Premium
SU‐F‐BRA‐01: A Procedure for the Fast Semi‐Automatic Localization of Catheters Using An Electromagnetic Tracker (EMT) for Image‐Guided Brachytherapy
Author(s) -
Damato A,
Viswanathan A,
Cormack R
Publication year - 2015
Publication title -
medical physics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.473
H-Index - 180
eISSN - 2473-4209
pISSN - 0094-2405
DOI - 10.1118/1.4925212
Subject(s) - brachytherapy , standard deviation , imaging phantom , computer science , digitization , computer vision , frame (networking) , nuclear medicine , algorithm , mathematics , artificial intelligence , medicine , surgery , statistics , telecommunications , radiation therapy
Purpose: To evaluate the feasibility of brachytherapy catheter localization through use of an EMT and 3D image set. Methods: A 15‐catheter phantom mimicking an interstitial implantation was built and CT‐scanned. Baseline catheter reconstruction was performed manually. An EMT was used to acquire the catheter coordinates in the EMT frame of reference. N user‐identified catheter tips, without catheter number associations, were used to establish registration with the CT frame of reference. Two algorithms were investigated: brute‐force registration (BFR), in which all possible permutation of N identified tips with the EMT tips were evaluated; and signature‐based registration (SBR), in which a distance matrix was used to generate a list of matching signatures describing possible N‐point matches with the registration points. Digitization error (average of the distance between corresponding EMT and baseline dwell positions; average, standard deviation, and worst‐case scenario over all possible registration‐point selections) and algorithm inefficiency (maximum number of rigid registrations required to find the matching fusion for all possible selections of registration points) were calculated. Results: Digitization errors on average <2 mm were observed for N ≥5, with standard deviation <2 mm for N ≥6, and worst‐case scenario error <2 mm for N ≥11. Algorithm inefficiencies were: N = 5, 32,760 (BFR) and 9900 (SBR); N = 6, 360,360 (BFR) and 21,660 (SBR); N = 11, 5.45*1010 (BFR) and 12 (SBR). Conclusion: A procedure was proposed for catheter reconstruction using EMT and only requiring user identification of catheter tips without catheter localization. Digitization errors <2 mm were observed on average with 5 or more registration points, and in any scenario with 11 or more points. Inefficiency for N = 11 was 9 orders of magnitude lower for SBR than for BFR. Funding: Kaye Family Award