Pathology Image Classification for Hypermutated Tumor Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for determining hypermutated cancer are labor-intensive and time-consuming, and they fail to accurately identify all hypermutated cancers, particularly through genetic analysis and mismatch repair mechanism examination.
Innovation Solution
A system utilizing machine learning to analyze hematoxylin and eosin stained pathological images, converting RGB colors to Z values, and dividing images into tiles for training a determination model to quickly and accurately classify hypermutated cancers based on image data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If comprehensive gene analysis is performed to determine hypermutated cancer, then measurement precision is improved, but loss of time and productivity deteriorate
Solution Approach 1:
The patent replaces the mechanical/genetic analysis system with an optical imaging system. Instead of performing comprehensive gene analysis through laboratory techniques, the invention uses whole slide images captured by a microscope and processed by deep learning algorithms to predict hypermutation status, thereby eliminating time-consuming wet lab procedures while maintaining diagnostic accuracy.
Solution Approach 2:
The patent creates a digital copy (whole slide image) of the pathological section that can be analyzed computationally. This digital representation allows the system to infer genetic characteristics (hypermutation status) from morphological features visible in the stained tissue image, avoiding the need for physical genetic material extraction and analysis.
2Measurement precision
If comprehensive gene analysis is performed to determinehypermutated cancer, then measurement precision is improved, but device complexity and ease of operation deteriorate
Solution Approach 1:
The patent makes the pathological imaging system multi-functional by enabling it to not only visualize tissue morphology but also predict genetic characteristics (hypermutation status) through integrated deep learning algorithms. This universal approach allows a single imaging system to perform both traditional histological examination and molecular subtype classification.
Solution Approach 2:
The patent introduces a deep learning model as an intermediary between the visual image data and the genetic characteristic prediction. This intermediary translates morphological features from whole slide images into predictions of hypermutation status, bridging the gap between visual pathology and molecular genetics without requiring direct genetic analysis infrastructure.
3Measurement precision
If mismatch repair mechanism examination is performed to determinehypermutated cancer, then measurement precision is improved partially, but productivity deteriorates
Solution Approach 1:
The patent performs preliminary action by capturing whole slide images during routine pathological examination, storing them for later analysis. The deep learning model can then process these pre-captured images to predict hypermutation status, allowing the system to prepare diagnostic data in advance without adding time to the immediate diagnostic workflow.
Solution Approach 2:
The patent enables continuous useful action by allowing the deep learning model to process multiple whole slide images sequentially without requiring repeated laboratory procedures. Once the imaging system captures the slides, the computational analysis can continuously evaluate them for hypermutation prediction, significantly increasing diagnostic throughput compared to individual genetic testing.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A system, a computer program, and a method for determining hypermutated cancer with higher accuracy than before is provided. The system for determining hypermutated cancer comprises, an input unit configured to be capable of inputting a plurality of first image data, a plurality of second image data and a plurality of third image data, wherein the first image data represents an image of a pathological section of stained hypermutated cancer, the second image data represents an image of a pathological section of cancer which is not hypermutated, and is stained same as the pathological section of the first image data, and the third image data represents an image of a pathological section of cancer which is newly determined whether hypermutated or not, and is stained same as the pathological section of the first image data; a holding unit configured to be capable of holding a first image data and a second image data; an image processing unit which is configured to be capable of performing a Z value conversion process for the first image data, the second image data and the third image data, converting each RGB color in each pixel into Z value in the CIE color system based on the entire color distribution of the first image data, the second image data or the third image data; wherein the image processing unit is furthermore configured to be capable of performing a division process dividing at least one of the first image data, the second image data, and third image data input into the input unit; a machine learning execution unit configured to be capable of generating a determination model determining whether a cancer is hypermutated or not, using the first image data and the second image data converted by the Z value conversion process and held by the holding unit as training data; and a determining unit configured to be capable of determining whether the third image data represents an image of hypermutated cancer or not, by inputting the third image data converted by the Z value conversion process into the determination model.