Ultrasound-target-shape-guided sparse regularization to improve accuracy of diffused optical tomography
A diffuse optical tomography (DOT) system for generating a functional image of a lesion region of a subject includes a source subsystem configured to generate optical waves, a probe coupled to the source subsystem and configured to emit the optical waves generated by the source subsystem toward the lesion region and to detect optical waves reflected by the lesion region, a detection subsystem configured to convert the optical waves detected by the probe to digital signals, and a computing device including a processor and a memory. The memory includes instructions that program the processor to receive the digital signals sent from the detection subsystem, calculate an initial estimate of the functional image by solving an inverse optimization problem with a shape-regularized Fast Iterative Shrinkage-Thresholding algorithm (FISTA), reconstruct the functional image iteratively using a Finite Difference Method (FDM) or a Finite Element Method (FEM), and display the functional image.
1. A diffuse optical tomography (DOT) system for generating a functional image of a lesion region of a subject, comprising:
a source subsystem configured to generate optical waves;
a probe coupled to the source subsystem and configured to emit the optical waves generated by the source subsystem toward the lesion region and to detect optical waves reflected by the lesion region;
a detection subsystem configured to convert the optical waves detected by the probe to digital signals; and
a computing device including a processor and a memory, the memory including instructions that program the processor to:
receive the digital signals sent from the detection subsystem;
receive lesion depth and shape information from a plurality of co-registered ultrasound images of the lesion region;
calculate an initial estimate of the functional image based on the digital signals and the lesion depth and shape information by solving an inverse optimization problem with a target-shape-regularized Fast Iterative Shrinkage-Thresholding algorithm (FISTA), wherein the calculation of the initial estimate is based upon depth dependent 1 regularization;
generate the functional image of the lesion region of the subject by performing a depth dependent 1 regularized non-linear Born iterative reconstruction method that comprises a non-linear Born iterative reconstruction method regularized by the depth dependent 1 regularization associated with the initial estimate, the functional image being generated by the processor being further programmed by the instructions to:
(1) generate, as part of the functional image, a lesion absorption map based on the initial estimate by solving the inverse optimization problem with the target-shape-regularized FISTA;
(2) update, using a Finite Difference Method (FDM), a photon-density wave distribution associated with the initial estimate to obtain an updated estimation of a weight matrix for use with the depth dependent 1 regularized non-linear Born iterative reconstruction method;
(3) generate an updated weight matrix based at least on the updated estimation of the weight matrix; and
(4) perform at least two iterations of the depth dependent 1 regularized non-linear Born iterative reconstruction method, at least one of the at least two iterations using the updated weight matrix; and
display the generated functional image.
2. The DOT system of claim 1 , wherein the instructions program the processor to calculate an initial estimate of the functional image using a dual-grid schedule to discretize the legion region with a fine grid and to discretize the background volume of the subject with a coarse grid to generate a weighted matrix for the lesion region and a weighted matrix for the background volume based on absorption of the optical waves on the subject.
3. The DOT system of claim 2 , wherein the instructions program the processor to perform reconstruction via the depth dependent 1 regularized non-linear Born iterative reconstruction method for generating the functional image iteratively by updating the photon-density wave distribution with a conjugate gradient method inside the fine grid.
4. The DOT system of claim 1 , further comprising an ultrasound imaging device coupled to the computing device and configured to generate the plurality of co-registered ultrasound images of the lesion region.
5. The DOT system of claim 1 , wherein the source subsystem is configured to generate optical waves at four optical wavelengths in a range from 730 nm to 830 nm.
6. The DOT system of claim 1 , wherein the depth dependent 1 regularized non-linear Born iterative reconstruction method is configured to reconstruct at least one of values or distributions of absorption coefficients associated with absorption of the optical waves on the subject.
7. A method for generating a functional image of a lesion region of a subject using diffuse optical tomography (DOT), the method comprising:
emitting optical waves toward the lesion region;
detecting optical waves reflected by the lesion region and converting the optical waves to digital signals;
receiving lesion depth and shape information from a plurality of co-registered ultrasound images of the lesion region;
calculating an initial estimate of the functional image based on the digital signals and the lesion depth and shape information by solving an inverse optimization problem with a shape-regularized Fast Iterative Shrinkage-Thresholding algorithm (FISTA), wherein the calculation of the initial estimate is based upon depth dependent 1 regularization;
generating the functional image of the lesion region the subject by performing a depth dependent 1 regularized non-linear Born iterative reconstruction method that comprises a non-linear Born iterative reconstruction method regularized by the depth dependent 1 regularization associated with the initial estimate, by:
(1) generating, as part of the functional image, a lesion absorption map based on the initial estimate by solving the inverse optimization problem with the target-shape-regularized FISTA;
(2) updating, using a Finite Difference Method (FDM), a photon-density wave distribution associated with the initial estimate to obtain an updated estimation of a weight matrix for use with the depth dependent 1 regularized non-linear Born iterative reconstruction method;
(3) generating an updated weight matrix based at least on the updated estimation of the weight matrix; and
(4) performing at least two iterations of the depth dependent 1 regularized non-linear Born iterative reconstruction method, at least one of the at least two iterations using the updated weight matrix; and
displaying the generated functional image.
