Dynamic Image Analysis for Pulmonary Blood Flow Calculation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current techniques face challenges in accurately calculating the pulmonary blood flow feature amount in chest dynamic images due to the three-dimensional movement of blood vessels, making it difficult to correctly align and evaluate blood flow in two-dimensional images, and the exclusion of main vessels leads to incorrect evaluations.
Innovation Solution
A dynamic image analysis system that includes a hardware processor to obtain and process chest dynamic images, extract the lung field region, calculate the blood flow rate feature amount, and set an upper limit value to exclude noise from main vessels, using deep learning for generating lung field masks and applying an organ back correction coefficient to accurately calculate the left-to-right ratio of pulmonary blood flows.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If alignment process is applied to chest dynamic images to extract blood flow features, then blood flow calculation can be performed, but measurement precision deteriorates because blood vessels move three-dimensionally making correct alignment difficult in two-dimensional images
Solution Approach 1:
The patent transitions from 2D image alignment to 3D spatial-coordinate-based extraction. Instead of attempting to align 2D images where blood vessels appear to move due to projection effects, the system uses 3D spatial coordinates from multiple viewing angles to directly calculate blood flow, eliminating the alignment problem by working in the third dimension.
Solution Approach 2:
The patent introduces a new intermediary coordinate system that maps blood vessel positions from 2D images to 3D space. By using spatial coordinates as an intermediary representation, the system can track blood flow without requiring direct alignment of 2D images, thus resolving the precision issue while maintaining calculation capability.
2Productivity
If main vessels are excluded from the evaluation region to calculate peripheral blood flow, then peripheral blood flow can be evaluated, but measurement precision deteriorates because peripheral vessels exist on the back of main vessels in an overlaid manner
Solution Approach 1:
The patent uses 3D spatial coordinates to separate overlapping vessels that appear overlaid in 2D images. By reconstructing the spatial positions of blood vessels in three dimensions, the system can distinguish between main vessels and peripheral vessels that are superimposed in the 2D projection, allowing accurate peripheral blood flow measurement without excluding main vessels.
Solution Approach 2:
The patent segments the blood vessel structure into different spatial layers using 3D coordinate information. This segmentation allows the system to separately identify and measure blood flow in peripheral vessels even when they are visually overlaid on main vessels in 2D images, improving measurement precision while maintaining evaluation capability.
Data Source
AI summary
A dynamic image analysis apparatus including a hardware processor that: obtains a chest dynamic image obtained by dynamic radiographing through radiation; extracts a lung field region from the dynamic image; calculates a feature amount about a blood flow rate, based on the lung field region; and limits a value of the calculated feature amount about the blood flow rate.