The following is a paper on using Kubios HRV Analysis Software for viewing HRV Data from the J&J I330 – by Christina Huang – California School of Professional Psychology, San Diego, CA
Protocol for Exporting J&J Data
1. After recording your session, exit out of J&J.
2. A small window will pop up in the middle of the screen asking you to select and name a client.
3. Click “Save to Database” and then “Export.”
4. A new window will pop up. On the top left, click on “HR/IBI” and make sure to double check where your file will be saved by clicking on the drop down menu. Also double check all the signals that will be exported in the bottom right hand side of the window.
5. Click ok. Your data should be exported in IBI format to the file you have chosen to save it in.
Protocol for HRV Biofeedback manual cleanup using Kubios HRV
1. Open Kubios HRV software, then open your data through this program in the File dropdown menu.
2. At the same time, open up the actual notepad or excel file that you are cleaning and position the window next to the Finnish program.
3. When you’ve opened the file in the Finnish program, it will show you what it actually looks like on the top right hand portion of the screen in the form of a graph with Time as the X-axis and RR(s) as the Y-axis. Each data point (blue dots in the picture) corresponds to each # in the Notepad or excel file and represents one second of data. The picture will help you to be able to see visually which data points might be ectopic. If the graph looks too hectic and you have trouble being able to see any trends, you can extend the range. There is a box directly under the graph to the very right labeled “Range(s).” To spread out the data, manually type in a smaller # here.
4. To the left hand side of the screen approximately in the middle, there is a box labeled “Remove trend components.” In the Methods dropdown menu, change the option from “None” to “Smoothn Priors.” This will add a curve to the graph by averaging out the center of the data points.
5. Unfortunately, the criteria for deciding what data points to clean is arbitrary. You just have to eyeball the data and see which data points look abnormal and which represent actual significant trends. For example, if the data points gradually go up, then you probably won’t need to delete any, but if there is a small cluster of data points that are far removed from everything else that do not appear to be part of a trend, it would be best to delete them. Remember, it’s better to average out any points that might be ectopic than to leave them as is. Methodologically, it’s probably a good idea to note which sets of data you have removed and to offer an explanation via your observations during the recording session. Ectopic beats can be caused by anything, such as the participant laughing, coughing, moving their hand, changing seating position, etc…
If your data has specific subsections in it (i.e. baseline, paced, stressor, recovery, etc…) and you want to examine each section separately but have not already manually separated them into discrete files, you can manipulate the graph by extending or narrowing the highlighted region to reflect the length of time of these segments. Double click on the graph until a compass appears, then drag the edge of the highlighted portion to correspond with the time segments you have. For example, if you are only interested in examining the paced breathing section which occurs from minutes 5-8, drag the highlighted portion to highlight only these minutes.
The statistics in the report will give you numbers calculated based off the highlighted regions only.
6. Using the picture to gauge which points you’ll need to delete, find these same beats in the notepad or excel file. This step is extremely tedious and the data point may be difficult to find. If you have trouble finding it, you can either use the seconds on the x-axis of the picture in the Finnish program or look at the #s in the file. For example, in the notepad or excel file, if all of the #s are in the 600 range and then you all of a sudden see a # that is 750, it’s probably the one you want to delete. Remember again, each of the data points you see in the Kubios graph represents each of the #s you have in IBI format and correspond by time as well, so if you are specifically looking at an ectopic beat that occurs at the 1 minute marker in Kubios, you can scroll 60 #s down in your data file to find that #.
7. Once you’ve found the data point you want to take out, go to the notepad/excel file and delete it. Replace it with the average of the 2 #s on either side of it, or in other words, the average of the #s directly preceding and following it.
8. Each time you’ve done this, resave the notepad or excel file and reopen it using the Finnish software to see what the new file looks like. This way, you can keep track of the changes you’ve made and not have to repeat any of your work in case something happens.
