Artifact Pixel Detection in Pathology Images With Multi-Resolution ML
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine-learning models for detecting artifact pixels in images used in digital pathology are inefficient due to challenges in generating accurate training data, handling various staining patterns, and integrating with digital pathology analyses without significant computing resource consumption.
Innovation Solution
A method involving image pre-processing algorithms and machine-learning models, such as U-Net, to generate and modify labels for training data, and apply convolutional layers at different resolutions to efficiently detect artifact pixels, reducing processing time and resource consumption.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If machine-learning models are trained to detect artifacts in IHC images, then detection accuracy is improved, but processing time and computing resources increase
Solution Approach 1:
The patent divides the image processing task into multiple resolutions, training and applying models at different scales. The system segments the detection process into coarse (lower resolution) and fine (higher resolution) stages, allowing artifact detection without requiring all processing to occur at maximum resolution, thus reducing overall processing time while maintaining accuracy.
Solution Approach 2:
The patent changes the resolution parameter dynamically throughout the pipeline. Training occurs at multiple resolutions, and the model applies different resolution levels based on image characteristics. This parameter change allows the system to achieve high detection accuracy when needed while reducing processing time for images that can be adequately evaluated at lower resolutions.
2Measurement precision
If machine-learning models are trained to detect artifacts in IHC images, then detection accuracy is improved, but computing resources increase
Solution Approach 1:
The computing workload is segmented across multiple resolution levels. Instead of training and applying a single high-resolution model, the system distributes computational tasks across different resolution tiers, reducing the peak computational resource requirements while maintaining comprehensive detection capability.
Solution Approach 2:
The system dynamically adjusts computational resource allocation by changing the resolution parameter. For images with artifacts or at critical decision points, the system allocates higher computational resources for high-resolution processing. For images that can be confidently evaluated at lower resolutions, computational resources are reduced, optimizing the balance between detection accuracy and resource consumption.
3Stability of the object's composition
If training data is generated using image pre-processing algorithms, then training data consistency is improved, but false labels occur reducing accuracy
Solution Approach 1:
The patent implements feedback mechanisms where the machine-learning model evaluates the labels generated by pre-processing algorithms and corrects false labels. The model uses its own predictions to validate and refine the training data, creating a self-correcting system that improves label accuracy while maintaining the consistency benefits of automated pre-processing.
Solution Approach 2:
The machine-learning model serves as an intermediary between the pre-processing algorithm and the final training data. It receives labels from the pre-processing algorithm, evaluates their accuracy, and produces corrected labels for training. This intermediary role allows the system to benefit from automated pre-processing consistency while filtering out false labels through the model's intelligent evaluation.
Data Source
AI summary
Method and systems for of using a machine-learning model to detect predicted artifacts at a target image resolution are provided. A machine-learning model trained to detect artifact pixels in images at a target image resolution is accessed. An image depicting at least part of the biological sample at an initial image resolution can be converted at the target image resolution. The machine-learning model is applied to the converted image to identify one or more artifact pixels from the converted image. Method and systems for training the machine-learning model to detect predicted artifacts at the target image resolution are also provided.