Add multiple installation options

eegnb/analysis/pipelines.py

@@ -37,7 +37,7 @@
 
 # EEG-Notebooks functions
 from eegnb import generate_save_fn
-from eegnb.analysis.utils import load_data,plot_conditions, load_csv_as_raw, fix_musemissinglines
+from eegnb.analysis.analysis_utils import load_data,plot_conditions, load_csv_as_raw, fix_musemissinglines
 from eegnb.analysis.analysis_report import get_html
 from eegnb.datasets import fetch_dataset
 from eegnb.devices.utils import EEG_INDICES, SAMPLE_FREQS

eegnb/analysis/streaming_utils.py

@@ -0,0 +1,204 @@
+import copy
+from copy import deepcopy
+import math
+import logging
+import sys
+from collections import OrderedDict
+from glob import glob
+from typing import Union, List
+from time import sleep, time
+import keyboard
+import os
+
+import pandas as pd
+import numpy as np
+import matplotlib.pyplot as plt
+import seaborn as sns
+from mne import create_info, concatenate_raws
+from mne.io import RawArray
+from mne.channels import make_standard_montage
+from mne.filter import create_filter
+from matplotlib import pyplot as plt
+from scipy import stats
+from scipy.signal import lfilter, lfilter_zi
+
+from eegnb import _get_recording_dir
+from eegnb.devices.eeg import EEG
+#from eegnb.devices.utils import EEG_INDICES, SAMPLE_FREQS
+
+# this should probably not be done here
+sns.set_context("talk")
+sns.set_style("white")
+
+
+logger = logging.getLogger(__name__)
+
+
+
+# Bjareholt Tools
+# ==================
+# From    https://github.com/ErikBjare/thesis/blob/master/src/eegclassify/clean.py
+# ------
+
+
+def channel_filter(
+    X: np.ndarray,
+    n_chans: int,
+    sfreq: int,
+    device_backend: str,
+    device_name: str,
+    low: float = 3,
+    high: float = 40,
+    verbose: bool = False,
+) -> np.ndarray:
+    """Inspired by viewer_v2.py in muse-lsl"""
+    if device_backend == "muselsl":
+        pass
+    elif device_backend == "brainflow":
+        if 'muse' not in device_name: # hacky; muse brainflow devices do in fact seem to be in correct units
+            X = X / 1000 # adjust scale of readings
+    else:
+        raise ValueError(f"Unknown backend {device_backend}")
+
+    window = 10
+    n_samples = int(sfreq * window)
+    data_f = np.zeros((n_samples, n_chans))
+
+    af = [1.0]
+    bf = create_filter(data_f.T, sfreq, low, high, method="fir", verbose=verbose)
+
+    zi = lfilter_zi(bf, af)
+    filt_state = np.tile(zi, (n_chans, 1)).transpose()
+    filt_samples, filt_state = lfilter(bf, af, X, axis=0, zi=filt_state)
+
+    return filt_samples
+
+
+
+
+
+def check(eeg: EEG, n_samples=256) -> pd.Series:
+    """
+    Usage:
+    ------
+
+    from eegnb.devices.eeg import EEG
+    from eegnb.analysis.utils import check
+    eeg = EEG(device='museS')
+    check(eeg, n_samples=256)
+
+    """
+
+    df = eeg.get_recent(n_samples=n_samples)
+
+    # seems to be necessary to give brainflow cnxn time to settle
+    if len(df) != n_samples: 
+      sleep(10) 
+      df = eeg.get_recent(n_samples=n_samples) 
+
+    assert len(df) == n_samples
+
+    n_channels = eeg.n_channels
+    sfreq = eeg.sfreq
+    device_backend = eeg.backend
+    device_name = eeg.device_name
+
+    vals = df.values[:, :n_channels]
+    df.values[:, :n_channels] = channel_filter(vals, 
+                                    n_channels,
+                                    sfreq,
+                                    device_backend,
+                                    device_name)
+
+    std_series = df.std(axis=0)
+
+    return std_series
+
+
+
+def check_report(eeg: EEG, n_times: int=60, pause_time=5, thres_std_low=None, thres_std_high=None, n_goods=2,n_inarow=5):
+    """
+    Usage:
+    ------
+    from eegnb.devices.eeg import EEG
+    from eegnb.analysis.utils import check_report
+    eeg = EEG(device='museS')
+    check_report(eeg)
+
+    The thres_std_low & thres_std_high values are the 
+    lower and  upper bound of accepted
+    standard deviation for a quality recording.
+
+    thresholds = {
+        bad: 15,
+        good: 10,
+        great: 1.5 // Below 1 usually indicates not connected to anything
+    }
+    """
+
+    # If no upper and lower std thresholds set in function call,
+    # set thresholds based on the following per-device name defaults
+    edn = eeg.device_name
+    flag = False
+    if thres_std_low is None:
+        if edn in thres_stds.keys():
+            thres_std_low = thres_stds[edn][0]
+    if thres_std_high is None:
+        if edn in thres_stds.keys():
+            thres_std_high = thres_stds[edn][1]
+
+    print("\n\nRunning signal quality check...")
+    print(f"Accepting threshold stdev between: {thres_std_low} - {thres_std_high}")
+
+    CHECKMARK = "√"
+    CROSS = "x"
+
+    print(f"running check (up to) {n_times} times, with {pause_time}-second windows")
+    print(f"will stop after {n_goods} good check results in a row")
+
+    good_count=0
+
+    n_samples = int(pause_time*eeg.sfreq)
+
+    sleep(5)
+
+    for loop_index in range(n_times):
+        print(f'\n\n\n{loop_index+1}/{n_times}')
+        std_series = check(eeg, n_samples=n_samples)
+
+        indicators = "\n".join(
+        [
+            f"  {k:>4}: {CHECKMARK if v >= thres_std_low and v <= thres_std_high else CROSS}  (std: {round(v, 1):>5})"
+                for k, v in std_series.iteritems()
+        ]
+                              )
+        print("\nSignal quality:")
+        print(indicators)
+
+        bad_channels = [k for k, v in std_series.iteritems() if v < thres_std_low or v > thres_std_high ]
+        if bad_channels:
+            print(f"Bad channels: {', '.join(bad_channels)}")
+            good_count=0  # reset good checks count if there are any bad chans
+        else:
+            print('No bad channels')
+            good_count+=1
+
+        if good_count==n_goods:
+            print("\n\n\nAll good! You can proceed on to data collection :) ")
+            break
+
+        # after every n_inarow trials ask user if they want to cancel or continue
+        if (loop_index+1) % n_inarow == 0:
+            print(f"\n\nLooks like you still have {len(bad_channels)} bad channels after {loop_index+1} tries\n")
+
+            prompt_time = time()
+            print(f"Starting next cycle in 5 seconds, press C and enter to cancel")    
+            while time() < prompt_time + 5:  
+                if keyboard.is_pressed('c'): 
+                    print("\nStopping signal quality checks!")
+                    flag = True
+                    break  
+        if flag: 
+            break  
+
+

