Volumetric Ultrasound Segmentation Using 2D Slices and 3D Shape Completion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional ultrasound imaging systems struggle to accurately segment objects like fibroids from volumetric ultrasound datasets due to their poorly defined boundaries, leading to inaccurate segmentation and prolonged processing times, which are critical for determining appropriate care pathways.
Innovation Solution
A method and system utilizing a two-dimensional segmentation model on parallel slices and a shape completion model to generate a three-dimensional shape model for objects, incorporating a user-identified seed point and neural networks for segmentation and shape completion.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional segmentation techniques are used on volumetric ultrasound datasets, then the system is simpler to implement, but the segmentation accuracy deteriorates due to poorly defined boundaries of objects like fibroids
Solution Approach 1:
The patent applies segmentation by dividing the volumetric ultrasound dataset into multiple two-dimensional slices. Each slice is processed independently through the segmentation model to identify object boundaries, and the results are integrated to form a complete three-dimensional segmentation. This approach improves measurement precision by allowing detailed analysis of each slice while managing system complexity through modular processing.
Solution Approach 2:
The patent transitions from two-dimensional slice processing to three-dimensional object reconstruction. By processing multiple 2D slices through the segmentation model and integrating the results in three-dimensional space, the system achieves accurate segmentation of objects with poorly defined boundaries while maintaining computational feasibility.
2Productivity
If conventional segmentation techniques are used, then the processing approach is simpler, but the processing time increases and becomes too long to be clinically useful
Solution Approach 1:
The volumetric dataset is segmented into multiple 2D slices that can be processed in parallel. This division allows the segmentation model to operate on smaller, more manageable units simultaneously, significantly reducing overall processing time and improving productivity while maintaining clinical utility.
Solution Approach 2:
The system processes a representative subset of slices through the full segmentation pipeline to generate accurate results, rather than exhaustively processing every single slice. This partial action approach achieves clinically useful processing speeds while maintaining sufficient segmentation accuracy for diagnostic purposes.
3Measurement precision
If conventional segmentation techniques are used, then the method is easier to implement, but the segmentation accuracy deteriorates for objects without clearly defined boundaries
Solution Approach 1:
The patent improves boundary detection accuracy by reconstructing three-dimensional objects from multiple two-dimensional slice segmentations. This dimensional transition allows the system to infer boundaries in three-dimensional space even when individual 2D slices show poorly defined boundaries, achieving superior boundary detection accuracy while using a relatively straightforward implementation approach.
4Measurement precision
If multi-planar 2D segmentation is performed to improve accuracy, then the segmentation precision improves, but the computational resources and system complexity increase
Solution Approach 1:
The system divides the computationally intensive 3D segmentation task into multiple independent 2D slice processing operations. Each slice requires minimal computational resources, and the results are integrated to achieve high precision. This segmentation approach maintains low energy consumption per operation while achieving high overall segmentation precision through parallel processing.
Data Source
AI summary
Various methods and ultrasound imaging systems are provided for segmenting an object. In one example, a method includes accessing a volumetric ultrasound dataset, receiving an identification of a seed point for an object in an image generated based on the volumetric ultrasound dataset, and implementing a two-dimensional segmentation model on a first plurality of parallel slices based on the seed point to generate a first plurality of segmented regions. The method includes implementing the two-dimensional segmentation model on a second plurality of parallel slices based on the seed point to generate a second plurality of segmented regions. The method includes generating a detected region by accumulating the first plurality of segmented regions and the second plurality of segmented regions. The method includes implementing a shape completion model to generate a three-dimensional shape model for the object, and displaying rendering of the object based on the three-dimensional shape model.