AI Radiomic Feature Comparison for Prostate MRI
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The Prostate Imaging—Reporting and Data System (PI-RADS) has limitations in differentiating clinically significant prostate cancer (sPC) from insignificant prostate cancer (inPC), leading to high false positive rates and interobserver variability in radiological data interpretation, particularly due to overlapping image features from benign conditions.
Innovation Solution
A method using AI-based evaluation of radiological data to identify pre-stored medical datasets similar to a current case dataset, reducing interobserver variability by quantifying definitive features and comparing them using a predefined AI-based method, which can include machine learning or deep learning techniques, to output the most similar datasets for enhanced assessment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If PI-RADS scoring system is used for prostate cancer detection, then standardization of prostate MRI examinations is achieved, but interobserver variability remains high due to heterogeneous signal changes from benign conditions
Solution Approach 1:
The patent replaces human observer interpretation with an AI-based system that automatically extracts radiomic features and compares them against pre-stored datasets. The AI system processes MRI data through deep learning models that quantify definitive features, eliminating the mechanical subjectivity of human observers and providing consistent, reproducible results across different cases and observers.
Solution Approach 2:
The patent introduces an intermediary AI-based feature extraction and comparison system between the MRI data and the diagnostic decision. This intermediary layer automatically extracts radiomic features, compares them against pre-stored datasets with known outcomes, and provides standardized assessments, thereby mediating between the raw imaging data and the diagnostic interpretation to reduce interobserver variability.
2Reliability
If PI-RADS scoring system is used for prostate cancer detection, then detection of clinically significant cancer is improved, but false positive rate increases due to overlapping image features from benign conditions
Solution Approach 1:
The patent segments the complex MRI data into multiple radiomic features (texture, shape, intensity, etc.) and evaluates each feature independently through deep learning models. This segmentation allows the system to distinguish between subtle patterns indicative of cancer versus benign conditions by analyzing multiple feature dimensions simultaneously, thereby reducing false positives while maintaining detection sensitivity.
Solution Approach 2:
The patent transforms the PI-RADS scoring system from a single-score assessment to a multi-parameter radiomic feature analysis. By changing from qualitative scoring to quantitative measurement of multiple radiomic parameters (texture heterogeneity, shape irregularity, intensity patterns), the system achieves more precise differentiation between clinically significant cancer and benign conditions, reducing false positive rates.
3Reliability
If AI-based evaluation method is implemented to extract definitive features, then interobserver variability is reduced through standardized comparison, but device complexity increases
Solution Approach 1:
The patent performs preliminary action by pre-storing and pre-evaluating numerous medical datasets with known outcomes before actual diagnosis. The AI system is pre-trained on these datasets to recognize patterns and extract radiomic features automatically. This preliminary preparation allows the system to handle new cases efficiently without requiring complex real-time processing, thereby reducing operational complexity while maintaining high reliability.
Data Source
AI summary
Similar pre-stored medical datasets are identified by comparison with a current case dataset. A current case dataset is provided and includes radiological data of a patient. A number of pre-stored medical datasets each including radiological data of other patients are provided. Each case dataset is evaluated according to a predefined AI-based method to obtain a number of definitive features for that case dataset. The definitive features of the current case dataset are compared with the definitive features of each pre-stored medical dataset to identify a number of pre-stored medical datasets most similar to the current case dataset. The identified number of most similar pre-stored medical datasets are output.