X-Ray CT Component Estimation Using Deep Metric Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods struggle to accurately estimate component amounts inside a subject due to discrepancies between actual phenomena and simulation results, which cannot be mathematically expressed, leading to incorrect diagnoses across various fields.
Innovation Solution
An X-ray CT apparatus and information processing method utilize deep metric learning to embed real and pseudo spectral reflectances as feature vectors in different feature spaces, allowing for accurate estimation of internal component amounts by matching these vectors, thereby reducing the influence of discrepancies between actual phenomena and simulation results.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If simulation results are used to estimate component amounts, then measurement can be performed, but accuracy deteriorates due to discrepancies between actual phenomenon and simulation
Solution Approach 1:
The patent segments the feature space into multiple subspaces using deep metric learning, where each subspace captures different characteristics of the data. This segmentation allows the system to handle the complexity of the relationship between simulation results and actual phenomena by breaking it down into manageable parts, thereby improving estimation accuracy despite simulation discrepancies.
Solution Approach 2:
The patent transforms the original feature space into a new dimensional space through deep metric learning embedding. By mapping features into a higher-dimensional space with different metric properties, the system can better capture the underlying relationships between simulation results and actual component amounts, resolving the accuracy-reliability contradiction.
2Measurement precision
If deep metric learning is used to embed feature vectors, then estimation accuracy improves, but computational complexity increases
Solution Approach 1:
The patent performs preliminary embedding of feature vectors into the target metric space during an offline training phase. This preliminary action allows the complex deep metric learning computations to be done in advance, so that during actual component amount estimation, the system only needs to perform simpler distance calculations in the pre-established feature space, reducing real-time computational complexity.
Data Source
AI summary
An X-ray CT apparatus of an embodiment includes processing circuitry. The processing circuitry configured to acquire a number of photons by performing a CT scan on a target event, generate a first feature amount from the number of photons, match the first feature amount to second feature amounts generated using a machine learning model from the number of photons of a reference event, the number of photons of the reference event being a number of photons generated from known component amounts of the reference event and having a nonlinear relationship with the known component amounts, and estimate the known component amounts corresponding to the second feature amount matched to the first feature amount as component amounts of the target event.