Labeling interfaces
Centaur Labs offers multiple labeling interfaces based on your needs. The recommended interface will depend on your data type, task complexity, and desired labeler population.
Overview
Our labeling interfaces are:
- DiagnosUs - our mobile labeling platform (iOS & Android)
This is our primary labeling interface and provides access to our vast mobile labeler network. This interface is suitable for most labeling needs and offers the most scale. - Desktop labeling - accessible via the desktop Centaur Platform
The solution is tailored to utilizing consultants (sourced by you or Centaur) or your own internal team. We recommend gathering multiple opinions on each case and leveraging our collective intelligence software to aggregate those opinions. This interface is often used when annotators don’t have access to mobile devices or prefer a larger screen. - Desktop DICOM viewer - accessible via the desktop Centaur Platform
The desktop DICOM viewer is also optimal for contracted consultants or your internal team. This specific solution is best for annotating DICOM images where desktop is preferred, similar to (2), and the annotation is particularly complex, requiring hours, not minutes on each case.
Summary
| Interface | Platform | Potential labelers | Best for |
|---|---|---|---|
| DiagnosUs | iOS 15+ or Android 13+ (iPhone, iPad, Android devices) | -Centaur mobile labeler network -Client team -Consultants | Most labeling tasks |
| Desktop labeling | Web browser | -Client team -Consultants | Labeling for your internal team |
| OHIF viewer | Web browser | -Client team -Consultants | Complex DICOM tasks |
Contact your project manager if you’re not sure which labeling interface to use - they will suggest the best option based on your needs.
Supported Annotation Types
(1) Classification
- Fixed answer choices (same answer choices for each case)
- Dynamic answer choices (answer choices vary based on case)
(2) Polygon, box, line, circle, and pixel segmentation
- Single class (one class of annotations)
- Multi-class (two classes of annotations e.g. vessels and veins on heart ultrasounds)
(3) Range Selection
- Named Entity Recognition
- Time Range Selection
(4) 2D Point Segmentation
- Annotation of precise (x, y) point coordinates (e.g., anatomical landmarks) in the Centaur DICOM Viewer; available for DICOM Viewer tasks only