Graph-Based Tumour Histology Analysis for Recurrence Risk Stratification

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current clinical pathways for predicting the risk of cancer recurrence in early-stage solid tumors rely heavily on traditional indicators and clinical expertise, lacking precise and personalized predictive markers, leading to uncertain necessity and efficacy of post-surgery treatments like adjuvant chemotherapy.

Innovation Solution

A computer-implemented method using histopathological image analysis, converting histological images into graph representations and processing them with a neural network comprising a graph isomorphism network and convolutional neural network to determine a risk factor for cancer recurrence, incorporating clinical parameters for improved accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If traditional indicators and clinical expertise are used to estimate risk, then the assessment process is simple and accessible, but the measurement precision and reliability of prognosis prediction are insufficient

Engineering Contradiction:
Improveprognosis prediction precisionVSAvoidanalysis system complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent replaces traditional manual clinical assessment with an automated deep learning system that processes histopathological images. The mechanical/manual process of clinical judgment is substituted with an artificial intelligence system comprising convolutional neural networks and attention mechanisms that automatically extract features and predict prognosis from digital pathology images, thereby improving measurement precision while managing complexity through automation

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The patent transforms the assessment from using traditional clinical parameters (TNM stage, lymph node count) to using high-dimensional image features extracted by deep learning models. The system changes the parameter space from discrete clinical indicators to continuous pixel-level features, enabling more precise prognosis prediction through multi-scale feature extraction and attention-based feature selection

Inventive Principle:
Principle #35Parameter changes

2Reliability

If traditional clinical indicators are used, then the ease of operation is maintained, but the reliability of treatment decision-making deteriorates due to disease heterogeneity

Engineering Contradiction:
Improvetreatment decision reliabilityVSAvoidneural network architecture complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent segments the histopathological image into multiple regions and scales, processing different portions of the tissue sample independently through separate neural network pathways. This segmentation allows the system to capture heterogeneous features across different tumor regions, improving treatment decision reliability by accounting for intra-tumoral variability while managing complexity through modular network architecture

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent adds spatial and hierarchical dimensions to the analysis by processing images at multiple scales and incorporating attention mechanisms that weigh different spatial regions differently. This dimensional expansion allows the system to capture both global tissue architecture and local cellular features, enhancing reliability without requiring a single overly complex model

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

3Measurement precision

If comprehensive histopathological analysis is performed, then the measurement precision of risk factors improves, but the loss of time in processing increases

Engineering Contradiction:
Improverisk factor prediction accuracyVSAvoidimage processing time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent performs preliminary processing of histopathological images by pre-training deep learning models on large datasets and pre-extracting feature representations. This preliminary action allows the system to quickly apply trained models to new patient samples, reducing processing time while maintaining high measurement precision through transfer learning and pre-computed feature hierarchies

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent extracts only the most relevant features from the full histopathological image using attention mechanisms and feature selection algorithms. Instead of processing all image data equally, the system identifies and extracts discriminative features that are most predictive of prognosis, thereby reducing processing time while maintaining or improving measurement precision through focused analysis

Inventive Principle:
Principle #2Taking out (Extraction)

Data Source

PatentEP4645338A1Histopathology-based solid tumour analysis
Publication Date: 2025.11.05 OXFORD CANCER BIOMARKERS
  • EP4645338A1 patent drawingFigure 1
  • EP4645338A1 patent drawingFigure 2
  • EP4645338A1 patent drawingFigure 3

AI summary

A computer-implemented method is provided, for predicting a risk factor for a patient based on histopathological image analysis. The method includes: receiving (110) at least one histological image of a solid tumour; converting (150) the histological image into a graph representation; processing (160) the graph representation using a neural network, wherein the neural network comprises a graph isomorphism network and a convolutional neural network; and determining (162) the risk factor based on an output of the neural network. Also provided is a method of training one or more neural networks for use in such a method.