3D Medical Image Segmentation via 2D Projection Retropropagation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current medical image segmentation tools struggle to accurately and efficiently segment thin or elongated anatomical features, such as blood vessels and bones, due to challenges like scale, noise, motion, and partial voluming in medical image data.
Innovation Solution
The method involves receiving a 2D projected rendering from a 3D medical image dataset, allowing user annotation to select regions of interest, and then retropropagating these annotations back into 3D space to generate a 3D segmentation mask using a retropropagation algorithm.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional segmentation algorithms are used on 3D medical image data, then segmentation can be performed on complex anatomical structures, but the processing time increases significantly and accuracy decreases for thin or elongated features
Solution Approach 1:
The patent projects 3D medical image data onto 2D planes to create simplified representations. By transforming the problem from 3D to 2D space, the segmentation process becomes computationally more efficient while maintaining accuracy for thin and elongated anatomical features. The projection approach allows algorithms to operate in a reduced dimensionality space, decreasing processing time significantly.
Solution Approach 2:
The patent divides the 3D medical image data into multiple 2D projected views or slices. By segmenting the volumetric data into planar components, the system can process each projection independently and more efficiently, then reconstruct the complete segmentation. This segmentation strategy reduces the computational complexity of handling entire 3D volumes at once.
2Reliability
If traditional segmentation algorithms are used on 3D medical image data, then complete volumetric segmentation can be achieved, but the computational resources and processing power required become excessive
Solution Approach 1:
The patent reduces computational burden by projecting 3D data onto 2D planes. This dimensionality reduction allows standard algorithms to process the data with significantly lower computational resources while maintaining segmentation completeness. The 2D projections preserve essential anatomical information needed for accurate segmentation of thin and elongated structures.
Solution Approach 2:
The patent creates 2D projected copies or representations of the 3D medical image data. Instead of directly processing the full 3D volumetric data, the system works with these simplified 2D projections that capture the essential features. These projected copies serve as surrogates that require far less processing power while maintaining segmentation reliability.
3Measurement precision
If manual annotation of 3D medical image data is performed, then accurate ground truth data can be generated, but the time and effort required increase substantially
Solution Approach 1:
The patent enables annotators to work in 2D projected space rather than 3D volumetric space. This dimensional transformation makes manual annotation significantly faster and more intuitive, as annotators can see and label structures in simplified planar views. The 2D projections maintain the spatial relationships needed for accurate annotation while reducing the cognitive and temporal burden.
Solution Approach 2:
The patent creates 2D projected representations that serve as annotation targets. Instead of requiring annotators to navigate and label complex 3D volumes, the system provides simplified 2D copies where annotation can be performed rapidly. These projected views preserve the essential anatomical information needed for accurate ground truth generation while dramatically improving annotation throughput.
Data Source
AI summary
Various methods and systems are provided for image data segmentation. In one example, a method includes receiving a first segmentation input selecting a first set of pixels of a two-dimensional (2D) projected rendering, the 2D projected rendering generated from a 3D medical image dataset, retropropagating the selected first set of pixels to 3D space based on a mapping between the 2D projected rendering and the 3D medical image dataset to form a 3D segmentation mask, and saving the 3D segmentation mask in memory and/or applying the 3D segmentation mask to the 3D medical image dataset, wherein the 2D projected rendering is an intensity projection rendering.