Real-Time Medical Image Segmentation Stabilization

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing medical imaging methods face challenges in achieving stable and accurate real-time extraction of anatomical planes due to movement during scanning, which compromises the balance between computational complexity and segmentation stability.

Innovation Solution

A computer-implemented method that utilizes previous segmentation results from the same imaging session to initialize a segmentation model, adjusting computational complexity based on the availability and quality of these results, thereby stabilizing and accelerating the segmentation process.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Loss of information

If multiple images are captured at different positions and times, then comprehensive information about the target can be obtained, but the complexity of processing and aligning these images increases

Engineering Contradiction:
Improvecomprehensive informationVSAvoidprocessing complexity
Core Design Contradiction:
Loss of informationVSDevice complexity

Solution Approach 1:

The patent segments the image processing task into distinct modules: capturing images at multiple positions, detecting features in each image, matching features across images, and generating the final assembled image. This segmentation allows each module to be optimized independently and reduces overall processing complexity by breaking down the complex task of processing multiple images into manageable steps.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent introduces feature points as intermediary elements that connect different images. By detecting features in each image and matching them across images, the system creates a common reference framework that simplifies the alignment and assembly process. The feature matching acts as an intermediary step that mediates between the individual images and the final assembled result.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Loss of information

If images are captured at different positions and times, then comprehensive information can be obtained, but the time required for processing increases

Engineering Contradiction:
Improvecomprehensive informationVSAvoidprocessing time
Core Design Contradiction:
Loss of informationVSLoss of time

Solution Approach 1:

The patent performs preliminary feature detection and extraction in each image before the matching and assembly process. By pre-processing each image to identify and extract feature points, the system prepares the data in advance, which reduces the time required during the subsequent matching and assembly operations. This preliminary action allows for more efficient processing of the complete image set.

Inventive Principle:
Principle #10Preliminary action

3Measurement precision

If feature matching is performed between multiple images, then accurate position determination can be achieved, but the computational load increases

Engineering Contradiction:
Improveposition determination accuracyVSAvoidcomputational load
Core Design Contradiction:
Measurement precisionVSPower

Solution Approach 1:

The patent applies local quality by focusing feature matching on specific local regions of interest rather than processing the entire image space. By identifying and matching only the relevant feature points in each image, the system achieves accurate position determination while reducing the computational load that would result from processing all pixels and data across multiple images.

Inventive Principle:
Principle #3Local quality

Data Source

PatentEP4639473B1Image acquisition method
Publication Date: 2026.04.29 KONINKLIJKE PHILIPS NV
  • EP4639473B1 patent drawingFigure 1~2
  • EP4639473B1 patent drawingFigure 3
  • EP4639473B1 patent drawingFigure 4

AI summary

A medical imaging and segmentation method comprising utilizing segmentation results obtained for 3D frames earlier in a same imaging session to initialize a segmentation model for subsequent 3D frames. This accelerates and stabilizes segmentation results, particularly in the context of a scanning procedure in which segmentation is performed in real time (on-the-fly) for each acquired frame.