Guidance method based on 3d-2d pose estimation and 3d-ct registration with application to live bronchoscopy

Abstract

A method provides guidance to the physician during a live bronchoscopy or other endoscopic procedures. The 3D motion of the bronchoscope is estimated using a fast coarse tracking step followed by a fine registration step. The tracking is based on finding a set of corresponding feature points across a plurality of consecutive bronchoscopic video frames, then estimating for the new pose of the bronchoscope. In the preferred embodiment the pose estimation is based on linearization of the rotation matrix. By giving a set of corresponding points across the current bronchoscopic video image, and the CT-based virtual image as an input, the same method can also be used for manual registration. The fine registration step is preferably a gradient-based Gauss-Newton method that maximizes the correlation between the bronchoscopic video image and the CT-based virtual image. The continuous guidance is provided by estimating the 3D motion of the bronchoscope in a loop. Since depth-map information is available, tracking can be done by solving a 3D-2D pose estimation problem. A 3D-2D pose estimation problem is more constrained than a 2D-2D pose estimation problem and does not suffer from the limitations associated with computing an essential matrix. The use of correlation-based cost, instead of mutual information as a registration cost, makes it simpler to use gradient-based methods for registration.

1 . A video-based method for providing a pose estimate of an endoscope in conjunction with a live endoscopic procedure, the method comprising:
 acquiring 3D image data of a target structure in advance of a live endoscopic procedure;   receiving a frame of live endoscopic video image data including the target structure; and   registering the frame of endoscopic video image data with acquired 3D image data to determine a pose estimate of the endoscope.   
     
     
         2 . The method of  claim 1 , wherein the pose estimate is used to guide the endoscope to a suspect site. 
     
     
         3 . The method of  claim 2 , wherein the suspect site is a mediastinal lymph node. 
     
     
         4 . The method of  claim 1 , wherein the endoscope is a bronchoscope. 
     
     
         5 . The method of  claim 1 , wherein the target structure is the airway tree. 
     
     
         6 . The method of  claim 5 , wherein the registration is performed either near to, or far from, a bifurcation I an airway tree. 
     
     
         7 . The method of  claim 1 , wherein the registration comprises an initial registration step to initiate guidance of the endoscope. 
     
     
         8 . The method of  claim 1 , wherein the registration step includes the steps of:
 a) estimating a three-dimensional location of the endoscope using (i) known motion information from said live endoscopic video image data, and (ii) local depth information obtained from the previously acquired 3D image data; and   b) determining a new pose of the endoscope based on the 3D location estimated in step (a).   
     
     
         9 . The method of  claim 8 , further including the step of performing a fine registration step to minimize errors associated with estimating the 3D motion of the endoscope. 
     
     
         10 . The method of  claim 9 , wherein the fine registration step is performed manually. 
     
     
         11 . The method of  claim 8 , wherein the registration step further includes the steps of:
 (c) updating the previously acquired 3D image data in accordance with the new pose; and   (d) repeating steps (a) through (c) until the guidance is terminated.   
     
     
         12 . The method of  claim 1 , including the step of receiving a sequence of consecutive video image frames as opposed to a single frame. 
     
     
         13 . The method of  claim 12 , wherein the sequence of video image frames is 5 consecutive video frames. 
     
     
         14 . A video-based method for registering previously acquired 3D image data and live endoscopic video image data of a patient to obtain a current pose of an endoscope comprising:
 a) performing an initial registration to register the previously acquired 3D image data and a frame of the live video image data to obtain a current pose and a current depth map;   b) selecting a plurality of points associated with the live video image data;   c) tracking the points over a plurality of consecutive frames to estimate the two-dimensional (2D) motion of the tracked points;   d) deriving a three-dimensional (3D) motion of the endoscope using the 2D motion of the tracked points and the current depth map;   e) determining a new pose based on the current depth map and 3D motion of the endoscope; and   f) updating the current pose based upon the new pose.   
     
     
         15 . The method of  claim 14 , wherein steps b) through f) are repeated until reaching a suspect site. 
     
     
         16 . The method of  claim 15 , wherein said site is along an airway. 
     
     
         17 . A video-based method for providing a pose estimate of an endoscope in conjunction with a live endoscopic procedure, the method comprising:
 receiving a frame of live endoscopic video image data of a target structure from the endoscope; and   registering previously acquired 3D image data of the target structure and the frame of endoscopic video image data of the target structure to provide a pose estimate of the endoscope; and   wherein the step of registering is performed using information arising from said frame of endoscopic video image data and the previously acquired 3D image data.   
     
     
         18 . The method of  claim 17 , wherein the step of registering the previously acquired 3D image data and the live video image data is used for guiding said endoscope to a suspect site. 
     
     
         19 . The method of  claim 18 , wherein said suspect site is a mediastinal lymph node. 
     
     
         20 . The method of  claim 17 , wherein the endoscope is a bronchoscope. 
     
     
         21 . The method of  claim 17 , wherein said target structure is the airway tree. 
     
     
         22 . The method of  claim 21 , wherein said registration is performed either near or far from a bifurcation. 
     
     
         23 . The method of  claim 17 , wherein said registering step is an initial registration step to start guidance of the endoscope. 
     
     
         24 . The method of  claim 18 , wherein the registration step includes the steps of:
 a) estimating a three-dimensional location of the endoscope using (i) known motion information from said live endoscopic video image data, and (ii) local depth information obtained from the previously acquired 3D image data; and   b) determining a new pose of the endoscope based on the 3D location estimated in step (a).   
     
     
         25 . The method of  claim 24 , further including the step of performing a fine registration step to minimize errors associated with estimating the 3D motion of the endoscope. 
     
     
         26 . The method of  claim 25 , wherein the fine registration step is performed manually. 
     
     
         27 . The method of  claim 24 , wherein the registration step further includes the steps of:
 (c) updating the previously acquired 3D image data in accordance with the new pose; and   (d) repeating steps (a) through (c) until the guidance is terminated.   
     
     
         28 . The method of  claim 17 , wherein said receiving comprises receiving a sequence of consecutive video image frames. 
     
     
         29 . The method of  claim 28 , wherein said sequence of video image frames is 5 consecutive video frames. 
     
     
         30 . A video-based method for registering previously acquired 3D image data and live endoscopic video image data of a patient to obtain a current pose of an endoscope comprising:
 g) performing an initial registration to register the previously acquired 3D image data and a frame of the live video image data to obtain a current pose and a current depth map;   h) selecting a plurality of points associated with the live video image data;   i) tracking the points over a plurality of consecutive frames to estimate the two-dimensional (2D) motion of the tracked points;   j) deriving a three-dimensional (3D) motion of the endoscope using the 2D motion of the tracked points and the current depth map;   k) determining a new pose based on the current depth map and 3D motion of the endoscope; and   l) updating the current pose based upon the new pose.   
     
     
         31 . The method of  claim 30  comprising repeating steps b) through f) until reaching a suspect site. 
     
     
         32 . The method of  claim 31 , wherein said site is along an airway.