Tissue Image Processing with Federated Learning for HER2 Assessment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for processing tissue images, particularly for evaluating HER2 expression, rely heavily on manual identification and lack accuracy, and deep learning models face challenges due to data scarcity and privacy issues, leading to inconsistent performance across institutions.
Innovation Solution
A method utilizing federated learning to construct a tumor segmentation model through staining component enhancement, which includes color deconvolution, random augmentation, and preprocessing, enabling accurate HER2 evaluation across institutions without compromising privacy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual identification methods are used for processing tissue images, then human experience can be applied to determine gene expression levels, but the accuracy and consistency of evaluation are insufficient
Solution Approach 1:
The patent replaces manual mechanical identification with an automated deep learning system. The tumor segmentation model automatically processes tissue images to identify tumor regions and evaluate HER2 expression, eliminating human subjectivity and improving both accuracy and consistency of evaluations across different institutions and pathologists.
Solution Approach 2:
The patent transforms the evaluation process by changing from qualitative manual assessment to quantitative automated analysis. The system uses standardized image processing parameters and algorithms to objectively measure HER2 expression levels, converting subjective pathologist judgments into consistent quantifiable results.
2Measurement precision
If deep learning models are trained using centralized data collection across institutions, then model performance can be improved, but data privacy and security are compromised
Solution Approach 1:
The patent segments the centralized training process into distributed local training units at each institution. Each institution trains models locally on its own data without sharing raw images, then contributes only model parameters or gradients to the federated learning server. This segmentation maintains data privacy while achieving collaborative model improvement.
Solution Approach 2:
The federated learning server acts as an intermediary that coordinates training across institutions without accessing actual patient data. It aggregates model updates from participating institutions and distributes improved models back, enabling collaborative learning while preserving data privacy through this intermediate coordination layer.
3Adaptability or versatility
If staining components are not enhanced in the dataset, then the data processing is simpler, but the model's ability to accurately evaluate HER2 expression across different institutions is reduced
Solution Approach 1:
The patent performs staining component enhancement as a preliminary action during data preprocessing. By applying color deconvolution and staining intensity adjustment before model training, the system standardizes images from different institutions, improving model generalizability. This preliminary processing ensures consistent input quality without requiring complex modifications during training.
Data Source
AI summary
The disclosure relates to the field of image processing technology and provides a method and system for processing tissue images on whole-slide imaging. The method includes: obtaining a target image; obtaining a dataset, enhancing the staining components of the dataset, using the enhanced dataset for federated learning to construct a tumor segmentation model, wherein the dataset includes several test images, and the test images are immunohistochemical images; inputting the target image into the tumor segmentation model to determine a tumor area within the target image; performing cell membrane staining detection on the tumor area, classifying the tumor area based on the integrity of the cell membrane and the level of staining, and grading based on the classification result. The disclosure can more accurately evaluate the level of HER2 in tissue images, optimize target images obtained from different institutions, and achieve cross-institutional model training without compromising user privacy.