Medical Image Processing With LMIP for 3D Vessel Annotation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for annotating and segmenting tubular organs in medical images, particularly blood vessels, face challenges in tracing small, curved, or intersecting vessels due to their complex geometries and noise, leading to inefficiencies and inaccuracies in manual annotation processes.
Innovation Solution
A tubular organ annotation method using local maximum/minimum intensity projections (LMIP) to enhance visibility and connectivity, followed by inverse mapping to generate a three-dimensional centerline, and a correction method to refine preliminary annotations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional iteration algorithms are used for tracing blood vessel centerlines, then the method can handle simple blood vessel structures, but it fails to trace small, curved, or intersecting blood vessels accurately
Solution Approach 1:
The patent transforms the three-dimensional blood vessel tracing problem into a two-dimensional projection problem. By projecting the 3D volume data onto a 2D plane along the blood flow direction, the complex 3D tracing task becomes a simpler 2D centerline detection task, which can then be mapped back to 3D space to achieve accurate tracing of small, curved, and intersecting vessels.
Solution Approach 2:
The patent introduces a projection image as an intermediary representation between the original 3D volume data and the final 3D centerline result. This intermediate 2D projection allows for easier identification of blood vessel paths, which are then transferred back to the 3D space through coordinate mapping, solving the difficulty of direct 3D tracing.
2Manufacturing precision
If manual annotation is performed by browsing multiple images sequentially from root to end point, then complete blood vessel annotation can be achieved, but it requires an extremely long period of time
Solution Approach 1:
The patent enables annotators to view blood vessels in a projected 2D view where the entire blood vessel path is visible simultaneously rather than requiring sequential browsing through multiple 3D slices. This dimensional transformation allows for rapid annotation while maintaining completeness, as the projection displays the full extent of blood vessels in a single view.
Solution Approach 2:
The patent performs projection and centerline identification in advance to create a reference annotation that can be directly transferred to the original 3D volume data. This preliminary processing eliminates the need for time-consuming manual browsing and annotation of each individual slice, dramatically improving annotation efficiency while maintaining accuracy.
3Illumination intensity
If contrast agent injection is used to strengthen image display of blood vessel centerline, then visibility of blood vessels is improved, but it requires additional invasive procedures and time
Solution Approach 1:
The patent replaces the physical/chemical method of contrast agent injection with a computational image processing method. By using projection algorithms to enhance the visibility of blood vessels in the image data, the system achieves improved visualization without requiring any invasive medical procedures or additional imaging agents, thereby reducing procedure complexity and risk.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
LMIP enhances the visibility of tubular organ centerlines, allowing accurate annotation of small structures and correcting incomplete annotations, thereby improving efficiency and accuracy in medical image processing.
Implementation Method 1
obtaining a projection image by implementing a local intensity projection on two-dimensional images structuring three-dimensional volume data rendering a tubular organ
Data Source
AI summary
A medical image processing apparatus according to an embodiment includes processing circuitry configured: to generate a projection image by implementing an intensity projection on a plurality of two-dimensional images structuring three-dimensional volume data rendering a tubular organ; to obtain a mapping matrix of the intensity projection; to annotate the tubular organ in the projection image; and to identify the tubular organ in the three-dimensional volume data, by inversely mapping the tubular organ annotated in the projection image onto the three-dimensional volume data while using the mapping matrix.