Premium
Poster — Thurs Eve‐29: Real‐time tumour tracking and dose adaptation utilizing 4D MR images
Author(s) -
Yun J,
Robinson D,
MacKenzie M,
Fallone BG
Publication year - 2008
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 - 10.1118/1.2965948
Subject(s) - scanner , contouring , imaging phantom , multileaf collimator , computer science , steady state free precession imaging , computer vision , magnetic resonance imaging , artificial intelligence , linear particle accelerator , feature (linguistics) , tracking (education) , nuclear medicine , physics , medicine , beam (structure) , radiology , computer graphics (images) , optics , psychology , pedagogy , linguistics , philosophy
Magnetic Resonance Imaging (MRI) is known to be the best imaging modality for soft‐tissue delineation. We will exploit this feature to develop and test a real‐time tumour tracking and dose adaptation algorithm based on 4D MR images, such as we hope to obtain using a Linac‐MR system under development by our research group. We have developed algorithms capable of auto‐contouring and tracking the motion of tumours in pseudo real‐time from 4D MR image sets acquired 4 times per second. To compensate for the time delays between imaging, image processing, the mechanical movement of Linac, and the actual treatment, an algorithm capable of predictively modeling tumour position based on 4D MR images, such as would be acquired just prior to patient treatment, is being developed. Based on these, an algorithm able to adjust Multileaf Collimator (MLC) movement and beam intensity depending on the position, depth, and density of tissue overlying the tumour as a function of time is under development. Our initial results, in which we used 4D MR images acquired from a 3T MR scanner with a Steady State Free Precession (SSFP) image sequence, demonstrated the feasibility of real‐time tumour contouring and tracking. By using an in‐house built MR compatible motion phantom and the same scanner with a Balanced Full Field Echo (BFFE) image sequence, we succeeded in acquiring dynamic MR images, and we confirmed the promising capability of our prediction algorithm, which can infer the position of tumour 0.25 seconds in advance.