Machine Learning of Breast Images for Objective ILC Ground Truth
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The low inter-observer agreement in diagnosing invasive lobular carcinoma (ILC) due to its distinctive discohesive growth pattern and the challenge of using AI-based algorithms without reliable ground truth labeling complicates therapeutic decision-making.
Innovation Solution
A computer-implemented method using a trained machine learning module to detect CDH1 biallelic genetic inactivation from digital images of breast tissue, incorporating supplemental patient information, and determining the presence of ILC based on ground truth detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If AI-based algorithms are used to diagnose invasive lobular carcinoma, then diagnostic accuracy may be improved, but their performance depends on ground truth labeling which is currently unreliable
Solution Approach 1:
The patent introduces CDH1 biallelic mutation status as an intermediary objective marker to bridge the gap between subjective histologic diagnosis and AI algorithm training. This molecular marker serves as a reliable ground truth that mediates the relationship between phenotypic observation and genetic reality, enabling accurate AI model development without relying on unreliable expert labeling.
Solution Approach 2:
The patent replaces the mechanical/expert-based histologic diagnosis system with a molecular genetics-based detection system. By substituting the subjective visual assessment mechanism with objective CDH1 mutation detection, the system achieves reliable ground truth labeling that can consistently train and validate AI algorithms for ILC diagnosis.
2Adaptability or versatility
If histologic subtyping is used for therapeutic decision making, then treatment personalization may be achieved, but low inter-observer agreement complicates the process
Solution Approach 1:
The patent changes the diagnostic parameter from subjective histologic morphology to objective molecular genetics (CDH1 biallelic mutation status). This parameter transformation converts the diagnostic process from one susceptible to observer variability into one based on immutable genetic facts, thereby enabling reliable therapeutic decision making through molecular phenotyping.
3Productivity
If machine learning models are trained on histologic images alone, then diagnostic speed may be improved, but accuracy is limited without molecular validation
Solution Approach 1:
The patent merges histologic image analysis with molecular genetics data into a unified diagnostic framework. By combining the speed advantages of AI image processing with the accuracy advantages of molecular validation, the system achieves both rapid and accurate diagnosis through integrated multi-modal analysis.
Data Source
AI summary
A computer-implemented method may diagnose invasive lobular carcinoma. The method may include receiving one or more digital images into a digital storage device, applying a trained machine learning module to detect a presence or absence of CDH1 biallelic genetic inactivation and/or CDH1 biallelic mutation from the received one or more digital images, and determining whether the patient has invasive lobular carcinoma using the detected presence or absence of the CDH1 biallelic genetic inactivation and/or CDH1 biallelic mutation as ground truth. The one or more digital images may include images of breast tissue of a patient.