OCT Signal Processing With Speckle Statistics for Density Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing optical coherence tomography (OCT) systems struggle to accurately estimate scatterer density and other characteristics of a sample due to variations in spatial resolution and interference from factors like aberration and signal-to-noise ratio, making precise diagnosis and evaluation challenging.
Innovation Solution
A measurement signal processing device and method using a model learning unit to determine model parameters and generate training data sets, incorporating a neural network with convolutional layers to estimate characteristics like scatterer density by minimizing differences between estimated and target values, accounting for factors like spatial resolution and aberration.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If scatterer density is estimated from OCT signal intensity, then estimation can be performed, but accuracy is insufficient due to various affecting factors
Solution Approach 1:
The patent introduces a speckle pattern as an intermediary element that connects the OCT signal to the scatterer density. By analyzing the speckle pattern's statistical properties (variance, skewness, kurtosis) rather than direct signal intensity, the system mediates between the complex OCT signal and the target parameter (scatterer density), achieving more accurate estimation while accounting for affecting factors like aberration and signal-to-noise ratio
Solution Approach 2:
The patent transforms the estimation approach by changing from direct intensity-based parameters to speckle pattern statistical parameters (variance, skewness, kurtosis). This parameter transformation allows the system to extract scatterer density information that is more robust against variations in signal-to-noise ratio, aberration, and other affecting factors, thereby improving measurement precision
2Measurement precision
If spatial resolution is not known in advance, then measurement can proceed, but scatterer density estimation becomes unreliable
Solution Approach 1:
The patent implements a feedback mechanism where the speckle pattern analysis provides information about the system's spatial resolution and aberration characteristics. By comparing the observed speckle pattern statistics with theoretical models, the system can infer and compensate for spatial resolution variations and aberration effects, thereby improving the reliability of scatterer density estimation without requiring prior knowledge of these parameters
3Measurement precision
If speckle pattern analysis is performed to estimate resolution, then scatterer density can be estimated, but accuracy is insufficient due to signal-to-noise ratio and aberration effects
Solution Approach 1:
The patent converts the harmful effects of speckle noise and aberration into beneficial information. By analyzing the statistical properties of the speckle pattern (variance, skewness, kurtosis) rather than treating it as noise to be eliminated, the system extracts meaningful information about scatterer density while compensating for aberration effects. The speckle pattern, previously considered a nuisance, becomes the key to accurate measurement
Solution Approach 2:
The patent employs a composite approach by combining multiple speckle pattern statistical parameters (variance, skewness, kurtosis) to form a comprehensive estimation model. This composite parameter approach allows the system to capture different aspects of the speckle pattern and compensate for various affecting factors simultaneously, achieving more accurate and robust scatterer density estimation
Data Source
AI summary
A model learning unit determines a model parameter for calculating an estimated value for each of a plurality of training sets, each including a measurement signal and at least one type of predetermined characteristic value indicating characteristics of the measurement signal as a target value, to minimize a difference between the estimated value calculated for the measurement signal using a predetermined mathematical model and the target value is minimized, determines characteristic value sets, each being a set of a plurality of types of characteristic values indicating characteristics of the measurement signal including the target value, with which, for each target value, the target value is common to a plurality of characteristic value sets, and generates, for each of the characteristic value sets, each of the plurality of training sets including the target value and a measurement signal having characteristics indicated by the plurality of types of characteristic values.