3D Dataset Annotation With 2D Label Propagation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current data annotation for machine-learning models, particularly for 3D medical images, relies heavily on manual human effort, leading to resource strain and often inadequate and inaccurate results.
Innovation Solution
An apparatus and method for automatic 3D data annotation using a processor configured to obtain and annotate 2D images from a 3D dataset, employing machine-learning models to propagate annotations across multiple images and allow user adjustment for model improvement.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual annotation by qualified annotators is used, then annotation accuracy may be maintained, but resource consumption and time requirements increase significantly
Solution Approach 1:
The system enables self-service annotation by training machine learning models to automatically annotate 3D medical images. The model learns from a small set of manually annotated examples and then autonomously performs annotation on remaining images, eliminating the need for continuous human involvement and significantly improving productivity while maintaining acceptable accuracy through iterative refinement
Solution Approach 2:
The system performs preliminary manual annotation on a small subset of images to create training data. This preliminary action enables the machine learning model to learn annotation patterns beforehand, which then allows automated annotation of the complete dataset, reducing overall human effort while maintaining accuracy
2Reliability
If manual annotation processes are used, then detailed annotations can be obtained, but the process requires tremendous human effort and strain resources
Solution Approach 1:
The system replaces the mechanical manual annotation process with an automated machine learning-based system. The ML model substitutes human annotators by learning from manually annotated examples and automatically generating annotations, thereby eliminating the time-consuming manual process while maintaining annotation quality through supervised learning and validation
3Productivity
If automated annotation is implemented, then productivity and speed improve, but annotation accuracy and reliability may deteriorate
Solution Approach 1:
The system implements feedback mechanisms where automated annotations are evaluated and used to retrain and refine the machine learning model. User corrections and validation results feed back into the training process, continuously improving model accuracy while maintaining high productivity through automated processing of the majority of images
Solution Approach 2:
The system applies partial manual annotation to a small subset of images rather than requiring complete manual annotation. This partial action provides sufficient training data for the ML model to achieve high accuracy on the remaining automated annotations, balancing productivity gains with maintained accuracy
Data Source
AI summary
Described herein are systems, methods, and instrumentalities associated with automatically annotating a 3D image dataset. The 3D automatic annotation may be accomplished based on a 2D annotation provided by an annotator and by propagating the 2D annotation through multiple images of a sequence of 2D images associated with the 3D image dataset. The automatically annotated 3D image dataset may then be used to annotate other 3D image datasets based on similarities between the first 3D image dataset and the other 3D image datasets. The automatic annotation of the first 3D image dataset and/or the other 3D image datasets may be conducted based on one or more machine-learning models trained for performing those tasks.