Anatomical measurements from ultrasound data

1. A computer-implemented method of generating a model for estimating a 3-D anatomical body measurement value from at least one 2-D ultrasound image including a contour of the anatomical body, the method comprising:

providing a set of 3-D ultrasound images of the anatomical body;

for each of the 3-D ultrasound images in the set of 3-D ultrasound images, determining a ground truth value of the 3-D anatomical body measurement;

generating from the set of 3-D ultrasound images, a set of 2-D ultrasound image planes, each including a contour of the anatomical body and, for each of the 2-D ultrasound image planes, estimating a value of the 3-D anatomical body measurement from at least one of an outline contour measurement and a cross-sectional measurement of the anatomical body in the 2-D ultrasound image plane;

training a machine-learning algorithm by inputting to the machine-learning algorithm the ground truth values of the 3-D anatomical body measurements, the corresponding estimated values of the 3-D anatomical body measurements, and at least one of: the corresponding outline contour measurements and the corresponding cross-sectional measurements of the anatomical body in the 2-D ultrasound image planes, to generate a model for estimating a 3-D anatomical body measurement value from at least one subsequently-acquired 2-D ultrasound image comprising at least one of an outline contour measurement and a cross-sectional measurement of a contour of the anatomical body.

2. The computer-implemented method of claim 1 , further comprising training the machine-learning algorithm to generate a further estimator function (g) to estimate the uncertainty in the estimated 3-D anatomical body measurement value that it generates in use, the training being done by inputting to the machine-learning algorithm the obtained ground truth values, the corresponding estimated values of the 3-D anatomical body measurement and at least one of the corresponding outline contour measurement and the corresponding cross-sectional measurement of the anatomical body in the 2-D ultrasound image planes.

3. The computer-implemented method of claim 1 , wherein determining a ground truth value of the 3-D anatomical body measurement comprises:

mapping a segmentation model of the anatomical body to said anatomical body within the 3-D ultrasound image; and

determining the ground truth value of the 3-D anatomical body measurement from the mapped segmentation model.

4. The computer-implemented method of claim 3 , wherein generating the set of 2-D image planes from the 3-D ultrasound image comprises generating the set of 2-D image planes from the 3-D ultrasound image based on the segmentation model that has been mapped to the anatomical body within the 3-D ultrasound image.

5. The computer-implemented method of claim 3 , wherein the anatomical body is a human heart and the segmentation model is a heart model.

6. The computer-implemented method of claim 1 , wherein generating the set of 2-D image planes comprises providing a first set of 2-D image slices providing a 2-chamber view of a human heart and a second set of 2-D image slices providing a 4-chamber view of the human heart.

7. The computer-implemented method of claim 1 , wherein generating the set of 2-D ultrasound image planes comprises providing a plurality of subsets of 2-D ultrasound image planes, wherein the 2-D ultrasound image planes in each subset are under a defined angle with a defined axis of the 3-D ultrasound images, said defined angles differing between subsets.

8. A computer-implemented method of estimating a 3-D anatomical body measurement value from at least one 2-D ultrasound image including a contour of the anatomical body using the model obtained by the computer-implemented method of any preceding claim, comprising:

providing the at least one 2-D ultrasound image including a contour of the anatomical body;

determining at least one of an outline contour measurement and a cross-sectional measurement of the anatomical body from the at least one 2-D ultrasound image;

providing at least one of the determined outline contour measurement and the determined cross-sectional measurement as an input to said model;

receiving an estimated 3-D anatomical body measurement value from the model in response to said input; and

receiving an estimated uncertainty in the value of the estimated 3-D anatomical body measurement value from the model in response to said input.

9. The computer-implemented method of claim 8 , further comprising identifying a plane orientation of the at least one 2-D ultrasound image relative to the anatomical body for which said estimated uncertainty is minimized and communicating said plane orientation to a user or to an ultrasound probe.

10. The computer-implemented method of claim 8 , wherein the anatomical body is the human heart and the estimated 3-D measurement value is a volume of at least a part of the human heart, and wherein providing the at least one 2-D ultrasound image including a contour of the anatomical body comprises:

providing a first 2-D ultrasound image including a contour of the human heart at a first point of a cardiac cycle;

providing a second 2-D ultrasound image including a contour of the human heart at a second point of a cardiac cycle; and

deriving an estimation of an ejection fraction of the human heart from a difference between the estimated volume for the first 2-D ultrasound image and the estimated volume for the second 2-D ultrasound image.

11. A non-transitory computer readable storage medium comprising computer readable program instructions embodied therewith for, when executed on a processor arrangement of an ultrasound image processing apparatus, causing the processor arrangement to implement the method of claim 1 .

12. A non-transitory computer readable storage medium comprising computer readable program instructions embodied therewith for, when executed on a processor arrangement of an ultrasound image processing apparatus, causing the processor arrangement to implement the method of claim 8 .