neuro-ml / dicom-csv

Utils for gathering, aggregation and handling metadata from DICOM files

Utils for gathering, aggregation and handling metadata from DICOM files.

Installation

From pip

pip install dicom-csv

or from GitHub

git clone https://github.com/neuro-ml/dicom-csv
cd dicom-csv
pip install -e .

Example join_tree

>>> from dicom_csv import join_tree
>>> folder = '/path/to/folder/'
>>> meta = join_tree(folder, verbose=2)
>>> meta.head(3)

AccessionNumber	AcquisitionDate	...	WindowCenter	WindowWidth
000002621237	20200922	...	-500.0	1500.0
000002621237	20200922	...	-40.0	400.0
000002621237	20200922	...	-500.0	1500.0
3 rows x 155 columns

Example load 3D image

from a series of dicom files (each containing 2D image)

from dicom_csv import join_tree, order_series, stack_images
from pydicom import dcmread
from pathlib import Path

# 1. Collect metadata from all dicom files
folder = Path('/path/to/folder/')
meta = join_tree(folder, verbose=2)

# 2. Select series to load
uid = '...' # unique identifier of a series you want to load,
            # you could list them by `meta.SeriesInstanceUID.unique()`
series = meta.query("SeriesInstanceUID==@uid")

# 3. Read files & combine them into a single volume
images2d = [dcmread(folder / row[1].PathToFolder / row[1].FileName) for row in series.iterrows()] 
image3d = stack_images(order_series(images2d))

Documentation

You can find the documentation here.