Graph Neural Network for Tissue Slide Image Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional machine learning techniques face difficulties in extracting high-level pathological information from whole slide images (WSI) for diagnosing anticancer drug responses, requiring medical expertise and being resource-intensive, while deep learning methods are inefficient due to high capacity and resource requirements.
Innovation Solution
A method and system using a graph neural network (GNN) to analyze images by extracting and sampling graphic data from tissue slide images, training parameters for readout and update functions, and deriving inference output data, reducing the need for expertise and resource usage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If traditional machine learning techniques are used to extract features from WSI information, then simple mathematical models can be used, but difficulty in extracting high-level pathological information such as anticancer drug response effects occurs
Solution Approach 1:
The patent segments the WSI analysis process into multiple levels: traditional machine learning handles simple feature extraction while deep learning models (CNNs, RNNs, GNNs) handle high-level pathological information extraction. This segmentation allows each method to operate within its optimal capability range, resolving the contradiction between ease of implementation and diagnostic accuracy.
Solution Approach 2:
The patent introduces hybrid models that combine traditional machine learning with deep learning components. These intermediary systems allow traditional ML to handle straightforward tasks while deep learning components process complex pathological patterns, achieving both ease of manufacture and high measurement precision through coordinated operation of different methodologies.
2Measurement precision
If deep learning techniques are used to input all pixels of WSI into a neural network, then high-level information can be extracted, but speed and efficiency are greatly degraded and much more resources are needed
Solution Approach 1:
The patent segments the image processing workflow into preprocessing/extraction phases that reduce data volume before deep learning analysis. By extracting relevant features and reducing WSI to smaller representative images or feature vectors beforehand, the system maintains deep learning's high diagnostic accuracy while significantly improving processing speed and reducing computational resource requirements.
Solution Approach 2:
The patent performs preliminary actions by pre-processing WSIs to extract and consolidate relevant information before feeding data to deep learning models. This includes feature extraction, image tiling, and data augmentation techniques that prepare the data in advance, allowing the neural network to process smaller, more focused inputs rather than raw pixel data, thereby improving efficiency without sacrificing accuracy.
Data Source
AI summary
An image analysis method and an image analysis system are disclosed. The method may include extracting training raw graphic data including at least one first node corresponding to a plurality of histological features of a training tissue slide image, and at least one first edge defined by a relationship between the histological features and generating training graphic data by sampling the first node of the training raw graphic data. The method may also include determining a parameter of a readout function by training a graph neural network (GNN) using the training graphic data and training output data corresponding to the training graphic data, and extracting inference graphic data including at least one second node corresponding to a plurality of histological features of an inference tissue slide image, and at least one second edge decided by a relationship between the histological features of the inference tissue slide image.