TIL Spatial Feature Extraction for NSCLC Immunotherapy Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for predicting response to immune checkpoint inhibitors in non-small cell lung cancer (NSCLC) are limited by the spatial and temporal heterogeneity of PD-L1 expression, resulting in only 20% of patients showing significant benefit, necessitating a more accurate and reliable predictive approach.
Innovation Solution
The use of computer-extracted features from digitized hematoxylin and eosin (H&E) stained images to quantify the spatial arrangement of tumor-infiltrating lymphocytes (TILs) and train machine learning classifiers to distinguish responders from non-responders, employing graph-based methods to capture the spatial interplay between TILs and tumor cells.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If PD-L1 expression is used as the current gold standard for predicting response, then the prediction method is simple and widely available, but the prediction accuracy is limited due to spatial and temporal heterogeneity
Solution Approach 1:
The patent segments the tissue sample into multiple regions and evaluates PD-L1 expression in each region separately, then integrates these regional assessments to account for spatial heterogeneity. This segmentation approach improves prediction accuracy by capturing the variability across different tissue areas rather than relying on a single bulk assessment.
Solution Approach 2:
The patent transitions from traditional two-dimensional histological section analysis to three-dimensional spatial mapping of PD-L1 expression across tissue architecture. By incorporating spatial coordinates and regional distribution patterns, the method adds a spatial dimension that captures heterogeneity and improves predictive accuracy.
2Reliability
If only 20% of patients show significant benefit from PD-1/PD-L1 targeted drugs, then the treatment has high efficacy for responders, but the majority of patients do not benefit, indicating poor predictive capability
Solution Approach 1:
The patent applies local quality assessment by evaluating PD-L1 expression in specific tissue regions rather than uniformly across the entire sample. By identifying and characterizing regions with high PD-L1 expression, the method captures local heterogeneity that is critical for accurate prediction of treatment response.
Solution Approach 2:
The patent performs preliminary spatial mapping and characterization of PD-L1 expression patterns before treatment decision-making. By pre-identifying the spatial distribution and heterogeneity of PD-L1 expression, the method provides more reliable predictive information that can guide treatment selection and avoid unnecessary treatment in non-responders.
Data Source
AI summary
Embodiments include controlling a processor to perform operations, the operations comprising: accessing a digitized image of a region of tissue demonstrating non-small cell lung cancer (NSCLC), detecting a member of a plurality of cellular nuclei represented in the image; classifying the member of the plurality of cellular nuclei as a tumor infiltrating lymphocyte (TIL) nucleus or non-TIL nucleus; extracting spatial TIL features from the plurality of cellular nuclei, including a first subset of features based on the spatial arrangement of TIL nuclei, and a second, different subset of features based on the spatial relationship between TIL nuclei and non-TIL nuclei; generating a set of graph interplay features based on the set of spatial TIL features; providing the set of graph interplay features to a machine learning classifier; receiving, from the machine learning classifier, a probability that the region of tissue will respond to immunotherapy, based, at least in part, on the set of graph interplay features; classifying the region of tissue as likely to respond to immunotherapy or unlikely to respond to immunotherapy based, at least in part, on the probability; and displaying the classification.