Issues with blink preprocessing

[acooperr1]

Hi there, I am new to analyzing pupil data and am running into issues in the preprocessing of the data. Specifically, I am having difficulty properly detecting blinks in my data. I have tried a lot of of combinations and haven't found any of them to be completely accurate for all of my subjects.

Below is my script:

#! /usr/bin/env python3
import sys
import os
import pypillometry as pp
import numpy as np
import matplotlib.pyplot as plt
from matplotlib.backends.backend_pdf import PdfPages

#File containing a list of subjs
input_file = "/Users/acoop/Dropbox/2023-fMRI/scripts/subj_list.txt"

#create lists to store subjects and scan numbers
subjects = []
scan_numbers = []

# Open the input file
with open(input_file, "r") as file:
    # Loop over each line in the file
    for line in file:
        # get subj name and scan number
        subj, scan_num = line.split()
        subjects.append(subj)
        scan_numbers.append(scan_num)

#create a pdf file to save plots 
pdf_file = "/Users/acoop/Desktop/subj_plots.pdf"

#create a PDF object 
pdf = PdfPages(pdf_file)

for subj, scan_num in zip(subjects, scan_numbers):
# Set data directories
    edf_dir = f"/Users/acoop/Dropbox/2023-fMRI/ParticipantFiles/{subj}/Oddball/{scan_num}"

# Loop through each file in the folder
    for filename in os.listdir(edf_dir):
    # Check if the file ends with .EDF
        if filename.endswith(".EDF"):
            prefix = filename

            # declare file name 
            pd_file = os.path.join(edf_dir, f"{prefix}.pd")
            print(f"Directories it's pulling from: {pd_file}")

#load picklefile, apply a 3Hz low-pass filter to it, downsample the signal to 50 Hz, detect blinks in the signal and merge short, successive blinks together. Store it all in object d
            d=pp.PupilData.from_file(pd_file)\
               .blinks_detect()\
               .blinks_merge()\
               .lowpass_filter(3)\
               .downsample(50)

            default = d.get_erpd("default", "DEFAULT_CUE")
            oddball = d.get_erpd("oddball", "ODDBALL_CUE")

            plt.figure(figsize=(15, 5))
            d.plot()
            # Add title with subject and scan number
            plt.title(f"Subject: {subj}, Scan Number: {scan_num}")
            pdf.savefig(plt.gcf())
            plt.close()

#get summary of the dataset and the operations applied to it 
            print(d)
#return information in the form of a dict
            d.summary()

pdf.close()
print(f"All plots have been saved to {pdf_file}")

Here are some of the various preprocessing options I've tried:

For .blinks_detect:
-min_duration=10
-min_offset_len=2
-min_onset_len=3
-vel_onset=-5
-vel_offset=5

For .blinks_merge:

• distance=130

For .lowpass_filter:
-3
-cutoff=5
-5

for .downsample:
-50
-100

Each time I have run this script with the various different options it has incorrectly identified the blinks by either identifying too many or not identifying all of them. I have also been unable to interpolate all of my subjects and run into errors after the first subject. The interpolate command I have been using is: d = d.blinks_interp_mahot()

Here are some plots of the preprocessed blink data with the default parameters set:

If you have any guesses on what the issue could be, please let me know, thank you!