Comparative Study On An Automatic Contouring Software For Delineating Organs At Risk In Intensity-Modulated Radiotherapy For Nasopharyngeal Carcinoma

To explore the speed, accuracy and user experience of self-developed automatic contouring software (software A) in delineating organs at risk (OARs) in intensity-modulated radiotherapy (IMRT) for nasopharyngeal carcinoma (NPC), compared with two commercially available radiotherapy treatment planning systems (TPS). The CT data of 850 patients with NPC were retrospectively reviewed to perform deep learning training with automatic delineation software, and then 100 patients with NPC who underwent IMRT were selected from 5 hospitals with different grades after the completion of learning. In each NPC case, 12 OARs of 8 types were delineated and timed using software A, B, and C, respectively (software B and C was currently marketed imported and domestic TPS, respectively), and the quality of automatic delineation was evaluated using the Dice similarity coefficient (DSC). The total score of software A user experience table was 100 points, including the fluency of using page, page layout and display function, etc., and 50 users graded software A. The delineation time of 12 OARs in NPC patients delineated with software A was 225.33 seconds (about 3.8 min), which was significantly (p = 0.000) shorter than that using software B (1081.31 seconds, about 18.0 min) and software C (1945.80 seconds, about 32.4 min). In terms of accuracy, the median DSC of automatic delineation using software A ranged from 0.76 to 0.90, and the maximum and minimum DSC value was 0.94 and 0.66 in bilateral eyes and oral cavity, respectively. It should be noted that the time for software A included manual modification. Among the 50 user experience scales, the lowest score was 88 points; the highest 96 points, and the average 91 points. In particular, a high score was obtained in the evaluation of software sensory experience with a full score of 40, with an average of 37. Software A can greatly improve the delineation speed while ensuring the accuracy of OARs contouring, showing superiority over the commercially available TPSs. By these findings, radiotherapy equipment may take a small step towards intelligent medical treatment and may be a clinical application tool with practical application value. In this study, digital human technology has been added at a later stage, this improving the accuracy of OARs delineation and correction.