Radiographic Image Classification by Acquisition Characteristic Clustering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for controlling the impact of image acquisition variability due to diverse protocols lack prior knowledge of these protocols, leading to uncertainty in clinical decisions and unreliable radiological evaluations.
Innovation Solution
A machine learning model is developed to automatically classify radiographic images by defining reference clusters based on image acquisition characteristics, using radiomics measurements and segmenting non-pathological ROIs or phantom-derived structures to group images with similar acquisition protocols, without requiring prior knowledge of these protocols.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If data harmonization is used to circumvent measurement variability, then measurement reliability is improved, but prior knowledge of acquisition protocols is required which is often difficult to obtain
Solution Approach 1:
A machine learning model acts as an intermediary that automatically infers acquisition protocol characteristics from the radiographic images themselves. The model takes images as input and outputs predicted acquisition characteristics, eliminating the need for direct access to protocol information while still enabling measurement harmonization. This intermediary approach bridges the gap between image data and protocol knowledge.
Solution Approach 2:
The patent replaces the mechanical/manual process of obtaining and inputting acquisition protocol information with an automated machine learning system. Instead of manually extracting or inputting protocol parameters, the ML model automatically predicts acquisition characteristics directly from image data, substituting the traditional information gathering mechanism with an intelligent automated system.
2Adaptability or versatility
If machine learning models are trained on diverse acquisition protocols, then model versatility is improved, but measurement variability increases
Solution Approach 1:
The patent segments the diverse acquisition protocols into distinct clusters based on their characteristics. Instead of treating all protocols as a single heterogeneous group, the ML model identifies and separates them into homogeneous clusters, allowing measurements to be compared within clusters where variability is minimized while maintaining overall model versatility across multiple clusters.
Solution Approach 2:
The patent changes the parameter space by transforming raw radiomics measurements into normalized or harmonized values that account for acquisition protocol differences. By applying parameter transformations based on predicted acquisition characteristics, the model maintains measurement precision across diverse protocols while preserving model versatility.
Data Source
Figure 1
Figure 2(a)~3a
Figure 3b~3c
AI summary
It is provided a method and system for generating a machine learning model for automatic classification of automatic classification of radiographic images acquired by various acquisition protocols. The method comprises the steps of: providing a plurality of radiographic images, detecting and segmenting in each of the radiographic image at least one regions of interest (ROI) as reference ROI, measuring at least one radiomic feature per reference ROI, identifying valid reference ROIs based on the measured radiomics values, and clustering the measured radiomics values of valid reference ROIs into at least two reference clusters according to a set of characteristics of image acquisition. It is further provided a method and system for classifying radiographic images by applying a machine learning model generated for automatic classification of radiographic images.