MRI Trained Model for Dementia Progression Prediction

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing technologies lack the capability to predict the degree of progression of dementia, which is crucial for supporting early diagnosis and selecting appropriate treatment methods.

Innovation Solution

A non-transitory computer-readable storage medium storing a program that includes a trained model for predicting the degree of dementia progression using MRI images and functional change information, employing machine learning techniques such as SVM, DNN, CNN, and RNN to analyze brain atrophy over time.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If machine learning methods are used to predict dementia progression, then diagnostic accuracy is improved, but device complexity increases

Engineering Contradiction:
Improvediagnostic accuracyVSAvoiddevice complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

A trained machine learning model acts as an intermediary between MRI image data and diagnostic conclusions. The model processes complex image analysis tasks internally, providing simplified prediction outputs (dementia progression degree) to clinicians without requiring them to understand the underlying complex algorithms

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent extracts and isolates the complex machine learning inference process into a separate trained model component. This allows the core diagnostic functionality to be improved through advanced algorithms while keeping the overall system architecture manageable by separating the complex processing from the user interface and data acquisition layers

Inventive Principle:
Principle #2Taking out (Extraction)

2Measurement precision

If multiple machine learning algorithms are employed, then prediction accuracy is improved, but ease of operation deteriorates

Engineering Contradiction:
Improveprediction accuracyVSAvoidease of operation
Core Design Contradiction:
Measurement precisionVSEase of operation

Solution Approach 1:

Multiple machine learning algorithms (SVM, DNN, CNN, RNN) are merged into a single integrated trained model. The system presents a unified interface that automatically selects and combines these algorithms, eliminating the need for operators to manually configure or switch between different complex algorithms while maintaining high prediction accuracy

Inventive Principle:
Principle #5Merging (Combining)

3Measurement precision

If brain atrophy analysis is performed over time, then measurement precision is improved, but loss of time increases

Engineering Contradiction:
Improvemeasurement precisionVSAvoidloss of time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The system performs preliminary analysis by automatically selecting relevant time points and preprocessing MRI images before detailed atrophy measurement. This preliminary action reduces the computational burden on subsequent processing steps, enabling long-term longitudinal analysis without proportionally increasing total processing time

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent implements optimized processing that skips redundant analysis steps. By using the trained model to quickly evaluate and filter less relevant temporal data, the system rushes through routine processing stages while maintaining comprehensive long-term analysis, thereby reducing overall processing time while preserving measurement precision

Inventive Principle:
Principle #21Skipping (Rushing through)

Data Source

PatentEP3918987B1Trained model
Publication Date: 2025.12.24 FUJIFILM CORP
  • EP3918987B1 patent drawingFigure 1
  • EP3918987B1 patent drawingFigure 2
  • EP3918987B1 patent drawingFigure 3~4

AI summary

There is provided a medical information acquisition device including an information acquisition unit that acquires functional change information obtained on the basis of a reference image and a past image acquired by capturing images of the same subject at a reference time and a past time closer to the past than the reference time, respectively, using a trained model.