8. The method of claim 7 , further comprising calculating an initial estimate of the functional image using a dual-grid schedule to discretize the legion region with a fine grid and to discretize the background volume of the subject with a coarse grid to generate a weighted matrix for the lesion region and a weighted matrix for the background volume based on absorption of the optical waves on the subject.
9. The method of claim 8 , further comprising performing reconstructing via the depth dependent 1 regularized non-linear Born iterative reconstruction method for generating the functional image iteratively by updating the photon-density wave distribution with a conjugate gradient method inside the fine grid.
10. The method of claim 7 , further comprising generating optical waves at four optical wavelengths in a range from 730 nm to 830 nm.
11. The method of claim 7 , wherein the depth dependent 1 regularized non-linear Born iterative reconstruction method is configured to reconstruct at least one of values or distributions of absorption coefficients associated with absorption of the optical waves on the subject.
12. At least one non-transitory computer-readable storage media having computer-executable instructions embodied thereon for generating a functional image of a lesion region of a subject using diffuse optical tomography (DOT), wherein when executed by at least one processor, the computer-executable instructions cause the processor to:
emit optical waves toward a lesion region of the subject;
detect optical waves reflected by the lesion region and converting the optical waves to digital signals;
receive lesion depth and shape information from a plurality of co-registered ultrasound images of the lesion region;
calculate an initial estimate of the functional image based on the digital signals and the lesion depth and shape information by solving an inverse optimization problem with a shape-regularized Fast Iterative Shrinkage-Thresholding algorithm (FISTA), wherein the calculation of the initial estimate is based upon depth dependent 1 regularization;
generate the functional image of the lesion region of the subject by performing a depth dependent 1 regularized non-linear Born iterative reconstruction method that comprises non-linear Born iterative reconstruction method regularized by the depth dependent 1 regularization associated with the initial estimate, the functional image being generated by the processor being further caused by the computer-executable instructions to:
(1) generate, as part of the functional image, a lesion absorption map based on the initial estimate by solving the inverse optimization problem with the target-shape-regularized FISTA;
(2) update, using a Finite Difference Method (FDM), a photon-density wave distribution associated with the initial estimate to obtain an updated estimation of a weight matrix for use with the depth dependent 1 regularized non-linear Born iterative reconstruction method;
(3) generate an updated weight matrix based at least on the updated estimation of the weight matrix; and
(4) perform at least two iterations of the depth dependent 1 regularized non-linear Born iterative reconstruction method, at least one of the at least two iterations using the updated weight matrix; and
display the generated functional image.
13. The computer-readable storage medium of claim 12 , wherein the computer-executable instructions further cause the processor to calculate an initial estimate of the functional image using a dual-grid schedule to discretize the legion region with a fine grid and to discretize the background volume of the subject with a coarse grid to generate a weighted matrix for the lesion region and a weighted matrix for the background volume based on absorption of the optical waves on the subject.
14. The computer-readable storage medium of claim 13 , wherein the computer-executable instructions further cause the processor to perform reconstruction via the depth dependent 1 regularized non-linear Born iterative reconstruction method for generating the functional image iteratively by updating the photon-density wave distribution with a conjugate gradient method inside the fine grid.
15. The computer-readable storage medium of claim 12 , wherein the computer-executable instructions further cause the processor to receive the plurality of co-registered ultrasound images of the lesion region from an ultrasound imaging device.
16. The computer-readable storage medium of claim 12 , wherein the computer-executable instructions further cause the processor to generate near-infrared (NIR) optical waves at four optical wavelengths in a range from 730 nm to 830 nm.
17. The computer-readable storage medium of claim 12 , wherein the depth dependent 1 regularized non-linear Born iterative reconstruction method is configured to reconstruct at least one of values or distributions of absorption coefficients associated with absorption of the optical waves on the subject.