Estimation of solitary pulmonary nodule diameters with a hybrid segmentation approach

1. A computer-implemented method for estimating a diameter of an object of interest, comprising: a computer to implement the steps of:

providing a volume of interest comprising a marker identifying a location of the object of interest;

extracting a sub-volume of voxels around the marker; initializing at least two volumes of the sub-volume, wherein each of the voxels has at least two values corresponding to the at least two volumes, respectively;

performing, iteratively, a diffusion operation and a reaction operation on the voxels to update at least two values;

comparing, for each voxel, the at least two values to a threshold to assign each voxel to one of the at least two volumes, wherein the assignment of the voxels is a segmentation result;

performing a boundary check to determine whether the object of interest is solitary;

estimating a diameter of the object of interest from the segmentation result upon determining the object of interest to be solitary; and

extracting the object of interest by performing ellipsoid fit on the sub-volume and estimating a diameter of the object of interest upon determining the object of interest to be non-solitary.

2. The computer-implemented method of claim 1 , wherein the ellipsoid fit is a Gaussian method.

3. The computer-implemented method of claim 1 , wherein if the segmentation result touches one of six boundaries of the sub-volume, it is defined as non-solitary.

4. The computer-implemented method of claim 1 , further comprising determining a ratio between the segmentation result and a volume of the sub-volume enclosing the segmentation result to determine if the segmentation has a reasonably spherical shape.

5. The computer-implemented method of claim 1 , wherein the at least two volumes include a foreground volume and a background volume.

6. The computer-implemented reactive-diffusion method of claim 5 , wherein the at least two values correspond to a foreground label and a background label, respectively.

7. The computer-implemented method of claim 1 , wherein a number of iterations of the performing of the diffusion and reaction operations is predetermined.

8. The computer-implemented method of claim 1 , wherein performing the diffusion operation comprises performing a linear diffusion.

9. The computer-implemented reactive-diffusion method of claim 8 , wherein the linear diffusion is given by x i Ý (s)=λ∇ 2 x i (s), where iε{0,1} and the diffusion rate λ is set to 0.5.

10. The computer-implemented method of claim 1 , wherein performing the reaction operation comprises performing a competition between the at least two values in a form of replicator dynamics implementing a fitness matrix, wherein the replicator dynamics updates the at least two values.

11. The computer-implemented reactive-diffusion method of claim 10 , wherein performing the reaction operation comprises enforcing that each of the at least two values is greater than or equal to zero and the sum of the at least two values is one.

12. The computer-implemented reactive-diffusion method of claim 11 , wherein the diffusion operation adds disturbance to break a tie between the at least two values in the reaction operation.

13. A computer storage medium readable by a computer, tangibly embodying a program of instructions executable by the computer to perform method steps for estimating a diameter of an object of interest, the method steps comprising:

providing a volume of interest comprising a marker identifying a location of the object of interest; extracting a sub-volume of voxels around the marker; initializing at least two volumes of the sub-volume, wherein each of the voxels has at least two values corresponding to the at least two volumes, respectively; performing, iteratively, a diffusion operation and a reaction operation on the voxels to update at least two values; comparing, for each voxel, the at least two values to a threshold to assign each voxel to one of the at least two volumes, wherein the assignment of the voxels is a

segmentation result; performing a boundary check to determine whether the object of interest is solitary; estimating a diameter of the object of interest from the segmentation result upon determining the object of interest to be solitary; and extracting the object of interest by performing ellipsoid fit on the sub-volume and estimating a diameter of the object of interest upon determining the object of interest to be non-solitary thus reducing the diameter estimation bias that is introduced with small dimensions modules.