CFOD-TS Biomarker for Breast Cancer Recurrence Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for assessing breast cancer aggressiveness and recurrence risk are expensive, tissue-destructive, and not widely available in clinical settings, lacking validated biomarkers from routine H&E stained images to predict outcomes based on collagen fiber organization.
Innovation Solution
A computational histomorphometric biomarker, Collagen Fiber Orientation Disorder from Tumor-associated Stroma (CFOD-TS), is extracted from routine H&E stained slides using deep learning techniques to classify patients as high-risk or low-risk of disease recurrence by analyzing collagen fiber orientation disorder features without requiring advanced microscopy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If genomic expression based molecular assays are used to determine cancer aggressiveness, then prediction accuracy is improved, but cost and tissue destructiveness increase
Solution Approach 1:
The patent uses digital copies of routine H&E stained tissue sections to extract collagen fiber orientation information through image processing algorithms. Instead of requiring expensive molecular assays or destructive advanced imaging, the system analyzes standardized histology slides that are already part of routine diagnostic workflow, thereby reducing cost and avoiding additional tissue destruction while maintaining prognostic accuracy
Solution Approach 2:
The invention leverages routine H&E stained tissue sections that are already disposed of or archived after standard diagnostic use. By extracting prognostic information from these routinely processed slides before they are discarded, the system provides a cost-effective alternative to expensive molecular assays without requiring additional specialized reagents or advanced imaging techniques
2Measurement precision
If advanced imaging techniques like SHG or laser-scanning multiphoton are used to assess collagen fiber architecture, then measurement precision is improved, but device complexity and availability decrease
Solution Approach 1:
The patent replaces complex advanced imaging techniques with analysis of routine H&E stained slides that are part of standard diagnostic workflow. The image processing system uses computational algorithms to extract collagen fiber orientation information from these readily available, low-cost slides, eliminating the need for expensive SHG or multiphoton imaging equipment while maintaining prognostic capability
Solution Approach 2:
The invention substitutes complex optical imaging systems with digital image processing and machine learning algorithms. Instead of using sophisticated hardware to visualize collagen fibers, the system processes standard H&E stained images through computational methods to extract orientation disorder features, thereby reducing device complexity and increasing clinical availability
3Ease of operation
If routine H&E stained images are used for collagen fiber analysis, then ease of operation and availability are improved, but measurement precision and validation status deteriorate
Solution Approach 1:
The patent transforms routine H&E stained images through specific image processing parameters and computational algorithms to extract quantitative collagen fiber orientation disorder features. By applying machine learning models trained on validated datasets, the system converts standard histology slides into prognostically relevant measurements, thereby achieving both clinical availability and measurement precision
Solution Approach 2:
The invention incorporates validation feedback loops where the image processing system is trained on labeled datasets with known outcomes, continuously refines its accuracy through feedback from clinical results, and adjusts its algorithms to maintain high measurement precision. This feedback mechanism ensures that the system achieves validated biomarker status while remaining based on routine, easily obtainable H&E stained slides
Data Source
AI summary
Embodiments discussed herein relate to accessing a digitized image associated with a patient of tissue demonstrating breast cancer pathology; segmenting a tumor region represented in the digitized image; segmenting collagen fibers represented in the tumor region; computing collagen vectors based on the segmented collagen fibers; generating an orientation co-occurrence matrix based on the collagen vectors; computing a collagen fiber orientation disorder feature based on the co-occurrence matrix; upon determining that the collagen fiber orientation feature exceeds a threshold value: generating a prognosis of the region of tissue as unlikely to experience breast cancer recurrence; upon determining that the collagen fiber orientation feature is less than or equal to the threshold value: generating a prognosis of the region of tissue as likely to experience breast cancer recurrence; classifying the patient as high-risk of recurrence or low-risk of recurrence based, at least in part, on the prognosis; and displaying the classification.