Purpose: MR-only treatment planning for skin High Dose Rate surface applicator brachytherapy (SABT) relies on accurate depiction of applicator channels on MR images. This work presents automatic detection of empty Freiburg Flap (FF) applicator catheters on optimized Pointwise Encoding Time Reduction with Radial Acquisition (PETRA) images, where catheters exhibit positive contrast.
Methods: An FF applicator (12 catheters with 24 spheres of 10mm diameter) was sandwiched between two plexiglass blocks and placed on a MagPhan TMR008 phantom in a 3T Siemens Vida. Coronal 0.8mm isotropic resolution 3D PETRA images (TR\TE\TI 3.3\0.07\100 ms, FOV 306x306mm², BW 407Hz/px) were acquired using UltraFlexLarge18 and Spine32 coils. A ROI encompassing a visible catheter and its surrounding beads on axial slices was drawn in MATLAB. After manual definition of the center column of the first sphere, the center and width of the following spheres was derived using known dimensions. Normalized signal intensity column profiles and their numerical gradients were computed. Profile mean and standard deviation (PM, SD) and gradient mean (GM) over the 13 columns width of each sphere was extracted. Locations of maximum PM found within 2 pixels of minima of SD and of GM were catheter position candidates for a sphere. A catheter was reconstructed by selecting the candidate position of a sphere within 1 pixel from its preceding one.
Results: Since FF catheters appeared as bright lines with dark borders inside bright circles, the characteristics of signal intensity profiles perpendicular to the catheters revealed accurately catheter position. The presented algorithm detected successfully all 12 catheters inside their 24 surrounding spheres. Six catheters were reconstructed as completely straight lines, five presented 0.8mm and one 1.6mm maximum deviation from a straight line.
Conclusion: Automated detection of empty catheters in FF applicators for skin SABT was feasible, based on their signal intensity properties on PETRA MRI.