Method and system for registration of ultrasound and physiological models to X-ray fluoroscopic images

Patent metadata

Field	Value
Publication number	US-9687204-B2
Application number	US-201213475048-A
Country	US
Kind code	B2
Filing date	May 18, 2012
Priority date	May 20, 2011
Publication date	Jun 27, 2017
Grant date	Jun 27, 2017

Abstract

A method and system for registering ultrasound images and physiological models to x-ray fluoroscopy images is disclosed. A fluoroscopic image and an ultrasound image, such as a Transesophageal Echocardiography (TEE) image, are received. A 2D location of an ultrasound probe is detected in the fluoroscopic image. A 3D pose of the ultrasound probe is estimated based on the detected 2D location of the ultrasound probe in the fluoroscopic image. The ultrasound image is mapped to a 3D coordinate system of a fluoroscopic image acquisition device used to acquire the fluoroscopic image based on the estimated 3D pose of the ultrasound probe. The ultrasound image can then be projected into the fluoroscopic image using a projection matrix associated with the fluoroscopic image. A patient specific physiological model can be detected in the ultrasound image and projected into the fluoroscopic image.

First claim

What is claimed is:

1. A method for registering an ultrasound image acquired using an ultrasound probe to a fluoroscopic image acquired using a fluoroscopic image acquisition device, comprising:
detecting a 2D location of the ultrasound probe in the fluoroscopic image using a trained machine learning based probe detector that extracts features from image patches of the fluoroscopic image, determines a probability score for each image patch, and selects the image patch having the highest probability score as the 2D location of the ultrasound probe;
estimating an initial machine learning based 3D pose of the ultrasound probe based on the detected 2D location of the ultrasound probe in the fluoroscopic image by initializing pose estimation using X and Y coordinates estimated from the detected 2D location of the ultrasound probe, estimating a 3D position of the ultrasound probe including X, Y, and Z coordinates based on the X and Y coordinates estimated from the detected 2D location by applying a trained machine learning based position classifier to the fluoroscopic image, and estimating a 3D position and orientation of the ultrasound probe including X, Y, and Z coordinates and a roll, pitch, and yaw based on the 3D position estimating using the trained machine learning based position classifier by applying a trained machine learning based position and orientation classifier to the fluoroscopic image;
iteratively refining the estimated initial machine learning based 3D pose of the ultrasound probe using 2D/3D registration based on the ultrasound and the fluoroscopic image to estimate a final 3D pose of the ultrasound probe;
mapping the ultrasound image to a 3D coordinate system of the fluoroscopic image acquisition device based on the estimated final 3D pose of the ultrasound probe;
estimating a patient specific physiological model of an anatomical structure in the ultrasound image; and
projecting the patient specific physiological model of the anatomical structure into the fluoroscopic image using a projection matrix associated with the fluoroscopic image.

2. The method of claim 1 , wherein the fluoroscopic image is a frame in a fluoroscopic image sequence the step of detecting a 2D location of the ultrasound probe in the fluoroscopic image comprises:
determining a reduced search space in the fluoroscopic image based on a probe location predicted from a detected probe location in a previous frame in the fluoroscopic image sequence; and
detecting the 2D location of the ultrasound probe in the reduced search space using a trained probe detector.

3. The method of claim 1 , wherein the estimated final 3D pose is a position and orientation of the ultrasound probe in the 3D coordinate system of the fluoroscopic image acquisition device used to acquire the fluoroscopic image.

4. The method of claim 3 , wherein the step of mapping the ultrasound image to a 3D coordinate system of the fluoroscopic image acquisition device based on the estimated final 3D pose of the ultrasound probe comprises:
calculating a transform to map the ultrasound image to the 3D coordinate system of the fluoroscopic image acquisition device based on the position and orientation of the ultrasound probe corresponding to the estimated final 3D pose using ultrasound configuration parameters.

5. The method of claim 1 , wherein the step of iteratively refining the estimated initial machine learning based 3D pose of the ultrasound probe using 2D/3D registration based on the ultrasound and the fluoroscopic image to estimate a final 3D pose of the ultrasound probe comprises:
(a) aligning a 3D model of the ultrasound probe to current estimated 3D pose;
(b) generating a digital reconstructed radiograph (DRR) from the aligned 3D model of the ultrasound probe;
(c) measuring a difference value between the fluoroscopic image and the DRR; and
(d) refining the current estimated 3D pose to reduce the difference value between the fluoroscopic image and the DRR.