9. Continue to do this for each and every single file.
10. When you put the data into SPSS, go to the toolbar of the Finnish program and go to “report sheet” under “View” in the top left hand corner.
11. In the Report, look at the left column of data that’s labeled “Time Domain Results.” Under “Statistical Measures,” you’ll see:
-Mean RR*
-STD RR (SDNN)
-Mean HR*
–STD HR
-RMSSD
–NN50
-pNN50
-RR Triangular Index
-TINN
You want to take note of everything but STD of mean HR and NN50 (underlined above).
12. In the bottom left column you’ll see “Frequency Domain Results” that are faintly highlighted blue. Of the 4 columns (peak; power: ms2; power: %; power: n.u.), take notes of the last 2 power columns (% and n.u.) for all bands: VLF, LF, HF, and LF/HF.
13. Still in the Frequency Domain on the right column, you’ll see the parametric spectrum labeled AR Spectrum for Auto-Regressive Peak. Take note of the highest first peak in this graph and its PSD value.
14. Under the Nonlinear Results at the bottom of the report, take note of the SD1 and SD2 under the “Poincare Plot” as well as the Approximate Entropy (ApEn) under the “Other” table. The Approximate Entropy tells you how regular the signal is.
*Each of these measures will be a separate column in SPSS.
Protocol for HRV Biofeedback automatic cleanup using Kubios HRV:
1. Open Kubios HRV software, then open your data through this program in the File dropdown menu.
2. At the same time, open up the actual notepad or excel file that you are cleaning and position the window next to the Finnish program.
3. When you’ve opened the file in the Finnish program, it will show you what it actually looks like on the top right hand portion of the screen in the form of a graph with Time as the X-axis and RR(s) as the Y-axis. Each data point (blue dots in the picture) corresponds to each # in the Notepad or excel file. The picture will help you to be able to see visually which data points might be ectopic. If the graph looks too hectic and you have trouble being able to see any trends, you can extend the range. There is a box directly under the graph to the very right labeled “Range(s).” To spread out the data, manually type in a smaller # here.
4. To the left hand side of the screen approximately in the middle, there is a box labeled “Remove trend components.” In the Methods dropdown menu, change the option from “None” to “Smoothn Priors.” This function averages out your data by finding trends and smoothing out the data points. Graphically, it will add a curve to the graph.
5. If your data has specific subsections in it (i.e. baseline, paced, stressor, recovery, etc…) and you want to examine each section separately but have not already manually separated them into discrete files, you can manipulate the graph by extending or narrowing the highlighted region. Double click on the graph until a compass appears, then drag the edge of the highlighted portion to correspond with the time segments you have. For example, if you are only interested in examining the paced breathing section which occurs from minutes 5-8, drag the highlighted portion to highlight only these minutes.
The statistics in the report will give you numbers calculated based off the highlighted regions only.
6. To the left hand side of the screen above the “Smoothn Priors” area, you will see a section labeled “Auto Correction” with a drop down menu allowing you to manually choose the degree of correction you would like the computer to apply to your data. The degrees of correction range from “very low” to “strong.” This function basically corrects your data for outliers and reflects your data with those outliers averaged out. The amount of auto correction you want to use will vary and depend on each segment of data you have.
Make sure that you document the amount of correction you have decided to use for each piece of data and that you mention this in your methods section.
7. When you put the data into SPSS, go to the toolbar of the Finnish program and go to “report sheet” under “View” in the top left hand corner.
8. In the Report, look at the left column of data that’s labeled “Time Domain Results.” Under “Statistical Measures,” you’ll see:
-Mean RR*
-STD RR (SDNN)
-Mean HR*
–STD HR
-RMSSD
–NN50
-pNN50
-RR Triangular Index
-TINN
You want to take note of everything but STD of mean HR and NN50 (underlined above).
9. In the bottom left column you’ll see “Frequency Domain Results” that are faintly highlighted blue. Of the 4 columns (peak; power: ms2; power: %; power: n.u.), take notes of the last 2 power columns (% and n.u.) for all bands: VLF, LF, HF, and LF/HF.
