Tomographic Lesion Detection Using Multi-Model Probability Integration

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing lesion detection methods in medical images face challenges in achieving high accuracy due to varying pixel values between lesion and normal regions, especially when factors like fat accumulation affect the luminance, leading to reduced detection precision.

Innovation Solution

Generate multiple lesion identification models tailored to different medical findings, such as with and without fat accumulation, and integrate their probability outputs based on image feature amounts to enhance detection accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If a single trained model is used for lesion detection, then the device complexity is low, but the detection accuracy deteriorates when pixel value differences between lesion and normal regions are small

Engineering Contradiction:
Improvedetection accuracyVSAvoidmodel complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent divides the detection task into multiple specialized models, each trained on specific medical findings (e.g., liver cirrhosis, fatty liver). This segmentation allows each model to specialize in detecting lesions under particular conditions, improving overall detection accuracy while managing complexity through modular architecture

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The system dynamically changes the parameters (which model to use and how to integrate results) based on the input medical image characteristics. By analyzing image features and selecting appropriate models accordingly, the system adapts to different pixel value relationships and medical conditions, resolving the contradiction between simplicity and accuracy

Inventive Principle:
Principle #35Parameter changes

2Measurement precision

If multiple specialized models are used to improve detection accuracy under different medical findings, then the detection accuracy improves, but the device complexity increases

Engineering Contradiction:
Improvedetection accuracyVSAvoidnumber of models
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent implements a dynamic model selection and integration mechanism that determines which specialized models to apply based on the characteristics of the input medical image. This dynamic approach allows the system to use multiple models when needed for high accuracy while avoiding unnecessary complexity when a single model suffices

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The patent introduces an intermediary integration mechanism that combines the outputs of multiple specialized models. This mediator layer manages the complexity of having multiple models by providing a unified interface and systematic method for integrating their results, thereby improving detection accuracy without proportionally increasing overall system complexity

Inventive Principle:
Principle #24Intermediary (Mediator)

Data Source

PatentEP4524886B1Lesion detection method and lesion detection program
Publication Date: 2026.05.13 FUJITSU LTD
  • EP4524886B1 patent drawingFigure 1
  • EP4524886B1 patent drawingFigure 2
  • EP4524886B1 patent drawingFigure 3

AI summary

A lesion detection method for a computer to execute: a learning process of classifying first tomographic images into a plurality of first tomographic image groups, and generating first lesion identification models for identifying whether or not unit image region included in a tomographic image is a specific lesion region; and a lesion detection process of calculating a first image feature amount based on second tomographic images, acquiring a probability that each of the unit image regions included in the second tomographic images is the specific lesion region from each of the first lesion identification models, calculating, for each of the unit image regions included in the second tomographic images, an integration value by integrating the probabilities acquired from each of the plurality of first lesion identification models, and detecting the specific lesion region from the second tomographic images based on the integration value.