Pathology Slide Severity Grading via Deep Learning Patch Analysis

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing histological severity grading systems for diseases like cancer face challenges in accurately determining the severity of biological tissues due to ambiguity in grade classification, leading to difficulties in training deep learning models to support a single grade, especially at tissue boundaries, which affects the accuracy of severity recognition and reporting.

Innovation Solution

A slide level severity determination method using a pre-trained deep learning model that divides pathology slide images into unit-sized partial images, calculates effective grades based on likelihood values and pre-assigned grade scores, and aggregates these to determine the overall severity of the slide image, employing techniques such as weighted averages or maximum values to refine severity grading.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If traditional histological severity grading systems are used to classify tissue severity, then a standardized grading framework is provided, but accuracy is reduced due to ambiguity in grade classification at tissue boundaries

Engineering Contradiction:
Improveseverity classification accuracyVSAvoidgrade determination consistency
Core Design Contradiction:
Measurement precisionVSReliability

Solution Approach 1:

The pathology slide image is divided into multiple partial images (patches or tiles), and each partial image is independently analyzed by the deep learning model to determine its severity grade. This segmentation allows for more granular severity assessment and reduces the ambiguity associated with classifying entire slides that contain heterogeneous tissue regions with different severity levels.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The invention changes the parameter of analysis from whole-slide level to patch-level, enabling the deep learning model to capture local tissue characteristics more accurately. By analyzing smaller regions independently, the model can determine severity grades with higher precision for each local area, and then aggregate these results to obtain an overall slide-level severity assessment.

Inventive Principle:
Principle #35Parameter changes

2Extent of automation

If deep learning models are trained to recognize severity grades in tissue images, then automated diagnosis is achieved, but accuracy decreases at tissue boundaries where grade classification is ambiguous

Engineering Contradiction:
Improveautomated severity determinationVSAvoidboundary tissue grade recognition accuracy
Core Design Contradiction:
Extent of automationVSMeasurement precision

Solution Approach 1:

By segmenting the whole slide into multiple partial images, the invention enables the deep learning model to process each region independently. This segmentation strategy allows the model to focus on local tissue characteristics without being confused by heterogeneous regions, thereby maintaining high accuracy in severity grade recognition even at tissue boundaries.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

Instead of attempting to classify the entire slide as a single homogeneous region, the invention applies partial action by analyzing only specific partial images that contain relevant diagnostic information. This approach allows the model to achieve high precision in severity determination by focusing on informative regions while excluding ambiguous boundary areas from the primary classification decision.

Inventive Principle:
Principle #16Partial or excessive action

3Measurement precision

If whole slide images are analyzed directly to determine disease severity, then comprehensive tissue assessment is possible, but computational complexity and processing time increase significantly

Engineering Contradiction:
Improveoverall severity determination accuracyVSAvoidprocessing time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The invention segments the large whole slide image into smaller partial images that can be processed independently and in parallel. This segmentation dramatically reduces the computational complexity and processing time compared to analyzing the entire slide as a single large image, while still maintaining comprehensive tissue assessment through aggregation of results from multiple patches.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The invention applies partial action by analyzing only representative partial images that contain diagnostic information rather than processing every pixel of the entire slide. This approach significantly reduces processing time and computational resources while maintaining the accuracy needed for clinical diagnosis by focusing on informative regions.

Inventive Principle:
Principle #16Partial or excessive action

Data Source

PatentEP4528751A1Patch level severity determination method, slide level severity determination method, and computing system for performing same
Publication Date: 2025.03.26 DEEP BIO
  • EP4528751A1 patent drawingFigure 1
  • EP4528751A1 patent drawingFigure 2~3
  • EP4528751A1 patent drawingFigure 4

AI summary

A patch level severity determination method, a slide level severity determination method, and a computing system for performing same are disclosed. According to one aspect of the present invention, provided is a method performed in a computing system including a deep-learning model pre-trained to provide determination results for partial images when each partial image obtained by dividing a pathology slide image is inputted, the method comprising the steps of: for each of a plurality of partial images obtained by dividing a pathology slide image, determining an effective grade for the partial image on the basis of a determination result for the partial image outputted by the deep-learning model that has received the partial image; and determining a slide level severity rating for the entire pathology slide image on the basis of the effective grade for each of the plurality of partial images that constitute the pathology slide image.