Breast Cancer Identification Through Confidence-Gated Image Validation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing breast cancer screening methods require significant workload from medical personnel and may lead to overdiagnosis due to the limitations of current AI algorithms.
Innovation Solution
A system and method that utilizes an evaluation unit to analyze breast images using AI, generating confident or unconfident results, requesting external validation for unconfident cases, and blocking access to initial results until external validation is complete, incorporating metadata analysis and image comparison for personalized reports.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If AI algorithms are used to interpret mammograms for early cancer detection, then cancer detection capability is improved, but workload for medical personnel and risk of overdiagnosis increases
Solution Approach 1:
The AI evaluation unit automatically analyzes mammogram images and generates diagnosis results with confidence levels, performing the initial screening work autonomously without requiring medical personnel to manually evaluate each image, thus reducing their workload while maintaining detection capability
Solution Approach 2:
The system introduces an intermediary confidence level mechanism that filters cases before human review. Only unconfident cases (those below the confidence threshold) are forwarded to medical personnel for additional evaluation, acting as a mediator to reduce the volume of work requiring human intervention
2Productivity
If AI algorithms interpret all mammograms, then screening efficiency is improved, but overdiagnosis occurs due to algorithm limitations
Solution Approach 1:
The system implements a feedback mechanism where unconfident cases are identified and forwarded for external evaluation. The results from external evaluation of unconfident cases can be used to retrain and improve the AI algorithm, creating a continuous improvement loop that reduces overdiagnosis while maintaining efficiency
Solution Approach 2:
The system changes the operational parameter by introducing a confidence level threshold. Instead of the AI making definitive diagnoses on all cases, it only provides confident diagnoses above the threshold, transforming uncertain cases into a separate category for external review, thereby reducing overdiagnosis while preserving screening efficiency
3Reliability
If external evaluation is requested for unconfident cases, then diagnostic reliability is improved, but processing time increases
Solution Approach 1:
The system segments the diagnostic process into two distinct paths: confident cases that are processed automatically and immediately, and unconfident cases that require external evaluation. This segmentation ensures that the majority of confident cases are resolved quickly without external review, minimizing overall processing time while maintaining reliability for uncertain cases
Data Source
AI summary
An exemplary embodiment of the present invention relates to a system for identifying breast cancer based on at least one image of a patient's breast, the system comprising: an evaluation unit configured to analyze the image, wherein the evaluation unit generates a diagnosis result that labels the image as confident if the confidence in the evaluation result exceeds a given confidence level, and otherwise generates a diagnosis result that labels the image as unconfident, a request unit configured to generate an evaluation-request signal that requests an additional external evaluation of the image, if the diagnosis result labels the image as unconfident, and a transfer unit having an input port for receiving the result of the additional external evaluation, and an output port for outputting the diagnosis result of the evaluation unit, wherein the transfer unit is configured to block access to the diagnosis result of the evaluation unit until the result of the additional external evaluation has been received.