Autoencoder Determination Control Device for Medical Image Anomaly Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for normal/abnormal data determination using probability distributions face challenges when input data features exhibit various probability distributions, leading to inaccurate differentiation due to entropic variations from formations and backgrounds in medical images.
Innovation Solution
A determination control device employing an autoencoder with encryption, noise addition, and decryption processes, along with a Gaussian mixture model for probability distribution estimation and parameter adjustment, to establish a determination standard based on low-dimensional feature entropy and membership coefficients.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If probability distribution is estimated from low-dimensional feature values extracted by autoencoder, then data dimensionality is reduced and computation is simplified, but determination accuracy deteriorates when input data features exhibit various probability distributions
Solution Approach 1:
The patent segments the probability distribution estimation process into multiple components: it separates the autoencoder's encryption function from the probability distribution modeling, and introduces a Gaussian mixture model to handle different distribution types. This segmentation allows the system to maintain low-dimensional features while improving accuracy by addressing the complexity of various probability distributions through specialized modeling components.
Solution Approach 2:
The patent changes parameters by introducing a threshold value for determining whether to use Gaussian mixture model or simple probability distribution, and by adjusting the number of components in the Gaussian mixture model. This parameter adaptation allows the system to switch between different modeling approaches based on the complexity of the data, maintaining both computational efficiency and determination accuracy.
2Use of energy by moving object
If simple probability distribution is used for determination, then computational cost is reduced, but determination accuracy deteriorates when entropic variations occur from formations and backgrounds
Solution Approach 1:
The patent implements a dynamic modeling approach where the system adapts its probability distribution model based on the data characteristics. It dynamically switches between simple probability distribution and Gaussian mixture model depending on whether the data exhibits complex entropic variations, allowing the system to maintain computational efficiency while improving accuracy when needed.
Solution Approach 2:
The patent introduces an intermediary Gaussian mixture model that mediates between the simple probability distribution and the complex data characteristics. This intermediary component allows the system to handle entropic variations from formations and backgrounds without requiring overly complex computations, thus balancing computational cost and determination accuracy.
3Quantity of substance
If autoencoder encryption function is used to extract low-dimensional features, then data compression is achieved, but determination accuracy deteriorates when probability distributions vary across different data types
Solution Approach 1:
The patent segments the feature extraction and probability distribution modeling into separate functional components. The autoencoder's encryption function is dedicated to dimensionality reduction, while the Gaussian mixture model is dedicated to handling probability distribution variations. This segmentation allows each component to optimize its specific function, maintaining low-dimensional representation while improving determination accuracy.
Solution Approach 2:
The patent changes parameters by introducing adaptability in the probability distribution modeling. It adjusts the modeling approach based on the data characteristics, using Gaussian mixture model when probability distributions vary across different data types. This parameter adaptation allows the system to maintain compressed low-dimensional features while compensating for distribution variations to improve determination accuracy.
Data Source
AI summary
A determination control device includes a processor that executes a procedure. The procedure includes estimating, as a probability distribution, a low-dimensional feature value obtained by encrypting input data, the low-dimensional feature value having a lower dimensionality than the input data, generating output data by decrypting a feature value resulting from adding noise to the low-dimensional feature value, and adjusting respective parameters of the encrypting, the estimating, and the decrypting, based on a cost including an error between the input data and the output data and including an entropy of the probability distribution, wherein, in a determination as to whether or not target input data is normal, a determination standard for the determination is controlled based on information obtained from another probability distribution estimated by encrypting the target input data with the parameters after adjusting.