Medical Image Processing Apparatus for Spinal Landmark Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The natural curvature of the spine makes it difficult to accurately detect and label vertebrae in medical images, leading to errors in automatic landmark detection and potential mis-labeling of vertebrae.
Innovation Solution
A medical image processing apparatus and method that involves obtaining the positions of anatomical landmarks, generating a spline based on these positions, normalizing the positions to one-dimensional data, adjusting the landmarks based on the normalized data, and re-mapping them to three-dimensional space to improve detection accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If automated landmark detection is used to identify vertebrae, then detection speed and workflow efficiency are improved, but detection accuracy deteriorates due to the natural curvature of the spine and similarity between neighbouring vertebrae
Solution Approach 1:
The patent divides the spine into multiple one-dimensional projections, each representing a different viewing angle or anatomical plane. By segmenting the three-dimensional detection problem into multiple one-dimensional sub-problems, the system can detect vertebrae more accurately in each projection while maintaining overall detection speed through automated processing of multiple views.
Solution Approach 2:
The patent transforms the three-dimensional vertebrae detection problem into multiple one-dimensional detection problems by creating projections along different axes. This dimensionality reduction allows detection networks to focus on specific anatomical planes, reducing the complexity of distinguishing neighbouring vertebrae while preserving the ability to identify all vertebrae through integration of multiple one-dimensional results.
2Extent of automation
If detection networks are used to label vertebrae, then automated labeling is achieved, but labeling accuracy deteriorates due to difficulty in distinguishing neighbouring vertebrae with similar appearance
Solution Approach 1:
The patent applies segmentation by creating multiple one-dimensional projections of the spine from different angles and anatomical planes. Each projection is independently labeled by detection networks, allowing the system to overcome the limitation of similar appearances in three-dimensional space by examining distinct one-dimensional views where vertebrae have more distinguishable characteristics.
Solution Approach 2:
The patent introduces one-dimensional projections as intermediary representations between the three-dimensional medical images and the final vertebrae labels. These intermediate one-dimensional views serve as mediators that simplify the discrimination task for detection networks, enabling more accurate automated labeling by transforming the input data into a form where neighbouring vertebrae are more easily distinguished.
3Reliability
If vertebrae are detected in three-dimensional space, then comprehensive anatomical information is captured, but detection difficulty increases due to the natural curvature of the spine
Solution Approach 1:
The patent reduces the three-dimensional detection problem into multiple one-dimensional projections, making the detection task easier by eliminating the complexity introduced by spinal curvature in 3D space. Each one-dimensional projection represents a simplified view where vertebrae appear as sequential structures, reducing detection difficulty while maintaining anatomical information through multiple complementary projections.
Solution Approach 2:
The patent segments the three-dimensional spine into multiple one-dimensional projections along different axes and anatomical planes. This segmentation transforms the complex curved three-dimensional structure into several simpler one-dimensional sequences, reducing detection difficulty in each segment while preserving comprehensive anatomical information through the combination of all segments.
Data Source
AI summary
A medical image processing apparatus comprises processing circuitry configured to:obtain a respective position for each of a plurality of anatomical landmarks in a three-dimensional space;generate a spline based on the positions for the plurality of anatomical landmarks; use the spline to normalize the positions to obtain one-dimensional data or two-dimensional data;adjust at least one of the anatomical landmarks based on the one dimensional data or two-dimensional data; andre-map the adjusting of at least one anatomical landmark to the three-dimensional space.