Decode DICOM files for medical imaging

Copyright 2019 The TensorFlow IO Authors.

View on TensorFlow.org

Run in Google Colab

View source on GitHub

Download notebook

Overview

This tutorial shows how to use tfio.image.decode_dicom_image in TensorFlow IO to decode DICOM files with TensorFlow.

Setup and Usage

Download DICOM image

The DICOM image we use in this tutorial from the NIH Chest X-ray dataset.

The NIH Chest X-ray dataset consists of 100,000 de-identified images of chest x-rays in PNG format, provided by NIH Clinical Center and could be downloaded through this link.

Google Cloud also provides a DICOM version of the images, available in Cloud Storage.

In this tutorial, we will download a sample file of the dataset from the GitHub repo

Xiaosong Wang, Yifan Peng, Le Lu, Zhiyong Lu, Mohammadhadi Bagheri, Ronald Summers, ChestX-ray8: Hospital-scale Chest X-ray Database and Benchmarks on Weakly-Supervised Classification and Localization of Common Thorax Diseases, IEEE CVPR, pp. 3462-3471, 2017

curl -OL https://github.com/tensorflow/io/raw/master/docs/tutorials/dicom/dicom_00000001_000.dcm
ls -l dicom_00000001_000.dcm

  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed
100   164  100   164    0     0    318      0 --:--:-- --:--:-- --:--:--   319
100 1024k  100 1024k    0     0  1101k      0 --:--:-- --:--:-- --:--:-- 1101k
-rw-rw-r-- 1 kbuilder kokoro 1049332 Feb 12 08:50 dicom_00000001_000.dcm

Install required Packages, and restart runtime

pip install -q tensorflow-io

Decode DICOM image

import matplotlib.pyplot as plt
import numpy as np

import tensorflow as tf

import tensorflow_io as tfio

image_bytes = tf.io.read_file('dicom_00000001_000.dcm')

image = tfio.image.decode_dicom_image(image_bytes, dtype=tf.uint16)

skipped = tfio.image.decode_dicom_image(image_bytes, on_error='skip', dtype=tf.uint8)

lossy_image = tfio.image.decode_dicom_image(image_bytes, scale='auto', on_error='lossy', dtype=tf.uint8)


fig, axes = plt.subplots(1,2, figsize=(10,10))
axes[0].imshow(np.squeeze(image.numpy()), cmap='gray')
axes[0].set_title('image')
axes[1].imshow(np.squeeze(lossy_image.numpy()), cmap='gray')
axes[1].set_title('lossy image');

png

Documentation

This package has two operations which wrap DCMTK functions. decode_dicom_image decodes the pixel data from DICOM files, and decode_dicom_data decodes tag information. tags contains useful DICOM tags such as tags.PatientsName. We borrow the same tag notation from the pydicom dicom package.

Getting DICOM Image Data

io.dicom.decode_dicom_image(
    contents,
    color_dim=False,
    on_error='skip',
    scale='preserve',
    dtype=tf.uint16,
    name=None
)

contents: A Tensor of type string. 0-D. The byte string encoded DICOM file
color_dim: An optional bool. Defaults to False. If True, a third channel will be appended to all images forming a 3-D tensor. A 1024 x 1024 grayscale image will be 1024 x 1024 x 1
on_error: Defaults to skip. This attribute establishes the behavior in case an error occurs on opening the image or if the output type cannot accomodate all the possible input values. For example if the user sets the output dtype to tf.uint8, but a dicom image stores a tf.uint16 type. strict throws an error. skip returns a 1-D empty tensor. lossy continues with the operation scaling the value via the scale attribute.
scale: Defaults to preserve. This attribute establishes what to do with the scale of the input values. auto will autoscale the input values, if the output type is integer, auto will use the maximum output scale for example a uint8 which stores values from [0, 255] can be linearly stretched to fill a uint16 that is [0,65535]. If the output is float, auto will scale to [0,1]. preserve keeps the values as they are, an input value greater than the maximum possible output will be clipped.
dtype: An optional tf.DType from: tf.uint8, tf.uint16, tf.uint32, tf.uint64, tf.float16, tf.float32, tf.float64. Defaults to tf.uint16.
name: A name for the operation (optional).

Returns A Tensor of type dtype and the shape is determined by the DICOM file.

Getting DICOM Tag Data

io.dicom.decode_dicom_data(
    contents,
    tags=None,
    name=None
)

contents: A Tensor of type string. 0-D. The byte string encoded DICOM file
tags: A Tensor of type tf.uint32 of any dimension. These uint32 numbers map directly to DICOM tags
name: A name for the operation (optional).

Returns A Tensor of type tf.string and same shape as tags. If a dicom tag is a list of strings, they are combined into one string and seperated by a double backslash \\. There is a bug in DCMTK if the tag is a list of numbers, only the zeroth element will be returned as a string.

Bibtex

If this package helped, please kindly cite the below:

@misc{marcelo_lerendegui_2019_3337331,
  author       = {Marcelo Lerendegui and
                  Ouwen Huang},
  title        = {Tensorflow Dicom Decoder},
  month        = jul,
  year         = 2019,
  doi          = {10.5281/zenodo.3337331},
  url          = {https://doi.org/10.5281/zenodo.3337331}
}

License

Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at

http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.