Duct Tissue Image Classification Using Morphology-Based Feature Regions

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing duct tissue feature extraction technologies rely on the overall quantity of nuclei, which is large and difficult to count accurately, leading to low accuracy in reflecting the features of duct tissue.

Innovation Solution

A method using computer vision and an intelligent microscope to adaptively select feature regions based on duct morphology, extract cell features, and integrate them with sieve pore features for accurate duct tissue classification, employing a duct tissue classifier trained with an SVM model.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of manufacture

If the overall quantity of nuclei is used as the main feature for duct tissue extraction, then the feature extraction process is simple, but the accuracy of reflecting duct tissue features is low due to the large and difficult-to-count number of nuclei

Engineering Contradiction:
Improvesimplicity of feature extraction processVSAvoidaccuracy of duct tissue feature extraction
Core Design Contradiction:
Ease of manufactureVSMeasurement precision

Solution Approach 1:

The patent segments the duct tissue image into multiple sub-regions and further divides nuclei into different categories (epithelial cell nuclei, stromal cell nuclei, inflammatory cell nuclei). This segmentation allows for more precise counting and analysis of specific cell types rather than treating all nuclei as a single group, thereby improving measurement precision while maintaining computational feasibility

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent introduces an artificial intelligence model as an intermediary to automatically identify, segment, and count different types of nuclei in duct tissue images. This AI intermediary handles the complex task of distinguishing and counting various cell types, resolving the contradiction by providing accurate measurements without requiring manual counting of the large total number of nuclei

Inventive Principle:
Principle #24Intermediary (Mediator)

2Measurement precision

If multiple feature regions are selected based on duct morphology, then the accuracy of duct tissue classification is improved, but the complexity of the processing system increases

Engineering Contradiction:
Improveaccuracy of duct tissue classificationVSAvoidcomplexity of processing system
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent dynamically adjusts the selection of feature regions based on the specific morphology of each duct tissue sample. The system adaptively determines which regions to analyze and what features to extract according to the unique characteristics of each tissue type, allowing high accuracy without requiring a fixed complex processing framework for all cases

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The patent employs a universal artificial intelligence model that can handle multiple tasks: image segmentation, feature extraction, cell counting, and classification. This multi-functional AI system replaces multiple specialized processing components, achieving high classification accuracy while actually reducing overall system complexity through consolidation

Inventive Principle:
Principle #6Universality (Multi-functionality)

Data Source

PatentEP4016380B1Computer vision based catheter feature acquisition method and apparatus and intelligent microscope
Publication Date: 2026.04.22 TENCENT TECHNOLOGY (SHENZHEN) CO LTD
  • EP4016380B1 patent drawingFigure 1~2
  • EP4016380B1 patent drawingFigure 3~5
  • EP4016380B1 patent drawingFigure 6

AI summary

A computer vision technology based catheter tissue feature acquisition method and apparatus, an intelligent microscope, a storage medium and a computer device. The method comprises: acquiring an image containing a catheter tissue (S201); in an image area corresponding to the catheter tissue of the image, determining at least two feature acquisition areas compatible with the catheter form of the catheter tissue(S202); acquiring cell features of cells of the catheter tissue in each of the feature acquisition areas(S203); and acquiring features of the catheter tissue on the basis of the cell features of the cells of the catheter tissue in each of the feature acquisition areas (S204).