Biomarker Expression Prediction from H&E Images for Faster Trial Screening
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The identification of patients eligible for clinical trials based on biomarker expression levels is time-consuming and costly due to the need for immunohistochemistry (IHC) screening, which can take several days and often results in a significant percentage of ineligible patients, slowing down recruitment.
Innovation Solution
A system utilizing machine learning models to predict biomarker expression levels from H&E stained images, enabling virtual staining and image preprocessing techniques to enhance computational efficiency and accuracy, allowing for the prediction of continuous biomarker levels without the need for extensive IHC screening.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If immunohistochemistry (IHC) screening is performed to identify patients eligible for clinical trials, then measurement precision of biomarker expression levels is improved, but loss of time and productivity deteriorate due to several days processing time
Solution Approach 1:
The patent creates a virtual copy of the IHC staining process by training a machine learning model to predict IHC biomarker expression levels from H&E stained images. This virtual staining approach replicates the information obtained from actual IHC staining without requiring the time-consuming laboratory process, thereby maintaining measurement precision while eliminating the time delay
Solution Approach 2:
The system performs preliminary screening using H&E stained images and machine learning predictions before committing to full IHC staining. This preliminary action identifies candidate patients who are likely to be IHC-positive, allowing the laboratory to perform IHC staining only on selected cases rather than all patients, thus reducing overall processing time
2Measurement precision
If immunohistochemistry (IHC) screening is performed to ensure accurate biomarker measurement, then measurement precision is improved, but device complexity and cost increase
Solution Approach 1:
The machine learning model serves multiple functions: it acts as a virtual IHC staining device, a prediction engine for biomarker levels, and a filtering mechanism for patient selection. By making the system multi-functional, the patent reduces reliance on separate complex IHC staining infrastructure while maintaining measurement precision through the unified model
3Reliability
If traditional IHC screening process is used for patient selection, then reliability of patient eligibility determination is improved, but productivity of clinical trial recruitment deteriorates
Solution Approach 1:
The virtual staining model creates a reliable copy of the IHC screening process that can be executed rapidly on existing H&E images. This digital copy maintains the diagnostic reliability of traditional IHC while enabling high-throughput screening of large patient cohorts, thereby increasing recruitment productivity without sacrificing eligibility determination accuracy
Data Source
AI summary
One or more methods of predicting expression levels. At least one of the methods includes preprocessing image data representing at least one biological image of a patient to generate preprocessed image data representing at least one preprocessed biological image of the patient; and applying a trained machine learning model to the preprocessed image data to predict, based at least partially on the at least one preprocessed biological image, an expression level of a biological indicator.