Pathological Image Analysis Using Dual AI Models

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing methods for diagnosing pathological images are prone to human subjective errors, especially when dealing with large quantities of images, which can lead to inaccurate diagnoses.

Innovation Solution

A method involving the use of two analysis models: a first analysis model for classifying image blocks based on the quantity or area of suspicious lesion components, and a second analysis model for detecting the location of suspicious lesion components in image blocks classified as second-type, to obtain a final analysis result for the pathological image.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If a doctor manually diagnoses pathological images, then the diagnosis can be performed with human judgment, but determining errors caused by human subjective factor occur and accuracy decreases

Engineering Contradiction:
Improvediagnostic accuracyVSAvoidconsistency of diagnosis
Core Design Contradiction:
Measurement precisionVSReliability

Solution Approach 1:

The patent replaces the mechanical system of manual visual inspection by doctors with an automated image processing system using deep learning models. The system segments pathological images into multiple blocks, processes each block through trained neural networks to identify lesion components, and aggregates results to provide automated diagnosis. This substitution eliminates human subjective factors and determines errors, achieving consistent and reliable diagnostic results across different users and occasions.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

2Measurement precision

If an experienced doctor performs diagnosis, then accuracy may be improved, but the process becomes dependent on doctor experience and is not easily replicable

Engineering Contradiction:
Improvediagnostic accuracyVSAvoiddependence on doctor experience
Core Design Contradiction:
Measurement precisionVSAdaptability or versatility

Solution Approach 1:

The patent transforms the diagnostic process from being dependent on variable human experience parameters to using fixed, learnable parameters embedded in deep learning models. The system trains neural networks on large datasets of pathological images with known outcomes, allowing the models to learn optimal diagnostic parameters automatically. This enables any user to achieve consistent diagnostic accuracy without requiring years of experience, as the system replicates expert knowledge through trained parameters.

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The patent creates a digital copy of expert diagnostic knowledge through trained deep learning models. By training on extensive datasets annotated by experienced pathologists, the system captures and replicates expert judgment patterns, feature recognition abilities, and diagnostic reasoning processes. This copied knowledge is then embedded in the model parameters, allowing the system to perform at expert levels without requiring the actual experts to be present, thereby eliminating dependence on specific individuals' experience.

Inventive Principle:
Principle #26Copying

3Measurement precision

If multiple analysis models are used to process image blocks, then diagnostic accuracy is improved, but the system complexity increases

Engineering Contradiction:
Improveanalysis accuracyVSAvoidsystem complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent divides the complex task of diagnosing entire pathological images into smaller, manageable segments by splitting images into multiple blocks. Each block is independently processed by analysis models to identify lesion components locally. This segmentation approach allows the use of multiple specialized analysis models without overwhelming system complexity, as each model only needs to handle local block analysis rather than entire images. The results are then aggregated to form the overall diagnosis, maintaining high accuracy while managing complexity through modular processing.

Inventive Principle:
Principle #1Segmentation

Data Source

PatentEP3971762B1Method, device and system for processing image
Publication Date: 2025.04.23 HUAWEI CLOUD COMPUTING TECHNOLOGIES CO LTD
  • EP3971762B1 patent drawingFigure 1~2
  • EP3971762B1 patent drawingFigure 3
  • EP3971762B1 patent drawingFigure 4~5

AI summary

This application relates to the artificial intelligence field, and provides an image processing method, an apparatus, and a system. The method includes: obtaining a plurality of image blocks, where the plurality of image blocks are obtained by segmenting a to-be-analyzed pathological image; inputting the plurality of image blocks to a first analysis model to obtain a first analysis result, where the first analysis model classifies each of the plurality of image blocks based on a quantity or an area of suspicious lesion components; inputting at least one second-type image block in the first analysis result to a second analysis model to obtain a second analysis result, where the second analysis model analyzes a location of a suspicious lesion component of each input second-type image block; and obtaining a final analysis result of the pathological image based on the first analysis result and the second analysis result. In the method, the pathological image is analyzed by using the two analysis models. Therefore, determining errors, caused by a human subjective factor, of the analysis result of the pathological image can be reduced.