Autonomous 3D Nervous System Segmentation Using Bony Landmarks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current technologies face challenges in autonomously segmenting three-dimensional nervous system structures from medical images, particularly with low-quality images, which affects the accuracy and safety of image-guided surgery and diagnostics.
Innovation Solution
A method and system using a pre-trained convolutional neural network (CNN) for autonomous segmentation of bony structures, combined with Select-Attend-Transfer gates and Generative Adversarial Networks, to process 3D medical images and determine the shape, location, and size of nervous system structures, improving image quality and enabling collision detection during surgical procedures.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If low-quality medical images are used, then radiation exposure is reduced, but segmentation accuracy deteriorates
Solution Approach 1:
The system performs preliminary actions by first segmenting bony structures from low-quality images, then using those segmentation results to guide the subsequent segmentation of nervous system structures. This preliminary segmentation of easier-to-detect bones provides structural information that compensates for the low image quality, enabling accurate nervous system segmentation without requiring high-quality images.
Solution Approach 2:
The patent introduces an intermediary approach by using bony structure segmentation as a mediator between the low-quality image and the final nervous system segmentation. The bony structures serve as intermediate landmarks that help locate and identify nervous system structures, bridging the gap between poor image quality and accurate segmentation requirements.
2Object-affected harmful factors
If low-quality images are used, then cost and radiation exposure are reduced, but identification of anatomic landmarks deteriorates
Solution Approach 1:
The system segments the image processing task into distinct components: first segmenting bony structures, then using those segments as references for identifying nervous system structures. This segmentation strategy allows the system to leverage the structural information from bones to compensate for poor soft tissue visibility in low-quality images.
Solution Approach 2:
The patent creates a computational copy or representation of the anatomical structure through automated bony structure segmentation. This digital model of the bones serves as a reference framework that can be overlaid with the original low-quality image to help identify nervous system structures, effectively creating a synthetic enhanced version of the image data.
3Productivity
If autonomous segmentation is implemented, then productivity is improved, but reliability deteriorates due to lack of human verification
Solution Approach 1:
The system implements self-service through autonomous segmentation, where the computer automatically performs the segmentation task without requiring human intervention. The automated pipeline processes images through multiple stages (bony structure segmentation, then nervous system segmentation) independently, achieving high productivity while maintaining reliability through the systematic approach and use of anatomical priors.
Data Source
Figure 1
Figure 2A~2B
Figure 2C~2D
AI summary
A method for autonomous segmentation of three-dimensional nervous system structures from raw medical images, the method comprising the following steps: receiving (301) a 3D scan volume comprising a set of medical scan images of a region of the anatomy; autonomously processing (302) the set of medical scan images to perform segmentation of a bony structure of the anatomy to obtain bony structure segmentation data; autonomously processing (304) a subsection of the 3D scan volume as a 3D region of interest (ROI) by combining (303) the raw medical scan images and the bony structure segmentation data, wherein the 3D ROI contains a subvolume of the bony structure with a portion of surrounding tissues, including the nervous system structure; autonomously processing the ROI (306) to determine the 3D shape, location, and size of the nervous system structures by means of a pre-trained convolutional neural network (CNN) (400).