Multimodal Diagnostic Data Fusion to Reduce False Positives
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing medical diagnostic methods, such as mammograms, have high false-positive rates and require invasive biopsies to confirm benign or malignant lesions, leading to unnecessary procedures and inefficiencies.
Innovation Solution
A method and apparatus for multimodal data fusion using machine learning to integrate medical imaging data with biomarker and electronic medical record data, employing early, intermediate, and late data fusion techniques to generate more accurate classification results.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If medical imaging alone is used for diagnosis, then the diagnostic process is simple and quick, but the false-positive rate is high and diagnostic accuracy is insufficient
Solution Approach 1:
The patent combines multiple data modalities (medical imaging data, biomarker data, and electronic medical record data) into a unified diagnostic system. This merging of diverse data sources enables more accurate diagnosis by integrating complementary information, directly resolving the contradiction between diagnostic accuracy and system complexity through comprehensive data fusion rather than relying on a single modality.
Solution Approach 2:
The diagnostic system functions as a composite information system that integrates heterogeneous data types (imaging, biomarkers, EHRs) similar to how composite materials combine different substances to achieve superior properties. This composite approach allows the system to leverage the strengths of each data modality while mitigating their individual limitations, achieving high diagnostic accuracy without excessive complexity.
2Reliability
If multiple data modalities are integrated, then diagnostic accuracy improves and false-positives reduce, but data processing complexity increases
Solution Approach 1:
The patent segments the diagnostic process into distinct stages: data acquisition from multiple modalities, intermediate fusion of imaging and biomarker data, and final fusion with EHR data. This segmentation allows each stage to process specific data types with appropriate methods, reducing overall processing complexity while maintaining diagnostic reliability through systematic multi-stage analysis.
Solution Approach 2:
The patent introduces intermediate fusion as a mediator stage between raw data acquisition and final diagnosis. This intermediate layer processes and integrates imaging and biomarker data before combining with EHR information, acting as a buffer that simplifies the complexity of direct multi-modal integration while preserving diagnostic reliability through structured intermediate processing.
3Measurement precision
If invasive biopsies are performed to confirm lesions, then diagnostic certainty is achieved, but unnecessary procedures and patient burden increase
Solution Approach 1:
The patent performs preliminary diagnostic assessment by integrating multiple data modalities (imaging, biomarkers, EHRs) before recommending invasive biopsies. This preliminary action uses less invasive methods to gather comprehensive information and achieve high diagnostic certainty, thereby reducing the need for unnecessary invasive procedures and minimizing patient burden while maintaining diagnostic precision.
Solution Approach 2:
The system uses feedback from multiple data sources (imaging features, biomarker levels, EHR history) to continuously refine diagnostic confidence. This feedback mechanism allows the system to achieve high diagnostic certainty through iterative analysis of accumulated evidence, reducing reliance on invasive biopsies as the default next step and thereby decreasing unnecessary patient procedures.
Data Source
AI summary
According to some embodiments, a method comprises obtaining data in multiple modalities, including data in an image modality and data in a biomarker modality; obtaining one or more trained machine-learning models; generating a first multi-modality group from the data in multiple modalities, wherein the first multi-modality group includes data in at least the image modality and in the biomarker modality; generating a group of intermediate data, wherein generating the group of intermediate data includes inputting the first multi-modality group into a machine-learning model that has been trained to extract features from input multi-modality data and output the features as the intermediate data; and generating a first classification result based on at least the group of intermediate data, wherein generating the first classification result includes inputting the intermediate data into a machine-learning model that has been trained to output the classification result based on the group of intermediate data.