Method for tracking soft tissue masses in images using directed graphs
View Patent ↗Region of interest (ROI) corresponding to a soft tissue mass are tracked in a training video acquired by sonography. The locations of the ROI are used to construct a directed graph in which each node represents a location of the tracked ROI, and the edges represent temporal relations of the ROIs. The soft tissue mass can also be tracked using the graph, and appropriate treatment can be delivered.
1. A computer implemented method for tracking a soft tissue mass, comprising the steps of:
acquiring a training video of a soft tissues mass inside a body using sonography, in which the training video includes a sequence of images;
marking a region of interest (ROI) corresponding to a location of the soft tissue mass in a first frame of the training video;
constructing a directed graph of nodes and edges in a memory while tracking the ROI in each image in the sequence of the training video, in which each node represent a location of the tracked ROI, and the edges represent temporal relations of the ROIs;
acquire a treatment video of the soft tissue mass using the sonography; and
tracking the soft tissue mass in the treatment video according to the directed graph.
2. The method of claim 1 , in which a location of the soft tissue mass is tracked.
3. The method of claim 2 , in which, the location is correlated to a cyclic behavior of respiration.
4. The method of claim 2 , in which the location is correlated to pulsation of a heart.
5. The method of claim 1 , in which a shape of the soft tissue mass is tracked.
6. The method of claim 1 , in which the soft tissue mass is a tumor, and the tracking is used to control a radiation beam directed at the tumor.
7. The method of claim 6 , in which the controlling turns the beam on and off.
8. The method of claim 6 , in which a depth of maximum exposure is controlled.
9. The method of claim 1 , further comprising;
increasing a resolution of the images according to the tracking.
10. The method of claim 1 , further comprising:
decreasing noise in the images according to the tracking.
11. The method of claim 1 , in which the sonography uses a Doppler ultrasound signal.
12. The method of claim 1 , in which the ROI is an entire image.
13. The method of claim 1 , in which the images are normalized.
14. The method of claim 1 , in which the tracking matches each image with the ROIs represented in the graph using a normalized cross-correlation to obtain a score.
15. The method of claim 1 , in which the tracking matches each image with the ROIs represented in the graph using a sum-of-squared difference metric.
16. The method of claim 15 , in which each edge is weighted according to a normalized number of transitions made from a current node to a child node, and the matching score.
17. The method of claim 1 , in which the images are stored in a buffer in a memory.
18. The method of claim 17 , in which images stored in the buffer are replaced according to a similarity score.
19. The method of claim 17 , in which images stored in the buffer as clusters, and the images in the clusters are replaced according to a cluster distance.
20. The method of claim 17 , in which images stored in the buffer as clusters using an affinity matrix, and an eigenvector decomposition.
21. The method of claim 1 , further comprising:
representing each image with a descriptor.
22. The method of claim 21 , in which the descriptor is a histogram, a covariance matrix, a Gabor filter, Haar wavelets, a mixture of Gaussian functions, textons, edgelets, or a histogram of oriented gradients.
23. The method of claim 1 , in which the tracking uses a breadth-first search of the graph.
24. The method of claim 1 , in which the tracking uses a normalized cross-correlation distance, a sum-of-squared distance, an exhaustive search, regression tracking, mean-shift tracking, gradient search, or particle tracking.
25. A system for tracking a soft tissue mass, comprising:
means for acquiring a training video of a soft tissues mass inside a body using sonography, in which the training video includes a sequence of images;
means for marking a region of interest (ROI) corresponding to a location of the soft tissue mass in a first frame of the training video;
means for constructing a directed graph of nodes and edges in a memory while tracking the ROI in each image in the sequence of the training video, in which each node represent a location of the tracked ROI, and the edges represent temporal relations of the ROIs;
means for acquire a treatment video of the soft tissue mass using the sonography; and
means for tracking the soft tissue mass in the treatment video according to the directed graph.