Multi-level contextual learning of data
View Patent ↗Described herein is a framework for automatically classifying a structure in digital image data are described herein. In one implementation, a first set of features is extracted from digital image data, and used to learn a discriminative model. The discriminative model may be associated with at least one conditional probability of a class label given an image data observation Based on the conditional probability, at least one likelihood measure of the structure co-occurring with another structure in the same sub-volume of the digital image data is determined. A second set of features may then be extracted from the likelihood measure.
1. A computer-implemented method for classifying a structure in digital image data, comprising:
(a) extracting a first set of features from the digital image data;
(b) learning a discriminative model based on the first set of features, wherein the discriminative model is associated with at least one conditional probability of a class label given an image data observation;
(c) determining, based on the at least one conditional probability, at least one likelihood measure of the structure co-occurring with another structure in a same sub-volume of the digital image data; and
(d) extracting a second set of features from the one or more likelihood measures.
2. The method of claim 1 wherein the extracting the first set of features from the image data comprises extracting at least one texture feature, gray-level feature, shape feature, or a combination thereof.
3. The method of claim 1 wherein the extracting the first set of features from the image data comprises applying a Hessian enhancement filter to the image data.
4. The method of claim 3 wherein the applying the Hessian enhancement filter comprises iteratively applying the Hessian enhancement filter to the image data until a stop condition is met.
5. The method of claim 1 wherein the learning the discriminative model comprises constructing a conditional random field model.
6. The method of claim 1 wherein the learning the discriminative model comprises learning one or more parameters of the discriminative model using a supervised classifier.
7. The method of claim 6 wherein the supervised classifier comprises a generative classifier.
8. The method of claim 7 wherein the generative classifier comprises a resilient subclass discriminant analysis classifier.
9. The method of claim 6 wherein the supervised classifier comprises a discriminative classifier.
10. The method of claim 9 wherein the discriminative classifier comprises a relevance vector machine classifier.
11. The method of claim 1 further comprising performing a graph-cut method to determine optimal labeling based on the discriminative model.
12. The method of claim 1 wherein the determining the at least one likelihood measure comprises determining at least one probability co-occurrence map.
13. The method of claim 12 wherein the extracting the second set of features comprises summing multiple probability co-occurrence maps.
14. A non-transitory computer readable medium embodying a program of instructions executable by a machine to perform steps for classifying a structure in digital image data, the steps comprising:
(a) extracting a first set of features from the digital image data;
(b) learning a discriminative model based on the first set of features, wherein the discriminative model is associated with at least one conditional probability of a class label given an image data observation;
(c) determining, based on the at least one conditional probability, at least one likelihood measure of the structure co-occurring with another structure in a same sub-volume of the digital image data; and
(d) extracting a second set of features from the one or more likelihood measures.
15. A system for classifying a structure in digital image data, comprising:
a memory device for storing computer readable program code; and
a processor in communication with the memory device, the processor being operative with the computer readable program code to perform steps for classifying a structure in digital image data, the steps comprising:
(a) extracting a first set of features from the digital image data;
(b) learning a discriminative model based on the first set of features, wherein the discriminative model is associated with at least one conditional probability of a class label given an image data observation;
(c) determining, based on the at least one conditional probability, at least one likelihood measure of the structure co-occurring with another structure in the same sub-volume of the digital image data; and
(d) extracting a second set of features from the one or more likelihood measures.
16. The system of claim 15 wherein the digital image data comprises medical images of at least a portion of a lung.
17. The system of claim 16 wherein the at least one likelihood measure comprises a likelihood measure of a lung nodule co-occurring with another pulmonary structure in the same sub-volume of the digital image data.
18. The system of claim 17 wherein the another pulmonary structure comprises a lung fissure, pleural or vessel.