Neural Network Reconstruction for Near-Infrared Tomography
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Near-infrared spectroscopy tomography faces challenges in reconstructing images quickly and accurately due to limited boundary measurement data and mixed noise in measured signals, leading to long reconstruction times and poor artifact suppression in traditional methods.
Innovation Solution
A neural network-based reconstruction method is employed, utilizing the Boltzmann radiation transmission equation and a BP neural network to process photon fluence rates and absorption coefficients, with a training process involving forward and back propagation to optimize weight adjustments and improve image reconstruction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional regularization methods are used for image reconstruction, then the reconstruction process can be completed, but the reconstruction time is relatively long and artifact suppression ability is weak
Solution Approach 1:
The patent replaces the traditional mechanical optimization-based reconstruction system with a neural network-based system. The neural network is trained offline using optimization algorithms, but during actual reconstruction, it performs rapid forward propagation to directly output reconstructed images, substituting the iterative mechanical optimization process with a streamlined neural computation process that achieves both speed and accuracy.
Solution Approach 2:
The patent applies preliminary action by pre-training the neural network offline using comprehensive training data and optimization algorithms. This preliminary training phase prepares the network weights and biases in advance, so that during actual reconstruction tasks, the network can rapidly produce accurate results without requiring time-consuming iterative optimization during the reconstruction process itself.
2Reliability
If traditional optimization methods are used to transform the reconstruction problem into a nonlinear optimization problem, then the reconstruction can be performed, but the artifact suppression ability remains weak
Solution Approach 1:
The patent substitutes the traditional mechanical optimization system with a neural network system that has been trained to recognize and suppress artifacts. The neural network learns the complex patterns of artifacts and reconstruction solutions during training, enabling it to automatically suppress artifacts during inference while maintaining high reconstruction efficiency without requiring iterative optimization.
3Measurement precision
If limited boundary measurement data with mixed noise is used, then the measurement process is practical, but the reconstruction accuracy deteriorates
Solution Approach 1:
The patent applies feedback by training the neural network with大量的 labeled training data that includes various noise conditions and ground truth images. During training, the network receives feedback through loss functions that compare its predictions with actual ground truth, allowing it to learn robust reconstruction patterns that can accurately reconstruct images even from noisy, limited measurement data. This feedback-driven training enables the network to compensate for information loss in the measurements.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach significantly reduces reconstruction time and enhances the accuracy of optical parameter distribution maps, achieving efficient and precise image reconstruction.
Implementation Method 1
A neural network-based reconstruction method is employed, utilizing the Boltzmann radiation transmission equation and a BP neural network to process photon fluence rates and absorption coefficients
Implementation Method 2
In the Boltzmann radiation transmission equation, transmission process of light is regarded as absorption and scattering process of photons in medium
Implementation Method 3
In the Boltzmann radiation transmission equation, transmission process of light is regarded as absorption and scattering process of photons in medium
Data Source
AI summary
The present disclosure disclose a near-infrared spectroscopy tomography reconstruction method based on neural network which belongs to the field of medical image processing. In the Boltzmann radiation transmission equation, transmission process of light is regarded as absorption and scattering process of photons in medium, and interaction between light and tissue is determined by absorption coefficient, scattering coefficient and phase function of the response scattering distribution. In the transmission, only the particle property of light is taken into account, not the fluctuation of light. Therefore, polarization and interference phenomena related to the fluctuation of light are not considered, and only the energy transmission of light is tracked. The reconstruction method based on BP neural network is used to reconstruct the distribution of optical absorption coefficient, reconstruction results of absorption coefficient distribution can be obtained by calculation. This method can not only reconstruct the absorption coefficient distribution accurately, but also has high computational efficiency.