values.
8. Notes about False Positives
The Data Insights default searches are very conservative such that it errors on having more false positive/less false negatives. The number of false positives are depenedent on the following factors:
- Noise levels
- P wave amplitudes for junctional searches and determining premature atrial complexes versus premature junctional complexes.
- Matching of template library