Real-time collimation and ROI-filter positioning in X-ray imaging via automatic detection of the landmarks of interest
A method for real-time collimation and ROI-filter positioning in X-ray imaging in interventional procedures includes acquiring an image of a region-of-interest (ROI) at a beginning of a medical intervention procedure on a subject, classifying the image based on low-level features in the image to determine a type of procedure being performed, determining a list of landmarks in the image from the type of procedure being performed, and loading a pre-trained landmark model for each landmark in the list of landmarks, where landmarks include anatomical structures of the subject and medical devices being used in the medical intervention procedure, and computing collimator settings of an X-ray imaging device from ROI filter margins and bounding boxes of the landmarks calculated using the landmark models.
1. A method for real-time collimation and ROI-filter positioning in X-ray imaging in interventional procedures, comprising the steps of:
acquiring an image of a region-of-interest (ROI) at a beginning of a medical intervention procedure on a subject;
classifying the image based on low-level features in the image to determine a type of procedure being performed;
determining a list of landmarks in the image from the type of procedure being performed, and loading a pre-trained landmark model for each landmark in the list of landmarks, wherein landmarks include anatomical structures of the subject and medical devices being used in the medical intervention procedure; and
computing collimator settings of an X-ray imaging device from ROI filter margins and bounding boxes of the landmarks calculated using the landmark models.
2. The method of claim 1 , further comprising presenting to an operator the computed collimator settings of the X-ray imaging device, and receiving a confirmation of the operator the computed collimator settings from the operator.
3. The method of claim 1 , further comprising, during the medical intervention procedure, detecting that a landmark has moved into the ROI, and recomputing the ROI filter margins and the collimator settings of the X-ray imaging device.
4. The method of claim 1 , wherein the landmark models are used to detect landmarks in the image.
5. The method of claim 4 , wherein landmarks include balloon markers, guiding catheters, guidewires, intravascular ultrasound transducers, vessels, ablation catheter electrodes, circumferential mapping catheter electrodes, coronary sinus catheter electrodes, calcium, and bone structures.
6. The method of claim 4 , wherein if no pre-trained landmark model exits for a landmark, the method further comprises initializing the landmark, receiving positive samples of the landmark from a user, receiving negative samples of the landmark from other regions of the image, determining an appearance of the image from an image patch that represents an appearance of the initialized landmark, and training a landmark model using the positive samples, the negative samples, and the image appearance.
7. The method of claim 6 , wherein the landmark is initialized with one of a line segment, a curve, a bounding box, a circle or other similar types of structures.
8. The method of claim 1 , wherein the low-level features include edges, Haar-like features, scale-invariant features, histogram of ordered gradients, and local ternary patterns (LTP).
9. The method of claim 1 , wherein the image is classified by a classifier based on the type of procedure being performed, wherein the procedure types include an electrophysiology procedure, an intravascular ultrasound procedure, a chronic total occlusion (CTO) procedure, and a balloon angioplasty procedure.
10. A method for real-time collimation and ROI-filter positioning in X-ray imaging in interventional procedures, comprising the steps of:
acquiring an image of a region-of-interest (ROI) at a beginning of a medical intervention procedure on a subject;
classifying the image based on a type of procedure being performed;
detecting landmarks in the image, wherein the landmarks being detected are determined from the type of procedure being performed, wherein landmarks include anatomical structures of the subject and medical devices being used in the medical intervention procedure; and
computing collimator settings of an X-ray imaging device from ROI filter margins and bounding boxes of the landmarks calculated using the detected landmarks.
11. The method of claim 10 , further comprising determining a list of landmarks from the type of procedure being performed and loading a pre-trained landmark model for each landmark in the list of landmarks, wherein the landmark models are used to detect landmarks in the image.
12. The method of claim 11 , wherein if no pre-trained landmark model exits for a landmark, the method further comprises initializing the landmark, receiving positive samples of the landmark from a user, receiving negative samples of the landmark from other regions of the image, determining an appearance of the image from an image patch that represents an appearance of the initialized landmark, and training a landmark model using the positive samples, the negative samples, and the image appearance.
13. The method of claim 10 , further comprising presenting to an operator the computed collimator settings of the X-ray imaging device, and receiving a confirmation of the operator the computed collimator settings from the operator.
14. The method of claim 10 , further comprising, during the medical intervention procedure, detecting that a landmark has moved into the ROI, and recomputing the ROI filter margins and the collimator settings of the X-ray imaging device.
15. The method of claim 10 , wherein the image is classified using low-level features, wherein the low-level features include edges, Haar-like features, scale-invariant features, histogram of ordered gradients, and local ternary patterns (LTPs).
16. A non-transitory program storage device readable by a computer, tangibly embodying a program of instructions executed by the computer to perform the method steps for real-time collimation and ROI-filter positioning in X-ray imaging in interventional procedures, the method comprising the steps of:
acquiring an image of a region-of-interest (ROI) at a beginning of a medical intervention procedure on a subject;
classifying the image based on low-level features in the image to determine a type of procedure being performed;
determining a list of landmarks in the image from the type of procedure being performed, and loading a pre-trained landmark model for each landmark in the list of landmarks, wherein landmarks include anatomical structures of the subject and medical devices being used in the medical intervention procedure; and
computing collimator settings of an X-ray imaging device from ROI filter margins and bounding boxes of the landmarks calculated using the landmark models.
17. The computer readable program storage device of claim 16 , the method further comprising presenting to an operator the computed collimator settings of the X-ray imaging device, and receiving a confirmation of the operator the computed collimator settings from the operator.
18. The computer readable program storage device of claim 16 , the method further comprising, during the medical intervention procedure, detecting that a landmark has moved into the ROI, and recomputing the ROI filter margins and the collimator settings of the X-ray imaging device.
19. The computer readable program storage device of claim 16 , wherein the landmark models are used to detect landmarks in the image.
20. The computer readable program storage device of claim 19 , wherein landmarks include balloon markers, guiding catheters, guidewires, intravascular ultrasound transducers, vessels, ablation catheter electrodes, circumferential mapping catheter electrodes, coronary sinus catheter electrodes, calcium, and bone structures.
21. The computer readable program storage device of claim 19 , wherein if no pre-trained landmark model exits for a landmark, the method further comprises initializing the landmark, receiving positive samples of the landmark from a user, receiving negative samples of the landmark from other regions of the image, determining an appearance of the image from an image patch that represents an appearance of the initialized landmark, and training a landmark model using the positive samples, the negative samples, and the image appearance.
22. The computer readable program storage device of claim 21 , wherein the landmark is initialized with one of a line segment, a curve, a bounding box, a circle or other similar types of structures.
23. The computer readable program storage device of claim 16 , wherein the low-level features include edges, Haar-like features, scale-invariant features, histogram of ordered gradients, and local ternary patterns (LTP).
24. The computer readable program storage device of claim 16 , wherein the image is classified by a classifier based on the type of procedure being performed, wherein the procedure types include an electrophysiology procedure, an intravascular ultrasound procedure, a chronic total occlusion (CTO) procedure, and a balloon angioplasty procedure.