A novel method for interactive multi-objective dose-guided patient positioning. (16th December 2016)

• Record Type: Journal Article
• Authors: Haehnle, Jonas; Süss, Philipp; Landry, Guillaume; Teichert, Katrin; Hille, Lucas; Hofmaier, Jan; Nowak, Dimitri; Kamp, Florian; Reiner, Michael; Thieke, Christian; Ganswindt, Ute; Belka, Claus; Parodi, Katia; Küfer, Karl-Heinz; Kurz, Christopher
• Is Part Of: Physics in medicine & biology. Volume 62:Number 1(2017:Jan.)
• Page Start: 165
• Page End: 185
• Publication Date: 2016-12-16
• Subjects: dose-guided positioning -- multi-criteria optimization -- image-guided radiotherapy; Biophysics -- Periodicals; Medical physics -- Periodicals; 610.153
• DOI: 10.1088/1361-6560/62/1/165
• Languages: English
• ISSNs: 0031-9155

Abstract

Abstract: In intensity-modulated radiation therapy (IMRT), 3D in-room imaging data is typically utilized for accurate patient alignment on the basis of anatomical landmarks. In the presence of non-rigid anatomical changes, it is often not obvious which patient position is most suitable. Thus, dose-guided patient alignment is an interesting approach to use available in-room imaging data for up-to-date dose calculation, aimed at finding the position that yields the optimal dose distribution. This contribution presents the first implementation of dose-guided patient alignment as multi-criteria optimization problem. User-defined clinical objectives are employed for setting up a multi-objective problem. Using pre-calculated dose distributions at a limited number of patient shifts and dose interpolation, a continuous space of Pareto-efficient patient shifts becomes accessible. Pareto sliders facilitate interactive browsing of the possible shifts with real-time dose display to the user. Dose interpolation accuracy is validated and the potential of multi-objective dose-guided positioning demonstrated for three head and neck (H&N) and three prostate cancer patients. Dose-guided positioning is compared to replanning for all cases. A delineated replanning CT served as surrogate for in-room imaging data. Dose interpolation accuracy was high. Using a 2 % dose difference criterion, a median pass-rate of 95.7% for H&N and 99.6% for prostate cases was determined in a comparison to exact dose calculations. For all patients, dose-guided positioning allowed to find a clinically preferable dose distribution compared to bony anatomy based alignment. For all H&N cases, mean dose to the spared parotid glands was below 26 G y (up to 27.5 G y with bony alignment) and clinical target volume (CTV) V 95 % above 99.1% (compared to 95.1%). For all prostate patients, CTV V 95 % was above 98.9% (compared to 88.5%) and V 50 G y to the rectum below 50 % (compared to 56.1%). Replanning yielded improved results for the H&N cases. For the prostate cases, differences to dose-guided positioning were minor.