Methods for automatic segmentation and temporal tracking

• US 8,483,432 B2
• Filed: 12/23/2009
• Issued: 07/09/2013
• Est. Priority Date: 12/23/2009
• Status: Active Grant

• Current Assignee: General Electric Company
• Original Assignee: General Electric Company
• Inventors: Patwardhan, Kedar Anil, Gupta, Sandeep N, Mills, David M, Dentinger, Aaron M, Yu, Yongjian
• Primary Examiner(s): Abdi, Amara
• Application Number:
  US12/645,781
• Publication Number:
  US 20110150274A1
• Time in Patent Office: 1,294 Days
• Field of Search: 382/103, 382128-132, 382/134, 382/173, 382/209, 600/407, 600/458, 600/467
• US Class Current: 382/103
• CPC Class Codes: A61B 8/0891 for diagnosis of blood vessels
• CPC Class Codes: A61B 8/483 involving the acquisition of a 3D volume of data
• CPC Class Codes: A61B 8/523 for generating planar views from image data in a user selectable plane not corresponding to the acquisition plane
• CPC Class Codes: G06T 2207/10136 3D ultrasound image
• CPC Class Codes: G06T 2207/20101 Interactive definition of point of interest, landmark or seed
• CPC Class Codes: G06T 2207/20116 Active contour; Active surface; Snakes
• CPC Class Codes: G06T 2207/30101 Blood vessel; Artery; Vein; Vascular
• CPC Class Codes: G06T 7/0012 Biomedical image inspection
• CPC Class Codes: G06T 7/149 involving deformable models, e.g. active contour models
• CPC Class Codes: G06T 7/251 involving models
• CPC Class Codes: G06T 7/277 involving stochastic approaches, e.g. using Kalman filters

Abstract

In one embodiment, a method of detecting centerline of a vessel is provided. The method comprises steps of acquiring a 3D image volume, initializing a centerline, initializing a Kalman filter, predicting a next center point using the Kalman filter, checking validity of the prediction made using the Kalman filter, performing template matching, updating the Kalman filter based on the template matching and repeating the steps of predicting, checking, performing and updating for a predetermined number of times. Methods of automatic vessel segmentation and temporal tracking of the segmented vessel is further described with reference to the method of detecting centerline.

Claims

1. A method of detecting and tracking a centerline of a vessel in real-time, the method comprising:

acquiring a 3D image volume;

initializing the centerline;

initializing a Kalman filter;

predicting a next center point using the Kalman filter;

checking validity of the prediction;

updating the centerline based on the predicted next center point;

performing template matching for estimating measurement error;

updating the Kalman filter based on the template matching; and

repeating the steps of predicting, checking, updating performing and updating for a predetermined number of times.

2. The method of claim 1, wherein the step of initializing comprises:

selecting an initial center point from the image volume;

acquiring a template of a vessel cross section from the 3D image volume; and

setting up centerline tracking parameters based on the initial center point and the acquired late.

3. The method of claim 2, wherein the step of acquiring template comprises:

obtaining a mask from thresholding and morphological processing of image data;

estimating approximate vessel dimensions based on the mask obtained;

specifying a gate size based on the vessel dimensions; and

creating a 2D template of the vessel cross-section centered at the initial center point and having the specified gate size.

4. The method of claim 3, wherein, the vessel dimension represents an internal diameter of a vessel segment.

5. The method of claim 2, wherein setting up centerline tracking parameters comprises:

obtaining one or more Kalman filter parameters, the Kalman filter parameters comprising process noise variance (Q), measurement error variance (R), prediction error variance (P) and Kalman gain (K); and

obtaining one or more template matching parameters, the template matching parameters comprising initial observation window size, motion vector and match-score threshold.

6. The method of claim 1, wherein initializing the Kalman filter comprises:

initializing the Kalman filter parameters using a preset configuration file.

7. The method of claim 5, wherein predicting the next center point comprises:

providing a first estimated center point and velocity;

providing values for Kalman filter parameters;

performing Kalman prediction;

obtaining a predicted center point and velocity; and

updating prediction error variance based on the predicted center point and velocity.

8. The method of claim 7, wherein the first estimated center point is the initial center point for a first iteration.

9. The method of claim 7, wherein checking validity of the prediction comprises:

checking if the predicted center point lies within a predetermined volume; and

checking if the prediction error variance is below an estimation threshold.

10. The method of claim 7, wherein the step of performing template matching comprises:

providing the predicted center point and velocity as inputs;

providing updated prediction error variance;

performing template matching;

obtaining a measured center point and velocity; and

,

obtaining measurement error variance.

11. The method of claim 10, wherein the template matching is performed based on Rayleigh approximation.

12. The method of claim 10, wherein the step of updating the Kalman filter comprises:

providing the measured center point and velocity and measurement error variance as inputs;

obtaining a second estimated center point and velocity; and

updating the variable Kalman Gain (K).

13. The method of claim 12, wherein the second estimated center point is the next center point.

14. A method of segmenting and tracking a centerline of a vessel comprising a plurality of cross sections in real-time, the method comprising:

acquiring a 3D image volume;

initializing one or more Kalman filter parameters using a preset configuration file;

selecting an initial center point within the vessel to be segmented;

performing an initial segmentation on a first image slice based on the initial center point, the first image slice being a 2D cross-section of the vessel;

creating a template of a vessel cross section based on the initial segmentation;

predicting a next center point that is a translation of the initial point along a beam direction using the Kalman filter;

correcting the next center point based on a measurement of the next center point in the image volume;

updating a centerline of the 3D image volume based on the next center point;

segmenting a second image slice based on the next center point; and

repeating the steps of predicting, correcting and segmenting until the vessel has been completely segmented in the 3D image volume.

15. The method of claim 14, wherein using the Kalman filter to predict a next center point comprises:

defining, a transition matrix for translating the initial center point by a spatial displacement along the beam direction, defining a prediction error variance P for the initial center point, predicting a next center point and a next prediction error variance based on the transition matrix, defining process noise variance for the prediction error variance and the series of center point measurements, correcting the center point prediction based on a measurement error variance, a measurement of the predicted center point, and the process noise variance of the center point measurements, and correcting the prediction error variance based on the measurement error variance and the process noise variance of the center point measurements.