Method and system of scanner automation for X-ray tube with 3D camera

Patent metadata

Field	Value
Publication number	US-9895131-B2
Application number	US-201514881526-A
Country	US
Kind code	B2
Filing date	Oct 13, 2015
Priority date	Oct 13, 2015
Publication date	Feb 20, 2018
Grant date	Feb 20, 2018

Abstract

A method and apparatus for X-ray tube scanner automation using a 3D camera is disclosed. An RGBD image of a patient on a patient table is received from a 3D camera mounted on an X-ray tube. A transformation between a coordinate system of the 3D camera and a coordinate system of the patient table is calculated. A patient model is estimated from the RGBD image of the patient. The X-ray tube is automatically controlled to acquire an X-ray image of a region of interest of the patient based on the patient model.

First claim

The invention claimed is:

1. A method for X-ray tube scanner automation, comprising:
receiving an RGBD image of a patient on a patient table from a 3D camera mounted on an X-ray tube;
calculating a transformation between a coordinate system of the 3D camera and a coordinate system of the patient table;
estimating a patient model from the RGBD image of the patient; and
automatically controlling the X-ray tube to acquire an X-ray image of a region of interest of the patient based on the patient model,
wherein estimating a patient model from the RGBD image of the patient comprises:
detecting a patient pose in the RGBD image using one or more machine learning-based pose classifiers;
detecting anatomical landmarks of the patient in the RGBD data based on the detected patient pose.

2. The method of claim 1 , wherein calculating a transformation between a coordinate system of the 3D camera and a coordinate system of the patient table comprises:
detecting table markers on the patient table in the RGBD image; and
estimating the transformation between the coordinate system of the 3D camera and the coordinate system of the patient table based on the detected table markers in the RGBD image.

3. The method of claim 2 , wherein the table markers on the patient table comprise a respective set of ring markers on opposite sides of the patient table with each set of ring markers having a number of rings each having an outer ring of a distinctive color and a white inner circle.

4. The method of claim 3 , wherein detecting table markers on the patient table in the RGBD image comprises:
detecting circular shapes in the RGBD image using a 3D Hough transform; and
determining whether the detected circular shapes in the RGBD image match one of the ring markers on the patient table based a brightness of the inner circle and a color distribution inside the outer ring.

5. The method of claim 1 , wherein estimating a patient model from the RGBD image of the patient comprises:
transforming a 3D point cloud representation of the RGBD image to align the RGBD image with a predetermined field of view of the patient table using the estimated transformation between the coordinate system of the 3D camera and the coordinate system of the patient table;
projecting the 3D point cloud representation to generate a reprojected image comprising a color and depth image pair; and
estimating the patient model using the reprojected image.

6. The method of claim 1 , wherein detecting a patient pose in the RGBD image using one or more machine learning-based pose detector comprises:
classifying the patient pose as head first or feet first using a first trained pose classifier; and
classifying the patient pose as prone or supine using one of a second trained pose classifier or a third trained pose classifier based on the classification of the patient pose as head first or feet first by the first trained pose classifier.

7. The method of claim 6 , wherein each of the first, second and, third trained pose classifiers is a multi-channel probabilistic boosting tree (PBT) classifier that extracts Haar features from a plurality of image channels associated with the RGBD image.

8. The method of claim 1 , wherein detecting anatomical landmarks of the patient in the RGBD data based on the detected patient pose comprises:
detecting each of a plurality of landmarks a directed acyclic graph (DAG) using respective machine learning-based landmark detectors for each of the plurality of landmarks based on relative positons of the plurality of anatomical landmarks in the DAG, wherein for each of the plurality of landmarks, the machine learning-based landmark detector is selected from a first trained landmark detector and trained landmark detector based on a classification of the patient pose as head first or feet first.

9. The method of claim 8 , wherein each of the machine learning-based landmark detectors for the plurality of landmarks is a multi-channel probabilistic boosting tree (PBT) classifier that extracts Haar features from a plurality of image channels associated with the RGBD image.

10. The method of claim 1 , wherein estimating a patient model from the RGBD image of the patient further comprises:
prior to classifying the patient pose, estimating a coarse position of the patient in the RGBD image using a machine learning-based full body detector.

11. The method of claim 1 , wherein estimating a patient model from the RGBD image of the patient further comprises:
estimating a plurality of body regions of the patient based on the detected anatomical landmarks.

12. The method of claim 1 , wherein estimating a patient model from the RGBD image of the patient further comprises:
fitting a human skeleton model to the RGBD image based on the detected anatomical landmarks.

13. The method of claim 1 , further comprising:
displaying the estimated patient model on a display device; and
receiving a user input of the region of interest of the patient on the displayed patient model.

14. The method of claim 1 , wherein calculating a transformation between a coordinate system of the 3D camera and a coordinate system of the patient table comprises:
receiving tube position control parameters of a control system of the X-ray tube; and
calculating a transformation between the coordinate system of the 3D camera and the coordinate system of the patient table using a kinematic calibration based on the tube position control parameters of the control system of the X-ray tube.

15. The method of claim 14 , wherein the kinematic calibration calibrates the coordinate system of the 3D camera, the coordinate system of the patient table, and coordinate systems of a kinematic chain of the X-ray tube control system.

16. The method of claim 15 , wherein the kinematic chain of the X-ray tube control system includes a tube origin, a tube holder vertical rotation center, a tube horizontal rotation center, and a collimator light field.

17. The method of claim 14 , wherein automatically controlling the X-ray tube to acquire an X-ray image of a region of interest of the patient based on the patient model comprises:
automatically adjusting the tube position control parameters of the control system of the X-ray tube to guide the X-ray tube to a target position and orientation in a coordinate system of the X-ray tube corresponding to the region interest on the estimated patient model in the coordinate system of the 3D camera using inverse kinematics based on the kinematic calibration.

18. The method of claim 17 , wherein the tube position control parameters of the control system of the X-ray tube comprise three translational parameters and two rotational parameters.

19. An apparatus for X-ray tube scanner automation, comprising:
means for receiving an RGBD image of a patient on a patient table from a 3D camera mounted on an X-ray tube;
means for calculating a transformation between a coordinate system of the 3D camera and a coordinate system of the patient table;
means for estimating a patient model from the RGBD image of the patient; and
means for automatically controlling the X-ray tube to acquire an X-ray image of a region of interest of the patient based on the patient model
wherein the means for calculating a transformation between a coordinate system of the 3D camera and a coordinate system of the patient table comprises:
means for detecting table markers on the patient table in the RGBD image; and
means for estimating the transformation between the coordinate system of the 3D camera and

Assignees

• Siemens Healthcare Gmbh

Inventors

• Chang Yao Jen
• Singh Vivek Kumar
• Ma Kai
• Nanke Ralf
• Dornberger Susanne
• Chen Terrence

Classifications

• A61B6/544

  dependent on patient size · CPC title

• A61B6/54 Primary

  Control of apparatus or devices for radiation diagnosis · CPC title

• A61B6/582

  Calibration · CPC title

• A61B6/0492

  using markers or indicia for aiding patient positioning · CPC title

• A61B6/488

  involving pre-scan acquisition · CPC title