3D Organ Segmentation Using Geometric Constrictions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing segmentation methods struggle to separate individual organs within connected systems like the digestive, respiratory, and circulatory systems, as there are no clear numerical differences at their boundaries, leading to anomalous results when using probabilistic models or training data.
Innovation Solution
An image processing method that calculates distance values, generates hierarchical layers, detects local maximum layers, and partitions the foreground region into segments based on these layers, allowing for the identification of individual organs without relying on probabilistic models or training data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing segmentation methods look for numerical differences at region boundaries, then segmentation can be performed when clear boundaries exist, but segmentation fails when there are no numerical differences at boundaries of connected organs
Solution Approach 1:
The invention changes the parameter used for segmentation from numerical intensity differences to geometric shape characteristics. By computing distance transforms and analyzing the geometry of the foreground region, the method can identify constricted locations that separate connected organs even when they have similar intensity values, thus resolving the contradiction between boundary detection accuracy and adaptability to connected organs
Solution Approach 2:
The invention applies segmentation by dividing the foreground region into multiple segments based on detected constricted locations. This allows connected organs to be separated into individual segments for independent analysis, enabling the system to handle cases where organs are connected but should be analyzed separately
2Adaptability or versatility
If probabilistic models or training data are used for segmentation, then segmentation can be performed on connected organs, but objective verification of performance becomes difficult and anomalous results occur with exceptional inputs
Solution Approach 1:
The invention makes the segmentation method self-service by using only the geometric properties of the input data itself, without requiring external probabilistic models or training data. The distance transform and concavity analysis operate directly on the foreground region geometry, enabling objective verification and consistent performance across different inputs while maintaining the ability to segment connected organs
3Loss of information
If doctors inspect all 2D images to estimate 3D organ structure, then complete organ structure can be recognized, but inspection time and data processing volume increase significantly
Solution Approach 1:
The invention segments the 3D foreground region into individual organ segments, allowing doctors to focus inspection on specific organs of interest rather than examining all 2D images for complete organ structure estimation. This reduces inspection time and data processing volume while maintaining the ability to recognize complete organ structures when needed
Solution Approach 2:
The invention transforms 2D image data into a 3D distance transform representation, enabling intuitive 3D visualization and inspection of organ structures. This dimensional change allows doctors to comprehend organ structures more efficiently by inspecting 3D segments rather than comparing numerous 2D images
Data Source
AI summary
The present invention provides an image processing method capable of intuitively recognizing the three-dimensional shape of an organ by limiting the inspection range to the vicinity of the diagnostic target organ in image diagnosis. Given binary three-dimensional data representing an object in three-dimensional space, the image processing method partitions the object into individual segments with boundaries on narrow parts of the object's shape. A series of organs with no numerical differences in the boundaries on the image data can be segmented into individual organs at geometrically constricted locations because the connected organs can be identified by their shape. As a result, it is possible to examine a limited area specific to the target organ during image diagnosis, and to perform the examination while viewing the intuitive 3D shape of the organ during image interpretation.