Whole Slide Image AI Quality Check for Pathology Artifacts
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing digital pathology systems face challenges in efficiently detecting and correcting quality artifacts in whole slide images, such as pen marks, air bubbles, out-of-focus areas, tissue folds, and scanning errors, which can lead to biased or inaccurate diagnoses.
Innovation Solution
An automated quality control system using an ensemble of deep learning models, including UNet and ResNet architectures, to analyze thumbnail images derived from whole slide images, detects and classifies various artifacts, and optionally rescans the slide if issues are found, providing objective and efficient quality control.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual quality control is used to detect artifacts in whole slide images, then detection accuracy can be maintained, but time consumption and resource requirements increase significantly
Solution Approach 1:
The patent replaces manual visual inspection with an automated deep learning-based detection system. The system uses convolutional neural networks to automatically identify artifacts such as pen marks, air bubbles, out-of-focus areas, tissue folds, and scanning errors in whole slide images, eliminating the need for human pathologists to manually review each image for quality control.
Solution Approach 2:
The system enables self-service quality control by automatically detecting and classifying artifacts without requiring human intervention. The deep learning model independently performs the quality assessment function, allowing the imaging system to self-evaluate and identify issues in the generated images.
2Manufacturing precision
If high-resolution imaging is used to capture whole slide images, then image quality improves, but detection of certain artifacts becomes more difficult and processing complexity increases
Solution Approach 1:
The patent segments the artifact detection task into multiple categories (first class artifacts: pen marks, air bubbles, out-of-focus, tissue folds; second class artifacts: misplaced coverslip, periodic scanning errors). The deep learning model is trained to distinguish between different artifact types and also generates tissue masks to differentiate artifacts from actual tissue structures, making detection more manageable and accurate.
Solution Approach 2:
The system introduces an intermediate representation in the form of a tissue mask that separates tissue regions from non-tissue regions. This mask serves as an intermediary that helps the system identify artifacts by comparing it with the original high-resolution image, thereby simplifying the detection of subtle artifacts without sacrificing resolution.
3Productivity
If automated deep learning models are used to detect artifacts, then processing speed and efficiency improve, but system complexity and computational resources increase
Solution Approach 1:
The deep learning model serves multiple functions within a single system: it detects various types of artifacts (pen marks, air bubbles, out-of-focus areas, tissue folds, misplaced coverslips, periodic scanning errors), generates tissue masks for comparison, and provides classification of artifact types. This multi-functionality reduces the need for separate specialized systems for each detection task.
Solution Approach 2:
The system performs preliminary quality control by automatically detecting artifacts before the images are used for diagnostic purposes. The deep learning model pre-screens the images, identifying and classifying artifacts in advance, which allows for early intervention or rejection of compromised images before they enter the diagnostic workflow.
Data Source
AI summary
Techniques of automated quality control for digital pathology whole slide images are presented. The techniques include obtaining a thumbnail image derived from a whole slide image of a pathology slide; determining whether the whole slide image includes an artifact in a first class of artifacts by providing the thumbnail image to an electronic neural network trained to detect artifacts in the first class of artifacts by analyzing a plurality of labeled training thumbnail images; generating a tissue mask representing tissue depicted in the thumbnail image; determining whether the whole slide image includes an artifact in a second class of artifacts by performing a comparison using the tissue mask; and providing an indication of whether the whole slide image includes an artifact in the first class of artifacts or an artifact in the second class of artifacts.