"""Process ADNI study data with adnipy."""
# Standard library imports
import warnings
from io import StringIO
# Third party imports
import pandas as pd
[docs]
def read_csv(file: str | StringIO) -> pd.DataFrame:
"""Return a csv file as a pandas.DataFrame.
Recognizes missing values used in the ADNI database.
Parameters
----------
file : str, pathlib.Path
The path to the .csv file.
Returns
-------
pd.DataFrame
Returns the file as a dataframe.
See Also
--------
standard_column_names
standard_dates
standard_index
"""
# empty values
na_values = ["-1", "-4"]
return pd.read_csv(file, na_values=na_values)
[docs]
def timedelta(old: pd.DataFrame, new: pd.DataFrame) -> pd.Series:
"""Get timedelta between timepoints.
Parameters
----------
old : pd.DataFrame
This is the older dataframe.
new : pd.DataFrame
This is the newer dataframe.
Returns
-------
pd.Series
The content will be timedelta values. Look into numpy for more options.
"""
old = old.reset_index()
old = old.set_index("Subject ID")
new = new.reset_index()
new = new.set_index("Subject ID")
return old["SCANDATE"] - new["SCANDATE"]
[docs]
def get_matching_images(left: pd.DataFrame, right: pd.DataFrame) -> pd.DataFrame:
"""Match different scan types based on closest date.
The columns 'Subject ID' and 'SCANDATE' are required.
Parameters
----------
left : pd.DataFrame
Dataframe containing the tau scans.
right : pd.DataFrame
Dataframe containing the mri scans.
Returns
-------
pd.DataFrame
For each timepoint there is a match from both inputs.
"""
left = left.set_index(["Subject ID", "SCANDATE"])
left = left.sort_index()
right = right.set_index(["Subject ID", "SCANDATE"])
right = right.sort_index()
missing_match = []
matching_images = []
right_subjects = right.index.get_level_values(0)
def closest_date(subject: pd.DataFrame, index: tuple) -> pd.Timestamp:
"""Get closest date from list."""
unique_dates = subject.index.unique()
return min(unique_dates, key=lambda x, index=index: abs(x - index[1])) # type: ignore[misc]
for index in left.index:
if index[0] in right_subjects:
subject = right.loc[index[0]]
date = closest_date(subject, index)
matching_image = right.loc[index[0], date]
image = left.loc[[index]]
image["Image ID_r"] = matching_image.to_numpy()[0]
matching_images.append(image)
else:
missing_match.append(index)
matching_images_df = pd.concat(matching_images)
matching_images_df = matching_images_df.rename(columns={"Image ID": "Image ID_l"})
if missing_match:
missing_match_str = str(set(missing_match))
message = "Could not find matching images for:" + missing_match_str
warnings.warn(message, stacklevel=1)
return matching_images_df