Autoencoder Training for Edema Diagnosis Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for diagnosing edema, which rely on the discrepancy between actual phenomena and simulation results, often lead to erroneous recognition due to ignoring discrepancies between observed and simulated data, affecting accuracy in medical and other fields.
Innovation Solution
An information processing device that acquires spectral reflectance data from edema images, converts it into feature quantities using a machine learning model, and trains the model based on the discrepancy between observed and restored spectral reflectance, ensuring accurate diagnosis even when discrepancies exist.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If an autoencoder is trained to restore observation values, then the restoration capability is improved, but the estimation accuracy of in-vivo components deteriorates due to discrepancy between simulation and actual phenomenon
Solution Approach 1:
The patent introduces a feedback mechanism where the discrepancy between simulated and actual observation values is fed back into the training process. The loss function incorporates both the restoration error and the discrepancy term, allowing the model to learn from and compensate for simulation-actual mismatches, thereby improving estimation accuracy while maintaining reliability.
Solution Approach 2:
The patent changes the training parameters by modifying the loss function to include a discrepancy term that penalizes differences between simulated and actual observation values. This parameter change ensures that the autoencoder learns to account for simulation-actual discrepancies, improving both restoration capability and estimation accuracy simultaneously.
2Ease of manufacture
If training is performed ignoring the discrepancy between observation and simulation, then the training process is simplified, but the diagnosis accuracy deteriorates due to erroneous recognition
Solution Approach 1:
The patent implements a feedback mechanism where the discrepancy between observed and simulated data is continuously monitored and fed back into the training process. This ensures that the model learns to account for discrepancies, improving diagnosis accuracy while maintaining a relatively simple training framework based on standard autoencoder architecture.
3Device complexity
If a loss function including only image feature quantity errors is used, then the training is simpler, but the model fails to account for discrepancy between actual and simulated data
Solution Approach 1:
The patent modifies the loss function parameters by adding a discrepancy term that specifically measures the difference between simulated and actual observation values. This parameter change enables the model to account for simulation-actual discrepancies without significantly increasing training complexity, as the additional term integrates seamlessly into the standard backpropagation framework.
Data Source
AI summary
An information processing device includes processing circuitry. The processing circuitry acquires observation data which is acquired when a target event is observed. The processing circuitry converts the observation data to a feature quantity of the target event using a machine learning model. The processing circuitry restores the observation data from the feature quantity using a numerical simulation model. The processing circuitry trains the machine learning model on the basis of a discrepancy between first observation data which is the observation data that has not been converted to the feature quantity and second observation data which is the observation data restored from the feature quantity.