AI-based region-of-interest masks for improved data reconstruction
Systems/techniques that facilitate AI-based region-of-interest masks for improved data reconstructions are provided. In various embodiments, a system can access a set of two-dimensional medical scan projections. In various aspects, the system can generate a set of two-dimensional region-of-interest masks respectively corresponding to the set of two-dimensional medical scan projections. In various instances, the system can generate a region-of-interest visualization based on the set of two-dimensional region-of-interest masks and the set of two-dimensional medical scan projections. In various cases, the system can generate the set of two-dimensional region-of-interest masks by executing a machine learning segmentation model on the set of two-dimensional medical scan projections.
1. A system, comprising:
a processor that executes computer-executable components stored in a computer-readable memory, the computer-executable components comprising:
a projection component that accesses a set of two-dimensional medical scan projections associated with a three-dimensional scan of a body part of a patient;
a mask component that generates a set of two-dimensional region-of-interest masks respectively corresponding to the set of two-dimensional medical scan projections, wherein the set of two-dimensional region-of-interest masks identify respective probabilities of pixels of two-dimensional medical scan projections of the set of two-dimensional medical scan projections belonging to one or more pathology regions; and
a reconstruction component that generates a three-dimensional region-of-interest visualization of the body part of the patient based on the set of two-dimensional region-of-interest masks and the set of two-dimensional medical scan projections, wherein the reconstruction component enhances one or more voxels of the three-dimensional region-of-interest visualization that are determined to be part of the one or more pathology regions based on respective probabilities of voxels of the three-dimensional region-of-interest visualization and a defined probability threshold, and wherein the respective probabilities of the voxels are based on the respective probabilities of pixels used to generate the voxels.
2. The system of claim 1 , wherein the mask component generates the set of two-dimensional region-of-interest masks by executing a machine learning segmentation model on the set of two-dimensional medical scan projections.
3. The system of claim 2 , wherein the machine learning segmentation model is configured to segment more than one of the set of two-dimensional medical scan projections at a time.
4. The system of claim 1 , wherein the projection component receives a three-dimensional medical scan volume, and wherein the projection component generates the set of two-dimensional medical scan projections by applying one or more projection algorithms to the three-dimensional medical scan volume.
5. The system of claim 4 , wherein the projection component receives a three-dimensional region-of-interest mask generated by a machine learning segmentation model based on the three-dimensional medical scan volume, and wherein the mask component generates the set of two-dimensional region-of-interest masks by applying one or more projection algorithms to the three-dimensional region-of-interest mask.
6. The system of claim 1 , wherein the reconstruction component generates a three-dimensional medical scan volume based on the set of two-dimensional medical scan projections, wherein the reconstruction component generates a three-dimensional region-of-interest mask based on the set of two-dimensional region-of-interest masks, and wherein the three-dimensional region-of-interest visualization includes slices of the three-dimensional region-of-interest mask overlaid onto respective slices of the three-dimensional medical scan volume.
7. The system of claim 6 , wherein the reconstruction component determines an overall likelihood of the one or more pathology regions being present in the three-dimensional medical scan volume, based on the three-dimensional region-of-interest mask comprising the respective probabilities of the voxels.
8. The system of claim 1 , wherein the three-dimensional region-of-interest visualization is a three-dimensional medical scan volume based on the set of two-dimensional medical scan projections and enhanced via weighting based on the respective probabilities of the voxels.
9. A computer-implemented method, comprising:
accessing, by a device operatively coupled to a processor, a set of two-dimensional medical scan projections associated with a three-dimensional scan of a body part of a patient;
generating, by the device, a set of two-dimensional region-of-interest masks respectively corresponding to the set of two-dimensional medical scan projections, wherein the set of two-dimensional region-of-interest masks identify respective probabilities of pixels of two-dimensional medical scan projections of the set of two-dimensional medical scan projections belonging to one or more pathology regions; and
generating, by the device, a three-dimensional region-of-interest visualization of the body part of the patient based on the set of two-dimensional region-of-interest masks and the set of two-dimensional medical scan projections, wherein the generating comprises enhancing one or more voxels of the three-dimensional region-of-interest visualization that are determined to be part of the one or more pathology regions based on respective probabilities of voxels of the three-dimensional region-of-interest visualization and a defined probability threshold, and wherein the respective probabilities of the voxels are based on the respective probabilities of pixels used to generate the voxels.
