Development and evaluation of an automatic interstitial catheter digitization tool for adaptive high-dose-rate brachytherapy

Abstract Purpose To develop and evaluate an automatic interstitial catheter digitization algorithm for use in adaptive high-dose-rate brachytherapy for gynecologic cancers using the Syed-Neblett template. Methods and Materials We developed an automatic catheter digitization tool, which uses a region growing algorithm in conjunction with a spline model of the catheters. Seed locations were selected in each catheter for the region growing algorithm. The region growing was constrained by a spline model of the catheters, which prevents intercatheter crossover or incorrect digitization due to air pockets. Plan reoptimization was performed on successive day computed tomography scans using dwell positions for the Day 1 computed tomography. This method was applied to 10 patients who had received high-dose-rate interstitial brachytherapy using the Syed–Neblett template. The prescribed dose was 18.75 or 20 Gy delivered in five fractions, twice daily, and more than 3 consecutive days. Dosimetric comparisons were made between automatically and manually digitized plans. Results The region growing algorithm was able to successfully digitize all catheters. The mean difference between automatic and manually digitized positions was 0.4 ± 0.2 mm. No significant difference was found in dosimetric parameters between the automatic and manually digitized plans. The mean D90% of the clinical target volume over all 3 days of treatment of the manual vs. reoptimized automatic plans was 94.3 ± 6.58% and 92.32 ± 8.34%, respectively ( p = 0.50). Conclusions The algorithm discussed in this article is the first developed for adaptive interstitial brachytherapy for a large number of catheters (14 on average). The algorithm has future potential in digitization quality assurance. A region growing algorithm was developed to automatically digitize interstitial catheters in high-dose-rate brachytherapy. This automatic digitization tool was shown to be accurate compared with manual digitization.