6. The method of claim 5 , wherein the step of iteratively refining the estimated initial machine learning based 3D pose of the ultrasound probe using 2D/3D registration based on the ultrasound and the fluoroscopic image to estimate a final 3D pose of the ultrasound probe further comprises:
(e) repeating steps (a)-(d) until the difference value between the fluoroscopic image and the DRR is less than a threshold.

7. The method of claim 1 , further comprising:
projecting the ultrasound image into the fluoroscopic image using a projection matrix associated with the fluoroscopic image.

8. The method of claim 1 , wherein the ultrasound image is a transesophageal echocardiography (TEE) image and the ultrasound probe is a TEE ultrasound probe used to acquire the TEE image.

9. An apparatus for registering an ultrasound image acquired using an ultrasound probe to a fluoroscopic image acquired using a fluoroscopic image acquisition device, comprising:
a processor; and
a memory storing computer program instructions, which when executed by the processor cause the processor to perform operations comprising:
detecting a 2D location of the ultrasound probe in the fluoroscopic image using a trained machine learning based probe detector that extracts features from image patches of the fluoroscopic image, determines a probability score for each image patch, and selects the image patch having the highest probability score as the 2D location of the ultrasound probe;
estimating an initial machine learning based 3D pose of the ultrasound probe based on the detected 2D location of the ultrasound probe in the fluoroscopic image by initializing pose estimation using X and Y coordinates estimated from the detected 2D location of the ultrasound probe, estimating a 3D position of the ultrasound probe including X, Y, and Z coordinates based on the X and Y coordinates estimated from the detected 2D location by applying a trained machine learning based position classifier to the fluoroscopic image, and estimating a 3D position and orientation of the ultrasound probe including X, Y, and Z coordinates and a roll, pitch, and yaw based on the 3D position estimating using the trained machine learning based position classifier by applying a trained machine learning based position and orientation classifier to the fluoroscopic image;
iteratively refining the estimated initial machine learning based 3D pose of the ultrasound probe using 2D/3D registration based on the ultrasound and the fluoroscopic image to estimate a final 3D pose of the ultrasound probe;
mapping the ultrasound image to a 3D coordinate system of the fluoroscopic image acquisition device based on the estimated final 3D pose of the ultrasound probe;
estimating a patient specific physiological model of an anatomical structure in the ultrasound image; and
projecting the patient specific physiological model of the anatomical structure into the fluoroscopic image.

10. The apparatus of claim 9 , wherein the estimated final 3D pose of the ultrasound probe is a position and orientation of the ultrasound probe in the 3D coordinate system of the fluoroscopic image acquisition device.

11. The apparatus of claim 10 , wherein mapping the ultrasound image to a 3D coordinate system of the fluoroscopic image acquisition device based on the estimated final 3D pose of the ultrasound probe comprises:
calculating a transform to map the ultrasound image to the 3D coordinate system of the fluoroscopic image acquisition device based on the position and orientation of the ultrasound probe corresponding to the estimated final 3D pose using ultrasound configuration parameters.

12. The apparatus of claim 9 , further comprising:
means for projecting

Assignees

• Mountney Peter
• Kaiser Markus
• Voigt Ingmar
• John Matthias
• Ionasec Razvan Ioan
• Boese Jan
• Comaniciu Dorin
• Siemens Healthcare Gmbh

Inventors

• Mountney Peter
• Kaiser Markus
• Voigt Ingmar
• John Matthias
• Ionasec Razvan Ioan
• Boese Jan
• Comaniciu Dorin

Classifications

• A61B6/4417

  related to combined acquisition of different diagnostic modalities · CPC title

• A61B6/4441

  the rigid structure being a C-arm or U-arm · CPC title

• A61B6/5247

  combining images from an ionising-radiation diagnostic technique and a non-ionising radiation diagnostic technique, e.g. X-ray and ultrasound · CPC title

• A61B8/4416

  related to combined acquisition of different diagnostic modalities, e.g. combination of ultrasound and X-ray acquisitions · CPC title

• A61B8/0883

  for diagnosis of the heart · CPC title