Adds TIB algorithm implementation (#22)

* adds implementation of TIB algorithm

* updates quick start docs

docs/source/index.rst

@@ -41,10 +41,10 @@ examples and more information on the functions can be found in the documentation
     data, sample_freq = paat.read_gt3x('path/to/gt3x/file')
 
     # Detect non-wear time
-    data.loc[:, "Non Wear Time"] = paat.detect_non_wear_time_syed2021(data, sample_freq)
+    data.loc[:, "Non Wear Time"] = paat.detect_non_wear_time_hees2011(data, sample_freq)
 
     # Detect sleep episodes
-    data.loc[:, "Sleep"] = paat.detect_sleep_weitz2022(data, sample_freq)
+    data.loc[:, "Sleep"] = paat.detect_time_in_bed_weitz2024(data, sample_freq)
 
     # Classify moderate-to-vigorous and sedentary behavior
     data.loc[:, ["MVPA", "SB"]] = paat.calculate_pa_levels(data, sample_freq)
@@ -58,9 +58,6 @@ examples and more information on the functions can be found in the documentation
 
 .. note::
 
-    In this example, methods of `Syed et al. (2021) <https://doi.org/10.1038/s41598-021-87757-z>`_
-    and `Weitz et al. (2022) <https://www.medrxiv.org/content/10.1101/2022.03.07.22270992>`_ and activity
-    thresholds of `Sanders et al. (2019) <https://doi.org/10.1080/02640414.2018.1555904>`_.
-    However, these are only examples. There are multiple methods implemented in PAAT
+    Note that these are only examples. There are multiple methods implemented in PAAT
     and the processing pipeline can easily be adjusted to individual needs. More examples
     can be found in the examples section.

paat/__init__.py

@@ -31,7 +31,7 @@
 from .estimates import calculate_pa_levels, create_activity_column
 from .features import calculate_actigraph_counts, calculate_vector_magnitude, calculate_brond_counts
 from .io import read_gt3x
-from .sleep import detect_sleep_weitz2022, detect_sleep_triaxial_weitz2022
+from .sleep import detect_sleep_weitz2022, detect_sleep_triaxial_weitz2022, detect_time_in_bed_weitz2024
 from .wear_time import detect_non_wear_time_naive, detect_non_wear_time_hees2011, detect_non_wear_time_syed2021
 
 try:

paat/sleep.py

@@ -12,9 +12,16 @@
 import numpy as np
 from torch import nn
 import torch
+import tensorflow as tf
+from tensorflow.keras import models
 
-from . import features
+# Hide GPU from visible devices
+tf.config.set_visible_devices([], 'GPU')
+#tf.compat.v1.disable_eager_execution()
+
+os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'
 
+from . import features
 
 class _SleepModel(nn.Module):
     def __init__(self, input_dim, hid_dim, output_dim, n_layers, dropout, batch_first=False):
@@ -49,6 +56,9 @@ def detect_sleep_weitz2022(data, sample_freq, means=None, stds=None):
     """
     Infer time in bed from raw acceleration signal using frequency features.
 
+    .. warning::
+        This method turned out to work not as accurately as initially thought. Use with care!
+
     Parameters
     ----------
     data : DataFrame
@@ -101,6 +111,9 @@ def detect_sleep_triaxial_weitz2022(data, sample_freq, resampled_frequency="1min
     """
     Infer time in bed from raw acceleration signal.
 
+    .. warning::
+        This method turned out to work not as accurately as initially thought. Use with care!
+
     Parameters
     ----------
     data : DataFrame
@@ -154,3 +167,56 @@ def detect_sleep_triaxial_weitz2022(data, sample_freq, resampled_frequency="1min
     predictions = np.repeat(predictions, seconds * sample_freq)
 
     return predictions
+
+def detect_time_in_bed_weitz2024(data, sample_freq, resampled_frequency="1min", means=None, stds=None, model=None):
+    """
+    Infer time in bed from raw acceleration signal.
+
+    Parameters
+    ----------
+    data : DataFrame
+        a DataFrame containg the raw acceleration data
+    sample_freq : int
+        the sampling frequency in which the data was recorded
+    resampled_frequency : str (optional)
+        a str indicating to what frequency the data should be resampled. This depends
+        on the model used to predict, defaults to 1min.
+    means : array_like (optional)
+        a numpy array with the channel means, will be calculated for the sample
+        if not specified
+    stds : array_like (optional)
+        a numpy array with the channel stds, will be calculated for the sample
+        if not specified
+    model : nn.Module (optional)
+        a loaded pytorch custom model.
+
+    Returns
+    -------
+    predicted_time_in_bed : np.array (n_samples,)
+        a numpy array indicating whether the values of the acceleration data were spent in bed
+
+    """
+    if resampled_frequency:
+        data = data[['X', 'Y', 'Z']].resample(resampled_frequency).mean()
+
+    X = data.reset_index()[["Y", "X", "Z"]].values.copy()
+
+    # If no means and stds are given, calculate subject's mean and std
+    # to normalize by this
+    if not means or not stds:
+        means, stds = X.mean(axis=0), X.std(axis=0)
+
+    # Normalize input
+    X = (X - means) / stds        
+
+    # Load model if not specified
+    if not model:
+        model_path = os.path.join(os.path.pardir, os.path.dirname(__file__), 'models', 'TIB_model.h5')
+        model = models.load_model(model_path)
+
+    predictions = (model.predict(X[np.newaxis], verbose=0).squeeze() >= .5)
+
+    seconds = pd.Timedelta(resampled_frequency).seconds
+    predictions = np.repeat(predictions, seconds * sample_freq)
+
+    return predictions

pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "paat"
-version = "1.0.0-beta.2"
+version = "1.0.0b3"
 description = "A comprehensive toolbox to analyse and model raw physical activity data"
 
 license = "MIT"
@@ -28,15 +28,14 @@ classifiers = [
 
 [tool.poetry.dependencies]
 python = ">=3.9,<3.11"  # Compatible python versions must be declared here
-pandas = "^1.2.4"
-scipy = "^1.6.2"
+pandas = "^1.5.3"
 psutil = "^5.8.0"
 glob2 = "^0.7"
 resampy = "^0.2.2"
 joblib = "^1.0.1"
-numpy = "^1.20.3"
+numpy = "1.23.5"
 bitstring = "^3.1.7"
-tensorflow = "^2.7.0"
+tensorflow = "^2.11.0"
 torch = "^1.10.1"
 agcounts = "^0.1.1"
 toml = "^0.10.2"