Method and apparatus for producing contrained medical image data
An apparatus for producing constrained medical image data, the apparatus including processing circuitry configured to: receive medical image data that includes or is obtained from scan data representing an anatomical region in which a sub-region is enhanced; predict, using a trained model, mask data from the medical imaging data, wherein the mask data is representative of the anatomical region without enhancement of the sub-region; and predict, using the trained model or a further trained model, subtraction data from the same medical image data, the subtraction data being representative of the same anatomical region, and the processing circuitry being further configured to apply at least one constraint to obtain constrained subtraction data.
1. An apparatus for producing constrained medical image data, the apparatus comprising:
processing circuitry configured to:
receive medical image data that comprises or is obtained from scan data representing an anatomical region in which a sub-region is enhanced;
predict, using a trained model, mask data from the medical image data, wherein the mask data is representative of the anatomical region without enhancement of the sub-region; and
predict, using the trained model or a further trained model, subtraction data from the same medical image data, wherein the subtraction data is representative of the same anatomical region, wherein
each of the medical image data, the mask data, and the subtraction data comprises a respective intensity for each of a plurality of locations corresponding to a plurality of locations within the anatomical region, and
the processing circuitry is further configured to apply at least two constraints to obtain constrained subtraction data, the at least two constraints comprising
a first constraint ensuring that, for each of the plurality of locations, a sum of an intensity in the subtraction data and an intensity in the mask data corresponds to an intensity in the medical image data, and
a second constraint ensuring that each of the intensities for each of the plurality of locations is non-negative.
2. The apparatus according to claim 1 , wherein the applying of the at least two constraints comprises performing an optimization procedure based on the mask data and the subtraction data.
3. The apparatus according to claim 1 , wherein the processing circuitry is further configured to obtain constrained mask data.
4. The apparatus according to claim 1 ,
wherein the first constraint comprises ensuring that each of the intensities for each of the plurality of locations is within a predetermined range of intensity values.
5. The apparatus according to claim 1 , wherein the applying of the at least two constraints comprises applying the at least two constraints within the trained model.
6. The apparatus according to claim 1 , wherein the trained model comprises a plurality of layers and the applying of the at least two constraints comprises applying the at least two constraints using at least one layer of the plurality of layers.
7. The apparatus according to claim 1 , wherein the applying of the at least two constraints is performed as part of the prediction of the mask data and/or the prediction of the subtraction data.
8. The apparatus according to claim 1 , wherein a further trained model is used to predict the subtraction data.
9. The apparatus according to claim 1 , wherein the trained model comprises a first convolutional neural network (CNN) and the further trained model comprises a second, independent CNN.
10. The apparatus according to claim 1 , wherein the predicting of the mask data comprises using at least two different trained models and/or the predicting of the subtraction data comprises using at least two different trained models.
11. The apparatus according to claim 1 , wherein the trained model comprises a multi-head CNN, and the trained model is used in the predicting of both the mask data and the subtraction data.
12. The apparatus according to claim 2 , wherein the optimization procedure comprises finding a closest possible subtraction volume to the predicted subtraction volume that satisfies the at least two constraints, and designating the closest possible subtraction volume as the constrained subtraction volume.
13. The apparatus according to claim 2 , wherein an objective of the optimization procedure comprises optimization of a 2-norm of a difference between predicted mask data and constrained mask data and a difference between predicted subtraction data and constrained subtraction data.
14. The apparatus according to claim 1 , wherein the processing circuitry is further configured to:
apply an additional trained model to the medical image data to obtain a segmentation of at least one metal object in the medical image data; and
use the segmentation to ensure that the at least one metal object is represented in the mask data.
15. The apparatus according to claim 1 , wherein the scan data represents the anatomical region with contrast agent present, the sub-region is enhanced in the medical image by presence of the contrast agent, and the mask data is representative of the anatomical region without the contrast agent.
16. The apparatus according to claim 1 , wherein the scan data comprises at least one of 3D rotational angiography data, 2D fluoroscopy data, CT data, MR data.
17. The apparatus according to claim 1 , wherein the scan data is obtained from a cone-beam C-arm acquisition.
18. A method for producing constrained medical image data, the method comprising:
receiving medical image data that comprises or is obtained from scan data representing an anatomical region in which a sub-region is enhanced;
predicting, using a trained model, mask data from the medical image data, wherein the mask data is representative of the anatomical region without enhancement of the sub-region;
predicting, using the trained model or a further trained model, subtraction data from the same medical image data, wherein the subtraction data is representative of the same anatomical region, and each of the medical image data, the mask data, and the subtraction data comprises a respective intensity for each of a plurality of locations corresponding to a plurality of locations within the anatomical region; and
applying at least two constraints to obtain constrained subtraction data, the at least two constraints comprising
a first constraint ensuring that, for each of the plurality of locations, a sum of an intensity in the subtraction data and an intensity in the mask data corresponds to an intensity in the medical image data, and
a second constraint ensuring that each of the intensities for each of the plurality of locations is non-negative.
19. The apparatus according to claim 1 , wherein
the processing circuitry is configured to:
receive medical image data that comprises or is obtained from scan data representing an anatomical region in which a sub-region is enhanced by presence of a contrast agent;
produce, using the trained model, mask data from the medical image data, wherein the mask data includes bone and/or soft tissue and is representative of the anatomical region without enhancement by the presence of the contrast agent in the sub-region;
produce, using the trained model or the further trained model, subtraction data from the same medical image data, wherein the subtraction data is representative of the same anatomical region and excludes the bone and/or soft tissue but includes the sub-region where the contrast agent is present; and
apply the at least two constraints to the subtraction data and the mask data to obtain constrained subtraction data.