Find Peaks in PPG/ECG data

[paulogil2010]

Hi everyone,

I'm trying to detect the max peaks in an ECG/PPG sample with findpeaks function from pracma package, but the results aren't as expected, not sure if I missing filters or the function isn't for that. The function returns some other peaks in the middle and in the bottom of the wave.

Example code:

data <- read.csv("wave_vector.csv")
wave_vector <- purrr::flatten_dbl(data)
plot(wave_vector,type='l') 
points(peaks,wave_vector[peaks])

peaks <- findpeaks(wave_vector, minpeakdistance = 100)
plot(wave_vector[peaks])
plot(wave_vector,type='l')
points(peaks,wave_vector[peaks])

My wave_vector is a vector of PPG or ECG signal and can be find on GitHub

Any help will be appreciated,

[jcblum]

Hi @paulogil2010! Welcome!

Thanks for providing the data, which made it easy for me to see what your problem was

So there are two things going awry in the code you posted — but the good news is that the findpeaks() part is working just fine!

The first issue is with how the CSV file gets imported. Note that I'm assuming here that you are importing the exact same CSV file you posted online. If that's not the case, this part might not apply to you.

library(pracma)

# Btw, here's an example of one way to write a CSV file import when constructing a 
# self-contained reproducible example that somebody else can easily run 
data_url <- "https://raw.githubusercontent.com/paulogil2010/hello-world/master/wave_vector.csv"
data_path <- file.path(tempdir(), "wave_vector.csv")
download.file(data_url, data_path)

data <- read.csv(data_path)
str(data)
#> 'data.frame':    1000 obs. of  2 variables:
#>  $ X: int  1 2 3 4 5 6 7 8 9 10 ...
#>  $ x: num  290 309 329 351 372 ...

There are two columns here, and one of them has been (confusingly!) given the automatic name of capital X. Where did this X column come from? Here are the first three lines of the CSV file:

cat(readLines(data_path, n = 3), sep = "\n")
#> "","x"
#> "1",289.8887
#> "2",308.60833

The file has row names, but read.csv() is interpreting them as a column of data that is missing a name. (To force them to be read in as row names in this case, you'd use read.csv("wave_vector.csv", row.names = 1)).

Flattening a data frame with two columns results in the column of row names being appended to the beginning of the column of data, which is not what you want!

str(purrr::flatten_dbl(data))
#>  num [1:2000] 1 2 3 4 5 6 7 8 9 10 ...

But flattening is overkill here — you can just pull the vector out of the data frame using dollar-sign or double-bracket indexing (or dplyr offers the pull() function to do the same thing).

wave_vector <- data$x
str(wave_vector)
#>  num [1:1000] 290 309 329 351 372 ...

# Finally! Find some peaks! :-)
peaks <- findpeaks(wave_vector, minpeakdistance = 100)

Now we encounter problem #2: peaks is a matrix, with column definitions given in the documentation:

The first column gives the height, the second the position/index where the maximum is reached, the third and fourth the indices of where the peak begins and ends — in the sense of where the pattern starts and ends.

peaks
#>           [,1] [,2] [,3] [,4]
#>  [1,] 486.8127  977  955  997
#>  [2,] 481.9355   15    1   41
#>  [3,] 450.8867  135  116  159
#>  [4,] 448.6920  858  838  882
#>  [5,] 448.6058  375  356  399
#>  [6,] 448.1287  495  475  519
#>  [7,] 447.9364  617  598  642
#>  [8,] 446.8559  255  235  279
#>  [9,] 444.3782  739  719  763

So to get the plot I think you’re looking for, you need to grab the values from the second column of peaks

plot(wave_vector, type = 'l')
points(peaks[, 2], wave_vector[peaks[, 2]])

^{Created on 2018-11-01 by the reprex package (v0.2.1)}

[paulogil2010]

@jcblum

plot(wave_vector, type = 'l') points(peaks[, 2], wave_vector[peaks[, 2]])

Thanks you very much @jcblum,

actually, to be easier, I have exported the variable to a csv file. I have filtered this variable with other libraries and functions previously. So probably when I exported it, something may have gone wrong. I didn't post all code to not be so extended.

I will try this in my code. Sorry again by the confusion, and I'm glad with your answer.

Thanks a million

[jcblum]

Glad to help!

I suspected you might be accessing your data in a different fashion than in the example. For the future, I strongly recommend learning how to use the reprex package to format reproducible examples for posting in forums like this one (I actually wrote almost that entire post in an R script file and reprex::reprex() turned it into what you see above, including the image upload ). By including the output along with the code and enforcing a decent amount of self-contained-ness, reprex makes it much easier to focus your helpers on the most important problem.