Introduction
Clinical challenge
State of art
Contribution of this paper
Structure of the paper
Methods
Bladder phantom design
Data acquisition
Reconstruction algorithms
Variable definitions
-
\(\Psi _w\): 3-D world coordinate frame
-
\(\Psi _c\): local 3-D coordinate frame attached to camera (origin at pinhole; z-axis in viewing direction)
-
\(H^w_{c_i}\): camera-world coordinate transformation matrix for camera frame i
-
\(N_F\): number of camera frames
-
\(N_L\): number of homologous points
-
\(N_{E_i}\): number of feature points in camera frame i
-
\(N_{A_j}\): number of camera frames in which homologous point j is visible
-
\(p_j\): location of j-th homologous point; \(p^w_j\): the 3-D location vector expressed in world coordinates
-
P: point cloud consisting of all reconstructed \(p_j\)
-
Q: vertices of surface mesh
-
\(G_j\): biconnected graph of all camera features corresponding to homologous point j with co-registrations as edges
Feature detection, matching and robust identification of homologous points
Point cloud generation
3-D mesh reconstruction, texture and 2-D atlas generation
Inter-session comparison
Results
In vivo study
Phantom study
Part | Time (mm:ss) |
---|---|
Feature detection | 0:14 |
Pairwise image registration (1st run) | 7:58 |
Improving pairwise image registration | 0:45 |
Point cloud construction with bundle adjustment | 31:00 |
2-D Atlas generation | 3:18 |
Manual steps and corrections | 1:00 |
Total | 44:15 |