In a three-dimensional ultrasound computed tomography (3D USCT) system, system errors such as transducer delay, transducer position deviation and temperature error will affect the quality of reconstructed images. Most of the existing calibration works use iterative methods to solve large-scale systems of linear equations. In our case, the transducer delay and position deviation calibration problem of the considered 3D USCT system is essentially to solve a linear system containing about 840,000 equations and 11,500 unknowns. For such a large system, the existing iterative methods require a lot of computation time and the accuracy also needs to be improved. Considering that neural networks have the ability to find optimized solutions for large-scale linear systems, we propose a neural network method for transducer delay and position deviation calibration. We designed a neural network to calibrate both delay and position solutions, together during the network training. We test the method with simulated system data where we add transducer delays in the range of 0.7~1.3 μs, position deviation in the range of -1~1 mm for the X- and Y-axis, and -0.3~0.3 mm for the Z-axis. Results show that the mean delay error is reduced to 0.15 μs, and the mean position error is reduced to 0.15 mm, after a neural network calibration process which takes about 11 minutes. The delay calibration result is better than the existing Newton method in literature, while our method is especially less time-consuming.
Breast cancer is the most common cancer for women worldwide. 3D Ultrasound Computed Tomography (3D USCT) is a novel imaging method for early breast cancer diagnosis, which allows reconstruction of quantitative tissue parameters like speed of sound and attenuation. For reconstruction we use the paraxial approximation of the Helmholtz equation as forward model. We have realized the forward solution, backprojection and reconstruction for a ring transducer arrangement. The reconstruction software was evaluated with data simulated with k-Wave, resulting in the mean error for the speed of sound map of 12.6 m/s for a pixel size of 0.3 mm. Spatial resolution was estimated with a resolution phantom containing circular inclusions with realistic speed of sound values for breast tissues, allowing maximum resolution of 2 mm. In this paper we show that our method has accurate forward solution, we present the new backprojection technique and initial results of reconstructing simulated data.
Ultrasound transmission tomography promises a high potential and novel imaging method for early breast cancer diagnosis; it can quantitatively characterize tissues or materials by the attenuation and speed of sound (SoS). Reconstruction of ultrasound transmission tomography is an inverse problem that can be solved iteratively based on a paraxial approximation of the Helmholtz equation as forward model, which is highly non-linear and time-consuming. In order to address these problems and reconstruct desired images, we design a dual domain network architecture for ultrasound transmission tomography reconstruction. It can enhance the information of measurement domain and directly reconstruct from pressure field measurements without using any initialization of reconstruction and fully connected layer. We train the network on simulated ImageNet data and transfer it for ultrasound transmission tomography images to avoid overfitting when the amount of ultrasound transmission tomography images is limited. Our experimental results demonstrate that a dual domain network produces significant improvements over state-of-the-art methods. It improves the measured structural similarity measure (SSIM) from 0.54 to 0.90 and normalized root mean squared error (nRMSE) from 0.49 to 0.01 on average concerning the SoS reconstruction, and from 0.46 to 0.98 for SSIM, from 353 to 0.03 for nRMSE on average concerning the attenuation reconstruction.
In ultrasound transmission tomography, image reconstruction is an inverse problem which is solved iteratively based on a forward model that simulates the wave propagation of ultrasound. A commonly used forward model is paraxial approximation of the Helmholtz equation, which is time-consuming. Hence developing optimizers that minimize the number of forward solutions is crucial to achieve clinically acceptable reconstruction time, while the state-of-the-art methods in this field such as Gauss-Newton conjugate gradient (CG) and nonlinear CG are not capable of reaching this goal. To that end, we focus on Jacobian-free optimizers or accelerators in this paper, since the computation of the Jacobian is expensive. We investigate the limited memory Broyden–Fletcher–Goldfarb–Shanno (L-BFGS) algorithm as a preconditioning technique due to its ability to efficiently approximate inverse Hessian without performing forward model or its adjoint. We show L-BFGS can reach a speedup of more than one order of magnitude for the noise-free case, while the method still halves the reconstruction time in presence of noise in the data. The performance drop is explained by perturbed gradients due to noise in the data. We also show when used alone as a quasi-Newton method, L-BFGS is competitive with the accelerated CG based methods regarding the number of iterations, and outperforms them regarding reconstruction time.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.