I have a numpy array that has times in the first column and the other columns contain signal values. The objective is to use a peak finding algorithm for all the signal value columns and get the peaks and their corresponding times. My code looks like this:

import numpy as np
from peakutils import indexes

arr = stress_data.to_numpy()
times_all = arr[:,0]
peaks_all = []
width = 125

for i in range(1,np.shape(arr)[1]):
    x = arr[:,i]
    xf = x - np.mean(x)    
    threshold = 0.1*np.average(xf) / np.max(xf)
   
    # Find x-coordinates of peaks in signal
    peaks = indexes(xf, thres = threshold, min_dist = width)
        
    sg = [times_all[peaks], xf[peaks]]
    peaks_all.append(sg)

This code works as expected but the problem is that it takes a long time to run.
I tried optimizing this with the np.apply_along_axis and defining a processing function for it but doing it that way took even longer. Here is the code I used for that:

def process_column(x):
        xf = x - np.mean(x)
        threshold = 0.1 * np.average(xf) / np.min(xf)
        valleys = indexes(xf, thres=threshold, min_dist=width)
        return [times_all[valleys], xf[valleys]]

    valleys_all = np.apply_along_axis(process_column, axis=0, arr=data_all.T)

About the numpy array: Besides the time column, there are 10 columns to be processed and there are 10.8 million rows. Any help wiyh optimizing this would be appreciated