Method and system for medical decision support using organ models and learning based discriminative distance functions
View Patent ↗A method and system for providing medical decision support based on virtual organ models and learning based discriminative distance functions is disclosed. A patient-specific virtual organ model is generated from medical image data of a patient. One or more similar organ models to the patient-specific organ model are retrieved from a plurality of previously stored virtual organ models using a learned discriminative distance function. The patient-specific valve model can be classified into a first class or a second class based on the previously stored organ models determined to be similar to the patient-specific organ model.
1. A method for providing medical decision support using virtual organ models, comprising:
generating a virtual patient-specific organ model from medical image data of a patient; and
retrieving, from a plurality of previously stored virtual organ models, one or more similar organ models to the patient-specific organ model using a learned discriminative distance function, wherein the learned discriminative distance function is trained based on equivalence constraints assigned to training data.
2. The method of claim 1 , wherein the medical image data comprises at least one of a 3D volume or a 4D sequence of medical image data.
3. The method of claim 1 , wherein the step retrieving, from a plurality of previously stored virtual organ models, one or more similar organ models to the patient-specific organ model using a learned discriminative distance function, from a plurality of previously stored organ models, one or more similar organ models to the patient-specific organ model using a learned discriminative distance function comprises:
detecting the one or more similar organ models to the patient-specific organ model using a learned discriminative distance function based on features of the patient-specific organ model, the features including clinical measurements extracted from the patient-specific organ model.
4. The method of claim 1 , wherein the patient-specific organ model comprises at least one of an aortic valve model, a mitral valve model, or a pulmonary valve model.
5. The method of claim 1 , wherein the learned discriminative distance function is trained based on the equivalence constraints assigned to the training data using one of Random Forests or a boosting algorithm.
6. The method of claim 1 , further comprising:
classifying the patient-specific organ model into one of a first class or a second class based on the retrieved one or more similar organ models to the patient-specific organ model.
7. The method of claim 6 , wherein the step of classifying the patient-specific organ model into one of a first class or a second class based on the retrieved one or more similar organ models to the patient-specific organ model comprises:
clustering the plurality of previously stored organ models into a first cluster corresponding to the first class and a second cluster corresponding to the second class; and
assigning the patient-specific organ model to one of the first cluster or the second cluster based on the retrieved one or more similar organ models to the patient-specific organ model.
8. The method of claim 6 , wherein the step of classifying the patient-specific organ model into one of a first class or a second class based on the retrieved one or more similar organ models to the patient-specific organ model comprises:
classifying the patient-specific organ model into one of a healthy class or a diseased class based on the retrieved one or more similar organ models to the patient-specific organ model.
9. The method of claim 6 , wherein the patient specific organ model comprises a patient-specific pulmonary valve model, and the step of classifying the patient-specific valve model into one of a first class or a second class based on the retrieved one or more similar organ models to the patient-specific organ model comprises:
classifying the patient-specific pulmonary valve model as suitable or non-suitable for percutaneous pulmonary valve implantation (PPVI) based on the retrieved one or more similar organ models.
10. The method of claim 1 , wherein the patient-specific organ model represents a diseased organ and the retrieved one or more similar organ models represent a healthy organ corresponding to the diseased organ represented by the patient-specific organ model, the method further comprising:
automatically determining personalized treatment targets for said diseased organ based on the retrieved one or more similar organ models.
11. The method of claim 1 , further comprising:
displaying a neighborhood graph visualizing similarities between the patient-specific organ model and the plurality of previously stored organ models based on similarity measures determined using the learned discriminative distance function.
12. An apparatus for providing medical decision support using virtual organ models, comprising:
means for generating a virtual patient-specific organ model from cardiac image data of a patient; and
means for retrieving, from a plurality of previously stored virtual organ models, one or more similar organ models to the patient-specific organ model using a learned discriminative distance function, wherein the learned discriminative distance function is trained based on equivalence constraints assigned to training data.
13. The apparatus of claim 12 , wherein the means for retrieving, from a plurality of previously stored virtual organ models, one or more similar organ models to the patient-specific organ model using a learned discriminative distance function comprises:
means for detecting the one or more similar organ models to the patient-specific organ model using a learned discriminative distance function based on features of the patient-specific organ model, the features including clinical measurements extracted from the patient-specific organ model.
14. The apparatus of claim 12 , wherein the learned discriminative distance function is trained based on the equivalence constraints assigned to the training data using one of Random Forests or a boosting algorithm.
15. The apparatus of claim 12 , further comprising:
means for classifying the patient-specific organ model into one of a first class or a second class based on the retrieved one or more similar organ models to the patient-specific organ model.
16. The apparatus of claim 15 , wherein the means for classifying the patient-specific organ model into one of a first class or a second class based on the retrieved one or more similar organ models to the patient-specific organ model comprises:
means for clustering the plurality of previously stored organ models into a first cluster corresponding to the first class and a second cluster corresponding to the second class; and
means for assigning the patient-specific organ model to one of the first cluster or the second cluster based on the retrieved one or more similar organ models to the patient-specific organ model.
17. The apparatus of claim 15 , wherein the means for classifying the patient-specific organ model into one of a first class or a second class based on the retrieved one or more similar organ models to the patient-specific organ model comprises:
means for classifying the patient-specific organ model into one of a healthy class or a diseased class based on the retrieved one or more similar organ models to the patient-specific organ model.
18. The apparatus of claim 15 , wherein the patient specific organ model comprises a patient-specific pulmonary valve model, and the means for classifying the patient-specific organ model into one of a first class or a second class based on the retrieved one or more similar organ models to the patient-specific organ model comprises:
means for classifying the patient-specific pulmonary valve model as suitable or non-suitable for percutaneous pulmonary valve implantation (PPVI) based on the retrieved one or more similar organ models.
19. The apparatus of claim 12 , wherein the patient-specific organ model represents a diseased organ and the retrieved one or more similar organ models represent a healthy organ corresponding to the diseased organ represented by the patient-specific organ model, the apparatus further comprising:
means for automatically determining personalized treatment targets for said diseased organ based on the retrieved one or more similar organ models.
20. The apparatus of claim 12 , further comprising:
means for displaying a neighborhood graph visualizing similarities between the patient-specific organ model and the plurality of previously stored organ models based on similarity measures determined using the learned discriminative distance function.
21. A non-transitory computer readable medium encoded with computer executable instructions for providing medical decision support using virtual organ models, the computer executable instructions defining steps comprising:
generating a patient-specific virtual organ model from medical image data of a patient; and
retrieving, from a plurality of previously stored virtual organ models, one or more similar organ models to the patient-specific organ model using a learned discriminative distance function, wherein the learned discriminative distance function is trained based on equivalence constraints assigned to training data.
22. The computer readable medium of claim 21 , wherein the computer executable instructions defining the step of retrieving, from a plurality of previously stored virtual organ models, one or more similar organ models to the patient-specific organ model using a learned discriminative distance function comprise computer executable instructions defining the step of:
detecting the one or more similar organ models to the patient-specific organ model using a learned discriminative distance function based on features of the patient-specific organ model, the features including clinical measurements extracted from the patient-specific organ model.
23. The computer readable medium of claim 21 , wherein the learned discriminative distance function is trained based on the equivalence constraints assigned to the training data using one of Random Forests or a boosting algorithm.
24. The computer readable medium of claim 21 , further comprising computer executable instructions defining the step of:
classifying the patient-specific organ model into one of a first class or a second class based on the retrieved one or more similar organ models to the patient-specific organ model.
25. The computer readable medium of claim 24 , wherein the computer executable instructions defining the step of classifying the patient-specific organ model into one of a first class or a second class based on the retrieved one or more similar organ models to the patient-specific organ model comprise computer executable instructions defining the steps of:
clustering the plurality of previously stored organ models into a first cluster corresponding to the first class and a second cluster corresponding to the second class; and
assigning the patient-specific organ model to one of the first cluster or the second cluster based on the retrieved one or more similar organ models to the patient-specific organ model.
26. The computer readable medium of claim 24 , wherein the computer executable instructions defining the step of classifying the patient-specific organ model into one of a first class or a second class based on the retrieved one or more similar organ models to the patient-specific organ model comprise computer executable instructions defining the step of:
classifying the patient-specific organ model into one of a healthy class or a diseased class based on the retrieved one or more similar organ models to the patient-specific organ model.
27. The computer readable medium of claim 24 , wherein the patient specific organ model comprises a patient-specific pulmonary valve model, and the computer executable instructions defining the step of classifying the patient-specific organ model into one of a first class or a second class based on the retrieved one or more similar organ models to the patient-specific organ model comprise computer executable instructions defining the step of:
classifying the patient-specific pulmonary valve model as suitable or non-suitable for percutaneous pulmonary valve implantation (PPVI) based on the retrieved one or more similar organ models.
28. The computer readable medium of claim 21 , wherein the patient-specific organ model represents a diseased organ and the retrieved one or more similar organ models represent a healthy heart valve corresponding to the diseased organ represented by the patient-specific organ model, further comprising computer executable instructions defining the step of:
automatically determining personalized treatment targets for said diseased organ based on the retrieved one or more similar organ models.
29. The computer readable medium of claim 21 , further comprising computer executable instructions defining the step of:
displaying a neighborhood graph visualizing similarities between the patient-specific organ model and the plurality of previously stored organ models based on similarity measures determined using the learned discriminative distance function.