examples/visual_cueing/01r__cueing_singlesub_analysis.py

@@ -26,7 +26,7 @@
 from mne.time_frequency import tfr_morlet
 
 # EEG-Notebooks functions
-from eegnb.analysis.utils import load_data,plot_conditions
+from eegnb.analysis.analysis_utils import load_data,plot_conditions
 from eegnb.datasets import fetch_dataset
 
 # sphinx_gallery_thumbnail_number = 1  

examples/visual_cueing/02r__cueing_group_analysis.py

@@ -22,7 +22,7 @@
 
 # EEG-Notebooks functions
 from eegnb.datasets import datasets
-from eegnb.analysis.utils import load_data
+from eegnb.analysis.analysis_utils import load_data
 
 # sphinx_gallery_thumbnail_number = 1
 

examples/visual_cueing/04r__cueing_group_analysis_winter2019.py

@@ -22,7 +22,7 @@
 from mne.time_frequency import tfr_morlet
 
 # EEG-Noteooks functions
-from eegnb.analysis.utils import load_data
+from eegnb.analysis.analysis_utils import load_data
 from eegnb.datasets import fetch_dataset
 
 # sphinx_gallery_thumbnail_number = 1

examples/visual_n170/01r__n170_viz.py

@@ -33,7 +33,7 @@
 from mne import Epochs,find_events
 
 # EEG-Notebooks functions
-from eegnb.analysis.utils import load_data,plot_conditions
+from eegnb.analysis.analysis_utils import load_data,plot_conditions
 from eegnb.datasets import fetch_dataset
 
 # sphinx_gallery_thumbnail_number = 3

examples/visual_n170/02r__n170_decoding.py

@@ -26,7 +26,7 @@
 from mne.decoding import Vectorizer
 
 # EEG-Notebooks functions
-from eegnb.analysis.utils import load_data
+from eegnb.analysis.analysis_utils import load_data
 from eegnb.datasets import fetch_dataset
 
 # Scikit-learn and Pyriemann ML functionalities

examples/visual_p300/01r__p300_viz.py

@@ -30,7 +30,7 @@
 from mne import Epochs,find_events
 
 # EEG-Notebooks functions
-from eegnb.analysis.utils import load_data,plot_conditions
+from eegnb.analysis.analysis_utils import load_data,plot_conditions
 from eegnb.datasets import fetch_dataset
 
