DicomAnnotator: a Configurable Open-Source Software Program for Efficient DICOM Image Annotation

Abstract

Modern, supervised machine learning approaches to medical image classification, image segmentation, and object detection usually require many annotated images. As manual annotation is usually labor-intensive and time-consuming, a well-designed software program can aid and expedite the annotation process. Ideally, this program should be configurable for various annotation tasks, enable efficient placement of several types of annotations on an image or a region of an image, attribute annotations to individual annotators, and be able to display Digital Imaging and Communications in Medicine (DICOM)-formatted images. No current open-source software program fulfills these requirements. To fill this gap, we developed DicomAnnotator, a configurable open-source software program for DICOM image annotation. This program fulfills the above requirements and provides user-friendly features to aid the annotation process. In this paper, we present the design and implementation of DicomAnnotator. Using spine image annotation as a test case, our evaluation showed that annotators with various backgrounds can use DicomAnnotator to annotate DICOM images efficiently. DicomAnnotator is freely available at https://github.com/UW-CLEAR-Center/DICOM-Annotator under the GPLv3 license.

Keywords: DICOM; Image annotation; Machine learning; Open source; Software design.

Fig. 1

Our approach to annotating multiple regions of interest in an image. The subfigures are the following: a the “L5” region identifier is selected, b the annotator provides a bounding polygon for L5 and then inputs the labels for L5 (no hardware, fractured), c the “L4” identifier is automatically selected after completing the last labels for L5, and d the annotator provides a bounding polygon for L4 and then inputs the labels for L4 (no hardware, normal). This process continues until image annotation is complete

Fig. 2

The steps to using DicomAnnotator to annotate an image dataset

Fig. 3

The login page demonstrating the username field, a dropdown to select a classification system, and radio buttons to determine the order of region identifiers shown on DicomAnnotator’s main page

Fig. 4

The main page of DicomAnnotator that is divided into 12 panels. Panel 1: radio buttons used to select an operation mode; panel 2: a group of buttons that allow the user to move between images, remove annotations, reset the image display, manually save annotations, and display help text; panel 3: text box showing details about the currently displayed image and the annotation process; panel 4: buttons used to set an image to unreadable when it is of low or non-diagnostic quality and to horizontally flip the image; panel 5: buttons used to flag/unflag an image for later review and to navigate through the flagged images; panel 6: commenting system where comments from any user are displayed and new comments can be added; panel 7: indicator of whether new annotations have been stored in the result file; panel 8: canvas displaying an image; panel 9: radio buttons for assigning an image label; panel 10: annotation table which has been configured to apply multiple annotations to each region of interest; panel 11: buttons used to toggle off the annotated points in the image and to invert the image’s grayscale; and panel 12: text boxes showing the identifiers of the regions that are not assigned the default region label like “Normal”

Fig. 5

The user’s interaction with the image in the annotation process. a The target vertebral body for annotation is identified by the user using the default display parameters. b The display window and level, zooming, and panning are employed to optimize visualization of the target vertebral body. c The boundary of the target vertebral body is marked by placing points at its four corners

Fig. 6

The final annotations of an example spine image

Fig. 7

The main page in the “View Only” mode demonstrating the annotations with the manipulation tools grayed out to prevent accidental alteration of the annotations. The user can return to the “Edit” mode by clicking the radio button in the upper left

Fig. 8

The confirmation dialog that is displayed when switching from the “View Only” mode to the “Edit” mode

Fig. 9

An example of comments displayed in the comment panel for an image

Fig. 10

The usability evaluation process

Fig. 11

The configuration file for the spine CT image annotation task in JSON format

Fig. 12

DicomAnnotator demonstrating display and annotations of an example sagittal lumbar spine CT image