Lymph Node Detection via Voxel Expansion and Prioritization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for detecting lymph nodes, bony lesions, and pancreatic masses from CT images face challenges due to low contrast and varying sizes, poses, shapes, and sparse distribution of these structures.
Innovation Solution
The proposed solution involves a computer-readable medium with program instructions that receive digital images, select voxels corresponding to soft tissue or bone cells, expand these voxels to identify endpoints, and determine the presence of soft tissue bodies or bones, associating the images with databases accordingly.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional methods (random forest classifier, Hessian matrix, radial structure tensor) are used to detect lymph nodes, then detection capability is provided, but detection accuracy is poor due to low contrast and varying characteristics of lymph nodes
Solution Approach 1:
The patent segments the lymph node detection process into multiple stages: initial voxel selection based on intensity thresholds, iterative expansion to capture full lymph node boundaries, and classification of expanded regions. This segmentation allows each stage to address specific challenges (low contrast, varying sizes, sparse distribution) independently, improving overall detection accuracy while managing complexity.
2Productivity
If manual review of all CT images is performed, then comprehensive analysis is achieved, but time consumption and cost are excessive
Solution Approach 1:
The system implements self-service through automated lymph node detection and prioritization. The computer-implemented method automatically processes CT images, identifies potential lymph nodes, calculates priority scores based on detection confidence and clinical relevance, and generates prioritized review lists. This automation reduces manual review time while maintaining comprehensive analysis through systematic processing of all images.
3Productivity
If all CT images are reviewed equally, then no images are prioritized, but clinical efficiency is reduced due to lack of prioritization
Solution Approach 1:
The patent implements feedback mechanisms where detection results from initial voxel analysis inform subsequent expansion operations, and priority scores based on detection confidence feed into the final prioritization ranking. The system continuously refines its analysis by using outcomes from each processing stage to guide subsequent steps, enabling efficient prioritization without excessive complexity through iterative feedback loops.
Data Source
AI summary
The disclosure provides example methods and a non-transitory computer-readable medium that has stored thereon program instructions that upon execution by a processor, cause performance of a set of acts including: (a) receiving a digital image, where the image comprises a plurality of voxels, (b) selecting from the plurality of voxels at least one voxel corresponding to soft tissue cells, (c) expanding the selected voxels to include adjacent voxels until an endpoint voxel is identified, (d) determining whether the expanded selected voxels indicate a presence of a soft tissue body, and (e) in response to a determination that the expanded selected voxels indicate the presence of the soft tissue body, associating the digital image with a database of digital images showing soft tissue bodies.