Medical Image Segmentation for Cancer Invasion Range Detection

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing image recognition technologies struggle to accurately determine the inclusion relationship between cancer regions and peripheral tissue regions in medical images, requiring significant specialization and a heavy workload to specify the presence or absence of cancer invasion into peripheral tissues.

Innovation Solution

An estimation model is trained to perform segmentation on cancer and muscularis propria regions by calculating probabilities for each pixel, using a loss function that minimizes the difference between actual and predicted inclusion relationships, allowing for accurate visualization of invasion ranges.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of operation

If an estimation model is used to determine inclusion relationships between cancer regions and peripheral tissue regions, then the workload and specialization requirements are reduced, but the accuracy of detecting cancer invasion may be insufficient

Engineering Contradiction:

ImproveworkloadVSAvoidaccuracy of cancer invasion detection

Core Design Contradiction:

Ease of operationVSMeasurement precision

Solution Approach 1:

The patent segments the image into multiple regions including cancer regions and peripheral tissue regions, and further segments these regions into pixel units. By calculating inclusion relationships at the pixel level rather than treating regions as whole units, the model achieves more precise detection of cancer invasion boundaries while maintaining automated operation.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent changes the parameter of measurement from region-level to pixel-level inclusion relationships. By calculating the degree of inclusion for each pixel pair between cancer and peripheral tissue regions, the model achieves higher precision in detecting invasion boundaries while maintaining computational feasibility through parameter transformation.

Inventive Principle:

Principle #35Parameter changes

2Manufacturing precision

If traditional image recognition methods are used without pixel-level segmentation, then the processing speed is faster, but the precision of specifying inclusion relationships is insufficient

Engineering Contradiction:

Improveprecision of inclusion relationship specificationVSAvoidprocessing speed

Core Design Contradiction:

Manufacturing precisionVSProductivity

Solution Approach 1:

The patent divides regions into pixel units and calculates inclusion relationships for each pixel pair. This segmentation enables precise specification of invasion boundaries at the pixel level, achieving high precision in inclusion relationship specification while the automated calculation process maintains efficient processing speed.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent replaces manual expert analysis with an automated estimation model that calculates pixel-level inclusion relationships. This substitution of mechanical/manual processes with computational algorithms achieves both high precision in specifying inclusion relationships and maintained processing speed through automation.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

3Measurement precision

If pixel-level probability calculation is performed for each pixel, then the accuracy of invasion range visualization is improved, but the computational complexity increases

Engineering Contradiction:

Improveaccuracy of invasion range visualizationVSAvoidcomputational complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent segments the image into pixels and calculates inclusion probabilities for each pixel pair between cancer and peripheral tissue regions. This segmentation enables accurate visualization of invasion ranges at pixel level, while the systematic approach to probability calculation manages computational complexity through structured processing.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent transforms the problem into calculating probability values for each pixel pair, changing the parameter from binary inclusion to probabilistic measurement. This parameter transformation achieves high accuracy in invasion range visualization while the probability framework provides a manageable computational approach.

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentUS12602778B2Information processing apparatus, information processing method, and information processing program

Publication Date: 2026.04.14 FUJIFILM CORP

AI summary

A plurality of training data in which range information indicating a range in which a first region and a second region including at least a part of the first region are present is added to each of a plurality of training images each including the first region and the second region are acquired. For each pixel of the training image, a probability that the pixel is a portion of the first region that is not included in the second region is calculated by using an estimation model. A probability sum, which is a sum of the probabilities, is calculated for each of the plurality of training images. The estimation model is trained such that the probability sum calculated for each of training images in which the first region has the portion that is not included in the second region is increased and the probability sum calculated for each of training images in which the first region does not have the portion that is not included in the second region is zero.