10. Still in the Frequency Domain on the right column, you’ll see the parametric spectrum labeled AR Spectrum for Auto-Regressive Peak. Take note of the highest first peak in this graph and its PSD value.
11. Under the Nonlinear Results at the bottom of the report, take note of the SD1 and SD2 under the “Poincare Plot” as well as the Approximate Entropy (ApEn) under the “Other” table. The Approximate Entropy tells you how regular the signal is.
*Each of these measures will be a separate column in SPSS.
Explanation of Kubios reports statistics
I. Measures of HRV (Time-Domain Results)
a. SDNN: standard deviation of normal to normal peaks
i. SDNN is said to represent all the cyclic components responsible for variability in the recording period
b. RMSSD: root mean square of successive differences
c. pNN50: percentage of instances in which 2 consecutive R-R intervals differ by more than 50 msecs
i. pNN50 is more useful than the NN50 because the NN50 just tells you how many R-R intervals are greater than 50 msecs. However, without the percentage of how many intervals meet this criterium, the NN50 really becomes just an arbitrary number that can’t be compared to anything
ii. pNN50 also is said to reflect strong vagal modulation without being dependent on homeostatic changes or length of recording period
*are all different measures calculated in slightly different ways that assess the same construct of heart rate variability. The actual ways these 3 statistics are calculated are quite complicated but it’s good to have all of them as extra checks.
*R-R interval or interbeat interval is the heartbeat from R peak to R peak (if you look at an ECG and the heart beat is measured via pQRST). In other words, the RR interval represents the period of time in between each QRS complex)
II. The other measures under Time-Domain Results are unnecessary. The RR triangular index and TINN are complicated statistics that involve some crazy computations, so don’t bother.
III. Frequency-Domain Results (FFT Spectrum)
a. LF/HF ratio: the only number for LF/HF ratio is under the Power (ms2) column. The people at Kubios just didn’t feel like reporting it multiple times but this is the only number you need even though we generally don’t use the rest of the numbers under the Power (ms2) column. There’s new evidence now that if you take the natural log (ln) of LF and HF, it will give you vagal tone.
b. Power (%) vs Power (n.u.; “normal unit”): the difference is how they are calculated. Power % = HF/(HF+LF+VLF) whereas Power n.u. = HF/(HF+LF). That is, Power % takes into account the VLF as well. We take both into account but the only important difference seems to lie with individual preference. Different big names in the field prefer different numbers so just report both of them to please everyone.
c. No skin conductance is reported here simply because Kubios doesn’t do it. They just didn’t take it into account. If you really want to report SC, refer back to your actual numbers in your excel files. Otherwise, SC implies autonomic functioning which is implied in your HF, LF, and VLF ratios, which compare PNS to SNS activity. Up to you!
d. Frequency Bands (just as a review…)
i. VLF: represents parasympathetic nervous system (PNS) activity
ii. LF: blend of SNS + parasympathetic nervous system (PNS) activity
iii. HF: PNS activity
IV. Frequency-Domain Results (AR Spectrum)
a. PSD = power spectral density. This is HRV as it varies over time in which the integral (yes as in calculus) of the beats are calculated and shown visually via a graph. The graph shows you high and low frequency power values so that a balance between PNS and SNS functioning can be derived via the PSD s2/Hz. The “smoothin priors” function you clicked on when reporting the data basically takes the graph from the FFT spectrum and smoothes out the irregular parts of the graph so it becomes all nice and pretty and easy to interpret in the AR spectrum.
V. Nonlinear Results
a. SD1 vs SD2 in Poincare Plot: This involves graphing the data points on different axes and giving you the standard deviations of both of those axes. The x-axis (RRn(s)) is just the R-R interval (refer to part I of this section).
For questions or concerns, contact Christina Huang: or Richard Gevirtz, PhD: