Medical anatomy quantification: computer-aided diagnosis tool

1. A method for segmenting an anatomical part, comprising:

identifying a landmark on a contour/surface of a 2D/3D model of the anatomical part in an ultrasound image;

assigning a weight to the landmark;

identifying different appearance patterns of a region around the landmark based on a previously stored training set; and

applying a filter to the different appearance patterns of the region around the landmark in order to identify contours of the anatomical part.

2. The method according to claim 1 , wherein the landmark on the contour/surface of the 2D/3D model of the anatomical part is identified by comparing the 2D/3D model of the anatomical part with a plurality of reference models.

3. The method according to claim 1 , wherein the weight is assigned to the landmark based on an angle between a propagation direction of an ultrasound wave front and a vector normal with respect to the landmark.

4. The method according to claim 1 , wherein a fuzzy clustering algorithm is used to identify the different appearance patterns of the region around the landmark.

5. The method according to claim 1 , wherein the training set includes appearance information from adjacent landmarks within a predetermined length from the landmark.

6. The method according to claim 1 , wherein the filter characterizes textures of the anatomical part with different dominant sizes.

7. The method according to claim 1 , wherein the different appearance patterns are created for a plurality of scales.

8. The method according to claim 1 , wherein the anatomical part is a kidney.

9. A method for segmenting a portion of an anatomical part, comprising:

identifying echogenic regions of the anatomical part in an ultrasound image to differentiate the portion of the anatomical part from other portions of the anatomical part;

generating, using the identified echogenic regions of the anatomical part, a positional map of the portion of the anatomical part; and

incorporating patient-specific constraints to delimitate the portion of the anatomical part.

10. The method according to claim 9 , wherein the echogenic regions are identified as fat located inside the portion of the anatomical part.

11. The method according to claim 9 , wherein the echogenic regions are identified by exploiting local phase properties of the anatomical part.

12. The method according to claim 9 , wherein the echogenic regions result in a natural anatomical constraint, and wherein the echogenic regions circumscribe the portion of the anatomical part.

13. The method according to claim 9 , wherein the positional map of the portion of the anatomical part is generated from the identified echogenic regions using alpha shapes.

14. The method according to claim 9 , wherein the patient-specific constraints are incorporated using the positional map of the portion of the anatomical part.

15. The method according to claim 9 , wherein the anatomical part is segmented by mimicking propagation of fluid inside the anatomical part, using the patient-specific constraints to control a propagation process.

16. The method according to claim 9 , wherein the anatomical part is a kidney, and the portion of the anatomical part is a collecting system.

17. A method for characterizing functionality of an anatomical part to identify severity of a medical condition, the method comprising:

delineating the anatomical part and segmenting a portion of the anatomical part in ultrasound images:

extracting morphological features of the anatomical part and the portion of the anatomical part using image analysis;

selecting an optimal subset of the morphological features using a supervised or an unsupervised feature selection framework; and

classifying each feature in the optimal subset as critical or non-critical based on a threshold and a classifier, the classifier being one of linear discriminant analysis or a support vector machine.

18. The method of claim 17 , further comprising:

grouping the extracted morphological features into a plurality of categories, the plurality of categories including size, geometric shape, curvature, and texture of the anatomical part and the portion of the anatomical part.

19. The method of claim 17 , further comprising:

identifying probability thresholds that maximize sensitivity of detecting severe cases of hydronephrosis using receiver operating characteristics (ROC).

20. The method of claim 17 , wherein the optimal subset is selected using an area under a receiver operating characteristics (ROC) curve.

21. The method of claim 18 , wherein the size includes a relative volume of the anatomical part and the portion of the anatomical part, relative area and perimeter of the anatomical part and the portion of the anatomical part, and maximum and minimum parenchyma thickness.

22. The method of claim 18 , wherein the geometric shape includes sphericity and eccentricity of the anatomical part and the portion of the anatomical part.

23. The method of claim 18 , wherein the curvature includes an average curvature of the anatomical part, and curvature dissimilarity between calices and the anatomical part.

24. The method of claim 17 , wherein the threshold is a parameter of a function of the anatomical part, the parameter of the function of the anatomical part including a washout half time of at least one of 20 mins, 30 mins, or 40 mins.

25. The method of claim 17 , wherein the anatomical part is a kidney, the portion of the anatomical part is a collecting system, and the medical condition is hydronephrosis.

26. The method of claim 17 , further comprising:

identifying the severity of the medical condition based on classifying each feature in the optimal subset as critical or non-critical.