Daily stereotactic ultrasound prostate targeting: inter-user variability

Purpose: Analysis of inter-user variability of patient setup for prostate RT using the BAT ultrasound-targeting device (Nomos Corp., Sewickley, PA). Methods: 18 prostate cancer patients participated in a study assessing variations in setup between users. Users were: 1 radiation oncologist (RO), 1 physicist, 5 RT (2 experienced, 3 inexperienced) and a radiologist. Study concept: (1) virtual patient alignment by all users based on standard axial and sagittal ultrasound images provided by the RO; (2) acquisition of individual freeze-frames by each user followed by virtual BAT shifts; (3) control CT-scan. We analyzed the range of BAT setups between users in x, y and z-axis and the residual setup error by CT/CT comparison and establishment of percent change in setup. Results: 163 virtual shifts were analyzed. The range of setups was 2.6± 1.5, 3.5± 1.1, and 4.4± 1.4 mm in x, y and z-direction for setups based on standard freeze-frames and 3.8 ±2.5, 5.1 ±3.0, and 5.6 ±2.8 mm based on individual freeze-frames. Experienced users reduced this range by 0.2 to 2.6 mm. Mean residual setup errors were 1.1 ±1.1 mm, 1.7± 1.2 mm, and 2.6± 1.8 mm in x, y, and z-axes (maximum observed residual errors 3.7, 4.6, and 6.0 mm). Mean reduction in setup error was 66.4 ±21.5% in experienced and 39.1 ±29.5% in inexperienced users (P < 0.0001). 3 of 163 virtual setups would have introduced new larger setup errors than the magnitude of the initial setup error. Conclusion: Ultrasound targeting for prostate cancer RT can be performed with high inter-user consistency and does lead to improved treatment setup in 98% of alignments. Although setup improvements were accomplished in virtually all analyzed cases, data derived from this study indicate that training in the appropriate use of this imaging technology, new to most radiation oncologists and therapists, is critical. Experienced use is correlated with a reduced variation of setups between users and higher degree of setup improvement when compared with performance of users new to the system.