 # sphinx_gallery_thumbnail_number = 3

examples/visual_p300/02r__p300_decoding.py

@@ -26,7 +26,7 @@
 from mne.decoding import Vectorizer
 
 # EEG-Notebooks functions
-from eegnb.analysis.utils import load_data
+from eegnb.analysis.analysis_utils import load_data
 from eegnb.datasets import fetch_dataset
 
 # Scikit-learn and Pyriemann ML functionalities

examples/visual_ssvep/01r__ssvep_viz.py

@@ -29,7 +29,7 @@
 from mne.time_frequency import tfr_morlet
 
 # EEG-Notebooks functions
-from eegnb.analysis.utils import load_data,plot_conditions
+from eegnb.analysis.analysis_utils import load_data,plot_conditions
 from eegnb.datasets import fetch_dataset
 
 # sphinx_gallery_thumbnail_number = 3

examples/visual_ssvep/02r__ssvep_decoding.py

@@ -27,7 +27,7 @@
 from mne.decoding import Vectorizer
 
 # EEG-Notebooks functions
-from eegnb.analysis.utils import load_data
+from eegnb.analysis.analysis_utils import load_data
 from eegnb.datasets import fetch_dataset
 
 # Scikit-learn and Pyriemann ML functionalities

requirements.txt

@@ -1,18 +1,52 @@
-# Main repo requirements
-psychopy==2023.1.0
-psychtoolbox
+
+## ~~ Analysis Requirements ~~ 
+
 scikit-learn>=0.23.2
 pandas>=1.1.4
 numpy>=1.19.4,<1.24  # due to outdated libs not changing the names after: https://github.com/numpy/numpy/pull/22607
 mne>=0.20.8
 seaborn>=0.11.0
 pyriemann>=0.2.7
 jupyter
+gdown>=4.5.1
+matplotlib>=3.3.3
+pysocks>=1.7.1
+pyserial>=3.5
+h5py>=3.1.0
+pytest-shutil
+pyo>=1.0.3; platform_system == "Linux"
+keyboard==0.13.5
+airium>=0.1.0
+attrdict>=2.0.1
+attrdict3
+
+
+## ~~ Streaming Requirements ~~ 
+
 muselsl>=2.0.2
 pylsl==1.10.5  # due to https://github.com/NeuroTechX/eeg-notebooks/issues/187
 brainflow>=4.8.2
-gdown>=4.5.1
-matplotlib>=3.3.3
+pysocks>=1.7.1
+pyserial>=3.5
+h5py>=3.1.0
+pytest-shutil
+pyo>=1.0.3; platform_system == "Linux"
+keyboard==0.13.5
+airium>=0.1.0
+attrdict>=2.0.1
+attrdict3
+click
+
+
+## ~~ Stimpres Requirements ~~ 
+
+psychopy==2023.1.0
+psychtoolbox
+scikit-learn>=0.23.2
+pandas>=1.1.4
+numpy>=1.19.4,<1.24  # due to outdated libs not changing the names after: https://github.com/numpy/numpy/pull/22607
+mne>=0.20.8
+seaborn>=0.11.0
 pysocks>=1.7.1
 pyserial>=3.5
 h5py>=3.1.0
@@ -26,7 +60,6 @@ attrdict3
 # This might try to build from source on linux (since there are no wheels for Linux on PyPI) .
 # You can pass `--find-links=https://extras.wxpython.org/wxPython4/extras/linux/gtk3/ubuntu-20.04/` your `pip install` to use the prebuilt wheels at the link.
 wxPython>=4.0 ; platform_system == "Linux"
-click
 
 # pywinhook needs some special treatment since there are only wheels on PyPI for Python 3.7-3.8, and building requires special tools (swig, VS C++ tools)
 # See issue: https://github.com/NeuroTechX/eeg-notebooks/issues/29
@@ -37,24 +70,25 @@ pywinhook @ https://github.com/ActivityWatch/wheels/raw/master/pywinhook/pyWinho
 # See issue: https://github.com/psychopy/psychopy/issues/2876
 pyglet==1.4.11 ; platform_system == "Windows"
 
-# Test requirements
-mypy
-pytest
-pytest-cov
-nbval
 
-# Types
-types-requests
 
-# Docs requirements
+## ~~ Docsbuild Requirements ~~ 
+
+# Docs
 sphinx
 sphinx-gallery
 sphinx_rtd_theme
 sphinx-tabs
 sphinx-copybutton
 sphinxcontrib-httpdomain
-numpydoc
-recommonmark
-versioneer
-rst2pdf
-docutils
+
+# Tests
+mypy
+pytest
+pytest-cov
+nbval
+
+# Types
+types-requests
+
+