Applicator reconstruction and applicator shifts in 3D MR-based PDR brachytherapy of cervical cancer

To evaluate the methods of applicator reconstruction in 3D MR-based planning for brachytherapy of cervical cancer, and to investigate applicator shifts and changes in DVH parameters during PDR treatment.

For each application MR scans with applicator in situ were made: three T2-weighted (4.5 mm slices) Turbo Spin Echo (TSE) scans and a balanced Steady State Free Precession scan (1.5 mm). Three observers tested two applicator reconstruction methods: (A) directly on the bSSFP scan and (B) on a resampled combination of the three T2-weighted scans. For 10 patients MR imaging was repeated on the second day of each PDR fraction to determine applicator shifts and changes in DVH parameters.

For both applicator reconstruction methods the interobserver variation for the DVH parameters was comparable (average <1.5% in dose). Differences between the two methods were larger (up to 6.4% for target) and were related to position differences during MR scanning. The average applicator shift relative to the pelvic structures was 5-6 mm into the ventral direction and 3-4 mm cranially. For a single PDR fraction, the average D90 (HR-CTV) on 'day 2' was 0.2 (SD 2.0) Gy lower than that for day 1. The average increase in D(2 cc) (bladder) was 1.0 (SD 3.0) Gy(alphabeta3) for a single PDR fraction. If the effect of both fractions was combined, for 1 patient a total decrease of D90 of 7 Gy(alphabeta10) was found, whereas for another patient the total increase in bladder dose was 12 Gy(alphabeta3).

Applicator reconstruction on MR data is feasible. In the overall accuracy during PDR brachytherapy the reconstruction uncertainty is of minor importance. Applicator and/or organ movement during the course of the PDR fraction produce larger uncertainties.