Systems and methods for data and model-driven image reconstruction and enhancement
View Patent ↗Systems and methods are disclosed for image reconstruction and enhancement, using a computer system. One method includes acquiring a plurality of images associated with a target anatomy; determining, using a processor, one or more associations between subdivisions of localized anatomy of the target anatomy identified from the plurality of images, and local image regions identified from the plurality of images; performing an initial image reconstruction based on image acquisition information of the target anatomy; and updating the initial image reconstruction or generating a new image reconstruction based on the image acquisition information and the one or more determined associations.
1. A computer-implemented method of medical image reconstruction, the method comprising:
receiving an image acquisition of a target anatomy of a patient;
determining a vessel centerline model using the image acquisition of the target anatomy of the patient;
receiving a plurality of prior vessel centerline models corresponding to the target anatomy, the plurality of prior vessel centerline models being associated with prior image acquisitions of the patient and/or image acquisitions of prior patients;
determining a plurality of corresponding centerline points between points of the vessel centerline model and points of the plurality of prior vessel centerline models;
determining, for at least one point of the image acquisition, at least one local image prior by matching the at least one point of the image acquisition with a corresponding centerline point in at least one of the plurality of prior vessel centerline models, the at least one local image prior comprising an image of a localized region of one or more of the prior image acquisitions of the patient and/or a localized region of one or more of the image acquisitions of prior patients; and
determining an image reconstruction of the image acquisition of the target anatomy of the patient based on the at least one local image prior.
2. The method of claim 1 , wherein the plurality of prior vessel centerline models are determined in a machine learning training phase.
3. The method of claim 1 , wherein determining at least one local image prior further comprises:
merging image regions associated with each of the plurality of corresponding centerline points.
4. The method of claim 3 , wherein merging image regions comprises averaging image regions associated with each of the plurality of corresponding centerline points.
5. The method of claim 1 , wherein determining an image reconstruction of the image acquisition further comprises:
blending the at least one local image prior with a corresponding portion of the image acquisition.
6. The method of claim 1 , further comprising:
iteratively performing steps of determining at least one local image prior and determining an image reconstruction until a convergence between the image reconstruction and the image reconstruction of a prior iteration reaches a predetermined threshold.
7. The method of claim 1 , wherein at least a portion of the image reconstruction is outputted to a storage device and/or display.
8. A system for image reconstruction, the system comprising:
at least one memory storing instructions; and
at least one processor executing the instructions to perform operations comprising:
receiving an image acquisition of a target anatomy of a patient;
determining a vessel centerline model using the image acquisition of the target anatomy of the patient;
receiving a plurality of prior vessel centerline models corresponding to the target anatomy, the plurality of prior vessel centerline models being associated with prior image acquisitions of the patient and/or image acquisitions of prior patients;
determining a plurality of corresponding centerline points between points of the vessel centerline model and points of the plurality of prior vessel centerline models;
determining, for at least one point of the image acquisition, at least one local image prior by matching the at least one point of the image acquisition with a corresponding centerline point in at least one of the plurality of prior vessel centerline models, the at least one local image prior comprising an image of a localized region of one or more of the prior image acquisitions of the patient and/or a localized region of one or more of the image acquisitions of prior patients; and
determining an image reconstruction of the image acquisition of the target anatomy of the patient based on the at least one local image prior.
9. The system according to claim 8 , wherein the plurality of prior vessel centerline models are determined in a machine learning training phase.
10. The system according to claim 8 , wherein determining at least one local image prior further comprises:
merging image regions associated with each of the plurality of corresponding centerline points.
11. The system according to claim 10 , wherein merging image regions comprises averaging image regions associated with each of the plurality of corresponding centerline points.
12. The system according to claim 8 , wherein determining an image reconstruction of the image acquisition further comprises:
blending the at least one local image prior with a corresponding portion of the image acquisition.
13. The system according to claim 8 , further comprising:
iteratively performing steps determining at least one local image prior and determining an image reconstruction until a convergence between the image reconstruction and the image reconstruction of a prior iteration reaches a predetermined threshold.
14. The system according to claim 8 , wherein at least a portion of the image reconstruction is outputted to a storage device and/or display.
15. A non-transitory computer readable medium for use on a computer system containing computer-executable programming instructions for performing a method of medical image reconstruction, the method comprising:
receiving an image acquisition of a target anatomy of a patient;
determining a vessel centerline model using the image acquisition of the target anatomy of the patient;
receiving a plurality of prior vessel centerline models corresponding to the target anatomy, the plurality of prior vessel centerline models being associated with prior image acquisitions of the patient and/or image acquisitions of prior patients;
determining a plurality of corresponding centerline points between points of the vessel centerline model and points of the plurality of prior vessel centerline models;
determining, for at least one point of the image acquisition, at least one local image prior by matching the at least one point of the image acquisition with a corresponding centerline point in at least one of the plurality of prior vessel centerline models, the at least one local image prior comprising an image of a localized region of one or more of the prior image acquisitions of the patient and/or a localized region of one or more of the image acquisitions of prior patients; and
determining an image reconstruction of the image acquisition of the target anatomy of the patient based on the at least one local image prior.
16. The computer-executable programming of claim 15 , wherein the plurality of prior vessel centerline models are determined in a machine learning training phase.
17. The computer-executable programming of claim 15 , wherein determining at least one local image prior further comprises:
merging image regions associated with each of the plurality of corresponding centerline points.
18. The computer-executable programming of claim 17 , wherein merging image regions comprises averaging image regions associated with each of the plurality of corresponding centerline points.
19. The computer-executable programming of claim 15 , wherein determining an image reconstruction of the image acquisition further comprises:
blending the at least one local image prior with a corresponding portion of the image acquisition.
20. The computer-executable programming of claim 15 , further comprising:
iteratively performing steps determining at least one local image prior and determining an image reconstruction until a convergence between the image reconstruction and the image reconstruction of a prior iteration reaches a predetermined threshold.