Artifact Exclusion in Digital Pathology Image Normalization

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Digital pathology images often contain artifacts such as out-of-focus, staining, pressure, and fixation artifacts, which can lead to sub-optimal image processing and analysis, especially when using automatic image analysis programs.

Innovation Solution

A system that detects and excludes artifact regions from the image processing parameter determination, allowing for improved image processing by using a parameter based on relevant image areas, and provides a user interface for visualization and control of the detected regions.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If image processing is performed on the complete tissue including artifacts, then processing coverage is maximized, but measurement precision deteriorates due to distortion from undesirable regions

Engineering Contradiction:
Improveparameter determination accuracyVSAvoidprocessing coverage
Core Design Contradiction:
Measurement precisionVSArea of stationary object

Solution Approach 1:

The image processing method segments the tissue image into multiple regions based on staining characteristics. By dividing the image into different stained and unstained regions, the system can selectively process only the relevant stained areas for parameter determination, excluding artifact-containing unstained regions from influencing the normalization parameters. This segmentation enables precise parameter measurement while maintaining focus on diagnostically relevant areas.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The invention applies different processing qualities to different regions of the image. Stained regions undergo full normalization and parameter analysis, while unstained regions are identified and excluded from parameter determination. This local quality approach ensures that high-precision processing is applied only where necessary (in stained regions containing diagnostic information) while avoiding the distortion caused by artifact-prone unstained regions.

Inventive Principle:
Principle #3Local quality

2Productivity

If automatic image analysis is performed on images containing artifacts, then productivity is improved, but reliability deteriorates due to difficulty in artifact recognition

Engineering Contradiction:
Improveanalysis throughputVSAvoiddiagnosis accuracy
Core Design Contradiction:
ProductivityVSReliability

Solution Approach 1:

The system performs preliminary identification of stained and unstained regions before conducting the main parameter determination and analysis. By pre-segmenting the image and marking unstained regions that likely contain artifacts, the system prepares the data structure in advance to exclude these regions from subsequent processing. This preliminary action enables automatic analysis to proceed efficiently while maintaining reliability by preventing artifact contamination of diagnostic parameters.

Inventive Principle:
Principle #10Preliminary action

3Manufacturing precision

If normalization is performed on the complete field of view, then processing completeness is maximized, but manufacturing precision deteriorates due to inconsistent staining in artifact regions

Engineering Contradiction:
Improvenormalization consistencyVSAvoidnormalization coverage
Core Design Contradiction:
Manufacturing precisionVSArea of stationary object

Solution Approach 1:

The normalization process is segmented to operate only on stained regions identified through color deconvolution and region classification. By separating stained and unstained areas, the system applies normalization consistently across all stained regions while excluding unstained regions that would introduce inconsistency. This segmented normalization maintains manufacturing precision by ensuring uniform processing of diagnostically relevant areas without contamination from artifact-prone regions.

Inventive Principle:
Principle #1Segmentation

Data Source

PatentEP2780888B1Processing an image containing one or more artifacts
Publication Date: 2019.08.14 KONINKLIJKE PHILIPS NV
  • EP2780888B1 patent drawingFigure 1
  • EP2780888B1 patent drawingFigure 2
  • EP2780888B1 patent drawingFigure 3~4

AI summary

A system for processing an image comprises a region detector (1). The region detector comprises an artifact detector (7) for detecting a region comprising an artifact in the image. The system comprises a parameter determining unit (2) for determining a parameter, based on a portion of the image excluding the detected region. The system comprises an image processing module (3) for processing the image, using the derived parameter. The system comprises a display unit (5) for displaying the processed image with an indication of the detected region. The parameter determined by the parameter determining unit (2) can comprise a normalization parameter, and the image processing module (3) can be arranged for performing a normalization of the image according to the normalization parameter.