STENT VIEWING USING A LEARNING BASED CLASSIFIER IN MEDICAL IMAGING

• US 20110142318A1
• Filed: 12/06/2010
• Published: 06/16/2011
• Est. Priority Date: 12/10/2009
• Status: Active Grant

• Current Assignee: Siemens Healthcare GMBH (Siemens AG)
• Original Assignee: Siemens AG, Siemens Corp. (Siemens AG)
• Inventors: Lu, Xiaoguang, Comaniciu, Dorin, Chen, Terrence, Durlak, Peter, Pohl, Thomas
• Granted Patent: US 8,311,308 B2
• Time in Patent Office: Days
• US Class Current: 382/131
• CPC Class Codes: G06F 18/2148 characterised by the process organisation or structure, e.g. boosting cascade
• CPC Class Codes: G06T 2207/20224 Image subtraction
• CPC Class Codes: G06T 5/50 using two or more images, e.g. averaging or subtraction
• CPC Class Codes: G06V 10/446 using Haar-like filters, e.g. using integral image techniques
• CPC Class Codes: G06V 10/7747 Organisation of the process, e.g. bagging or boosting

Abstract

Stent viewing is provided in medical imaging. Stent images are provided with minimal or no user input of spatial locations. Images showing contrast agent are distinguished from other images in a sequence. After aligning non-contrast images, the images are compounded to enhance the stent. The contrast agent images are used to identify the vessel. A contrast agent image is aligned with the enhanced stent or other image to determine the relative vessel location. An indication of the vessel wall may be displayed in an image also showing the stent. A preview images may be output. A guide wire may be used to detect the center line for vessel identification. Various detections are performed using a machine-trained classifier or classifiers.

Claims

1. A method for stent viewing in medical imaging, the method comprising:

identifying from a plurality of frames of data a first sub-set of frames of data representing contrast agents, the frames of data representing a patient at different times during inflow of the contrast agent such that the first sub-set the frames of data represent the contrast agent and a second sub-set of the frames of data are free of contrast agent;

detecting markers for a stent in at least some of the frames of data;

selecting a reference contrast frame of data from the first sub-set of the frames of data, the reference contrast frame of data selected as a function of detected markers for the stent;

detecting a vessel wall in the reference contrast frame of data;

aligning the reference contrast frame of data with a stent output; and

outputting an image representing the stent, the image including a highlight of the vessel wall, the highlight aligned with the stent based on the aligning.

2. The method of claim 1 wherein identifying comprises identifying with a machine-trained classifier.

3. The method of claim 1 wherein identifying comprises:

determining, for locations represented by the frames of data, a score for each of the locations, the score based on whether the location is inside a vessel;

summing the scores for each of the frames of data; and

correlating a kernel representing in-flow with a plot of the summed scores as a function of relative time of the frames of data, the kernel representing the in-flow having an initial portion corresponding to no contrast agents.

4. The method of claim 1 further comprising:

selecting a third sub-set of the frames of data as a function of an ECG such that the third sub-set corresponds to low cardiac motion; and

outputting a pre-view image as a function of the third sub-set.

5. The method of claim 1 wherein detecting the markers comprises:

selecting a marker reference frame of data from the frames of data, the selecting being as a function of a distance between markers in the frames of data and probabilities of the markers in the frames being correct.

6. The method of claim 1 further comprising:

aligning at least some of the frames of data as a function of the markers;

combining the aligned at least some of the frames of data into a stent frame of data, the stent frame of data comprising the stent output; and

outputting a stent image as a function of the stent frame of data.

7. The method of claim 1 wherein aligning comprises rigidly transforming.

8. The method of claim 1 wherein detecting the vessel wall comprises:

detecting a guide wire;

forming a probabilistic map indicating a likelihood of locations being in a vessel;

deforming locations for the detected guide wire as a function of the probabilistic map; and

applying a cost function to the probabilistic map with the deformed locations of the guide wire as a positive seed, the application outputting locations of the vessel wall.

9. In a non-transitory computer readable storage medium having stored therein data representing instructions executable by a programmed processor for stent viewing in medical imaging, the storage medium comprising instructions for:

identifying images representing contrast agents from a fluoroscopy imaging sequence;

detecting a vessel wall from at least one of the identified images; and

generating an image of a stent from the fluoroscopy imaging sequence with the vessel wall enhanced relative to background.

10. The non-transitory computer readable storage medium of claim 9 wherein identifying the images representing contrast agents comprises identifying with a machine-trained classifier.

11. The non-transitory computer readable storage medium of claim 10 wherein identifying comprises:

determining, for locations represented by the images, a score for each of the locations, the score based on whether the location is inside a vessel and an output of the machine-trained classifier;

summing the scores for each of the images; and

correlating a kernel representing in-flow with a plot of the summed scores as a function of relative time of the images, the kernel representing the in-flow having an initial portion corresponding to no contrast agents.

12. The non-transitory computer readable storage medium of claim 9 wherein identifying comprises identifying an image corresponding to a beginning of in-flow of contrast agents in the fluoroscopy sequence.

13. The non-transitory computer readable storage medium of claim 9 wherein identifying comprises:

scaling the images to different scale levels;

for each scale level, applying steerable filters, the steerable filters enhancing ridges in the scaled images; and

classifying points along a ridge as being part of a vessel or not part of a vessel.

14. The non-transitory computer readable storage medium of claim 9 wherein generating comprises generating the image with a graphic overlay for the vessel wall.

15. The non-transitory computer readable storage medium of claim 9 further comprising:

detecting markers for the stent;

aligning the images as a function of the markers; and

compounding the aligned images;

wherein the image is generated as a function of the compounded images and includes a graphic overlay for the vessel wall.

16. The non-transitory computer readable storage medium of claim 9 wherein detecting the vessel wall comprises:

locating a center line of the vessel; and

determining vessel borders as a function of the center line;

wherein the locating and determining the vessel borders are performed without user indication of any location in the images.

17. A system for stent viewing in medical imaging, the system comprising:

a memory operable to store data representing a vessel with contrast agent and a guide wire for a stent;

a processor configured to detect the vessel from the data, determine a center line for the vessel from the guide wire, locate a vessel boundary from the vessel and the centerline, and generate an image from the data, the image highlighting the vessel boundary; and

a display operable to display the image.

18. The system of claim 17 wherein the processor is configured to detect the vessel from the data by forming a probabilistic map indicating a likelihood of locations being in the vessel using a machine-trained classifier.

19. The system of claim 18 wherein the processor is configured to determine the center line from the guide wire by detecting a guide wire and to locate the vessel boundary by deforming locations for the detected guide wire as a function of the probabilistic map and by applying a cost function to the probabilistic map with the deformed locations of the guide wire as a positive seed, the application outputting locations of the vessel boundary.

20. The system of claim 17 wherein the processor is configured to identify stent markers and align the vessel boundary with the stent as a function of the stent markers.

21. The system of claim 17 wherein the processor is configured to apply learning-based classifiers to detect balloon markers, the vessel boundary, and the guide wire.