10. The computer-implemented method of claim 9 , wherein the generating the set of two-dimensional region-of-interest masks includes executing, by the device, a machine learning segmentation model on the set of two-dimensional medical scan projections.
11. The computer-implemented method of claim 10 , wherein the machine learning segmentation model is configured to segment more than one of the set of two-dimensional medical scan projections at a time.
12. The computer-implemented method of claim 9 , further comprising:
receiving, by the device, a three-dimensional medical scan volume; and
generating, by the device, the set of two-dimensional medical scan projections by applying one or more projection algorithms to the three-dimensional medical scan volume.
13. The computer-implemented method of claim 12 , further comprising:
receiving, by the device, a three-dimensional region-of-interest mask generated by a machine learning segmentation model based on the three-dimensional medical scan volume; and
generating, by the device, the set of two-dimensional region-of-interest masks by applying one or more projection algorithms to the three-dimensional region-of-interest mask.
14. The computer-implemented method of claim 9 , further comprising:
generating, by the device, a three-dimensional medical scan volume based on the set of two-dimensional medical scan projections; and
generating, by the device, a three-dimensional region-of-interest mask based on the set of two-dimensional region-of-interest masks, wherein the three-dimensional region-of-interest visualization includes slices of the three-dimensional region-of-interest mask overlaid onto respective slices of the three-dimensional medical scan volume.
15. The computer-implemented method of claim 14 , further comprising:
determining, by the device, an overall likelihood of the one or more pathology regions being present in the three-dimensional medical scan volume, based on the three-dimensional region-of-interest mask comprising the respective probabilities of the voxels.
16. The computer-implemented method of claim 9 , wherein the three-dimensional region-of-interest visualization is a three-dimensional medical scan volume based on the set of two-dimensional medical scan projections and enhanced via weighting based on the respective probabilities of the voxels.
17. A computer program product for facilitating AI-based region-of-interest masks for improved data reconstruction, the computer program product comprising a computer-readable memory having program instructions embodied therewith, the program instructions executable by a processor to cause the processor to:
access a set of two-dimensional medical scan projections associated with a three-dimensional scan of a body part of a patient;
generate a set of two-dimensional region-of-interest masks respectively corresponding to the set of two-dimensional medical scan projections, wherein the set of two-dimensional region-of-interest masks identify respective probabilities of pixels of two-dimensional medical scan projections of the set of two-dimensional medical scan projections belonging to one or more pathology regions; and
generate a three-dimensional region-of-interest visualization of the body part of the patient based on the set of two-dimensional region-of-interest masks and the set of two-dimensional medical scan projections, wherein the generating comprises enhancing one or more voxels of the three-dimensional region-of-interest visualization that are determined to be part of the one or more pathology regions based on respective probabilities of voxels of the three-dimensional region-of-interest visualization and a defined probability threshold, and wherein the respective probabilities of the voxels are based on the respective probabilities of pixels used to generate the voxels.
18. The computer program product of claim 17 , wherein the processor generates the set of two-dimensional region-of-interest masks by executing a machine learning segmentation model on the set of two-dimensional medical scan projections.
19. The computer program product of claim 18 , wherein the machine learning segmentation model is configured to segment more than one of the set of two-dimensional medical scan projections at a time.
20. The computer program product of claim 17 , wherein the program instructions are further executable to cause the processor to:
receive a three-dimensional medical scan volume; and
generate the set of two-dimensional medical scan projections by applying one or more projection algorithms to the three-dimensional medical scan volume.