Sunteți pe pagina 1din 3

FIDUCIAL LOCALIZATION IN MEDICAL IMAGING 

 
TEAM 
Nikhil Sheoran, 14114031 
Meet Vora, 14115138 
Kanish, 16117040 
Pragya Choudhary, 15112062 
 
GIVEN DATA 
1. An initial 3D image of the object with clearly visible Fiducials in two planes. (Sample 1) 

 
 
2. A set of images with distance increasing gradually in one direction, in a given plane. 
(Multiple planes provided) 
 
SOLUTION 
● Based on the initial images with Fiducial visibles, for each fiducial :  
○ The ​Starting Point [​ Algo 1]​ o
​ f the Fiducial 
○ The ​Ending Point​ [Algo 1] of the Fiducial 
○ Observe the lengths of major and minor axis of the ellipse formed, thus obtaining 
the angle between normal to the plane and the fiducial. 
● Based on the set of images in different views 
○ Obtain the coordinates of the Center point of the fiducial in that view (from above 
images) 
○ The existence of the fiducial can be verified by ​observing​ [Algo 2] a continuous 
segment of length L*cos(alpha) for changes along the images in the direction 
normal to the plane of observation, where L is the standard Fiducial length, and 
alpha is the angle fiducial makes with the normal to the plane of observation. 
○ The location of the Fiducial can then be confirmed in Patient Coordinate System 
by dividing the distance b/w first and last image in that many equal partitions and 
getting the proportionate coordinate. 
 
ALGORITHMS 
● ALGORITHM 1 (OBTAINING THE LOCATION OF AN ELLIPSE) 
○ Currently the idea is to use a standard ellipse detection algorithm, like 
■ Based on Sorted Merging 
https://www.hindawi.com/journals/tswj/2014/481312/ 
■ Using 1-D Hough Transform 
https://in.mathworks.com/matlabcentral/fileexchange/33970-ellipse-dete
ction-using-1d-hough-transform 
○ Further optimizations to this could be made by utilizing properties related to 
dicom 
■ Two ellipses (Start and End) would be connected by a line, 
■ There would be two ellipses with a similar angle (major and minor axis 
proportion) in close vicinity, and 
■ Utilizing the information about the same fiducial in a perpendicular view 
plane, etc. 
 
● ALGORITHM 2 (OBSERVING THE POINTS FOR OCCURRENCE OF A FIDUCIAL) 
○ Observe a window of distance length, L*cos( α ) from the sequence of images. 
○ If the images have a first distorted observation (differing significantly from 
previous view or irregular shape) and varying as well as shifting μ units for a 
continuous β number of images, where β = L*cos(α)
D
*N i
where Ni is number of 
images for that plane of view and D is the actual distance between first and last 
image in that view. 
○ Thus, we have obtained the location of the fiducial in the plane perpendicular to 
the already observed plane. 

S-ar putea să vă placă și