DicomAnnotator

DicomAnnotator facilitates annotation of DICOM images to generate labeled datasets for supervised machine learning tasks such as medical image classification, segmentation, and object detection.

Key Features:

• Configurable annotation tasks: Supports annotation configurations for tasks including classification, segmentation, and object detection on DICOM images.
• Multiple annotation types: Allows placement of multiple annotation types on an image or on specific regions within an image.
• Annotator attribution: Records and attributes annotations to individual annotators to track contributor-specific labels.
• DICOM support: Displays and handles DICOM-formatted images to maintain compatibility with medical imaging standards.

Scientific Applications:

• Dataset generation for supervised learning: Produces labeled DICOM datasets for training and validating machine learning models for classification, segmentation, and object detection.
• Spine image annotation evaluation: Applied to spine image annotation as a test case to assess annotation consistency across annotators.
• Collaborative annotation studies: Enables tracking of individual annotator contributions for inter-rater analysis and consensus studies.

Methodology:

Configurable placement of annotations on images or regions, annotator attribution, and DICOM image display were implemented, and the software was evaluated using a spine image annotation test case.