Personalized Breast CAD Using Predicted Images for Pathology Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing breast cancer screening technologies, particularly AI-based CAD systems, suffer from high false positive rates due to reliance on tomosynthesis and mammogram images alone, lacking personalized context and long-term risk estimation accuracy, and fail to differentiate between normal aging and pathological changes effectively.
Innovation Solution
A data-driven personalized breast CAD and health-tracking system using a neural network to predict future breast images based on past images and patient-specific data, incorporating a healthy breast latent manifold to identify pathological changes and quantify age-invariant breast health indices.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional CAD systems use machine learning techniques for detection based solely on breast tomosynthesis and mammogram images, then the system can process images quickly, but the false positive rate increases due to lack of personalized context and long-term risk estimation
Solution Approach 1:
The patent combines multiple data sources including breast images (tomosynthesis and mammogram), patient demographic information, lifestyle factors, and familial history into a unified personalized risk assessment system. This integration allows the system to maintain fast image processing while improving detection accuracy by contextualizing findings with personalized patient data, thereby reducing false positives.
Solution Approach 2:
The system performs multiple functions simultaneously: it processes breast images for cancer detection, calculates personalized risk scores based on demographic and lifestyle factors, tracks health progression over time, and provides tailored screening recommendations. This multi-functionality enables the system to maintain productivity while improving reliability through comprehensive personalized assessment.
2Measurement precision
If radiologists review all tomosynthesis images manually, then detection accuracy can be maintained, but the time required for screening increases significantly
Solution Approach 1:
The system segments the breast imaging data into multiple discrete layers and regions through tomosynthesis, allowing both automated analysis of individual layers and targeted radiologist review of suspicious areas. This segmentation enables maintaining high detection accuracy by examining all layers while reducing overall screening time by focusing human expertise only where needed.
Solution Approach 2:
The system provides feedback to radiologists through automated risk scores and highlighted regions of interest, allowing them to prioritize areas requiring manual review. This feedback mechanism maintains detection accuracy by ensuring thorough review of critical areas while significantly reducing the time required to screen all images by eliminating unnecessary manual examination of low-risk regions.
3Measurement precision
If the system incorporates multiple data sources and personalized context information, then long-term risk estimation accuracy improves, but the system complexity increases
Solution Approach 1:
The patent implements a nested structure where multiple data sources and processing layers are organized hierarchically. Patient demographic data, lifestyle factors, and familial history form the foundation layer, with breast imaging analysis nested within this framework. The system processes images through multiple nested algorithms that calculate risk scores at different levels of granularity, improving long-term risk estimation accuracy while managing complexity through structured organization.
Solution Approach 2:
The system performs preliminary processing of multiple data sources to create personalized risk profiles and baseline assessments before actual cancer detection. By pre-processing and organizing diverse data types into structured formats, the system reduces the complexity of integrating these sources during actual analysis while maintaining high risk estimation accuracy through comprehensive personalized context.
Data Source
AI summary
Systems and methods for risk-based breast cancer screening. The breast cancer screening techniques can identify and monitor women who may otherwise later be diagnosed with symptomatic and/or later-stage breast cancer. A personalized breast CAD and health-tracking system is provided that can differentiate pathological changes from normal changes in breast tomosynthesis images. A breast progression predictor can be a generative model that receives input breast images including past images captured at a past timepoint and current images captured at a current timepoint. The model uses the past images to generate predicted images for the current timepoint. Differences between the predicted images and the current images can be used to determine a likelihood of pathological change in the current images. When a pathological change is detected. The system can incorporate a broad spectrum of patient non-image information to further enhance and personalize the prediction of breast progression.