Medical Image Data Volume Reduction via Longitudinal Axis Intersection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for evaluating medical image data sets, particularly MR images, face challenges such as excessive memory requirements and difficulty in focusing on specific regions of interest due to the need for full reconstruction of the thorax, leading to inefficient diagnosis and visualization of target areas like the heart.
Innovation Solution
A method that generates a reduced data set by determining a longitudinal axis through two-dimensional cross-sections, calculating points of intersection, and creating partial cross-sections using scaling factors to produce a smaller data volume tailored to the target region, allowing for quicker navigation and visualization of section images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If full three-dimensional reconstruction of the thorax is performed to locate the heart, then the target region can be found, but the memory requirement increases to over 3.2 GB and the process becomes time-consuming
Solution Approach 1:
The patent divides the large thorax volume into multiple smaller slice volumes along the longitudinal axis. Each slice is processed independently to determine presence of the target region (heart), allowing localization without loading the entire thorax into memory simultaneously. This segmentation reduces peak memory requirements while maintaining accurate localization capability.
Solution Approach 2:
The patent performs preliminary localization of the target region by processing slices in sequence to identify which slice contains the heart. This preliminary action determines the bounding box coordinates before full reconstruction, enabling subsequent focused processing only on the relevant region rather than handling the complete thorax data set.
2Ease of operation
If conventional evaluation methods are used to manually focus on the target region, then the diagnosis can be performed, but the process requires multiple approximation stages and is time-consuming
Solution Approach 1:
The system automatically performs localization and bounding box determination without requiring manual intervention or multiple approximation stages. The computer-executable instructions autonomously process the slice volumes, identify the target region, and generate the focused view, eliminating the time-consuming manual focusing process while maintaining diagnostic accuracy.
3Loss of information
If the complete image data set is reconstructed, then full anatomical information is available, but the data volume exceeds 3.2 GB and navigation becomes difficult
Solution Approach 1:
The patent extracts only the necessary portion of the anatomical data by determining a bounding box that encompasses the target region (heart) within the thorax. Instead of reconstructing and navigating the complete 3.2 GB data set, the system extracts and processes only the relevant slice volumes containing the heart, reducing data volume while preserving all necessary anatomical information for diagnosis.
4Quantity of substance
If two-dimensional cross-sections are reconstructed without focusing on the target region, then memory requirements are lower, but spatial associations cannot be directly recognized
Solution Approach 1:
The patent transitions from viewing individual two-dimensional cross-sections to reconstructing three-dimensional slice volumes that preserve spatial relationships. By organizing slices into volumetric data structures with proper spatial coordinates, the system enables direct recognition of spatial associations (such as the position and shape of the heart) while still using reduced memory compared to full thorax reconstruction.
Data Source
AI summary
The present invention relates to a method, a device and a computer program product for evaluating medical image data sets, which consists of two-dimensional section images, in particular MR images, wherein a plurality of two-dimensional cross-sections (K1, K2, . . . Kn) and at least one two-dimensional longitudinal section (L1) of a target region (1) of a human or animal body are recorded and stored, a longitudinal axis (5) and points of intersection (6) are determined at the points at which the longitudinal axis (5) extends through the plurality of two-dimensional cross-sections (K1, K2 . . . Kn) and a reduced data volume (7) is generated, which is composed of partial cross-sections (k1, k2, . . . kn), which are generated by means of the points of intersection (6) and scaling factors (a1, a2, b1, b2) from the plurality of two-dimensional cross-sections (K1, K2, . . . Kn).