Educative Impact Of Automatic Head And Neck Cancer Patients Delineation On Radiation Oncology Residents

Abstract

To evaluate the educational impact on radiation oncology residents in training when introducing an automatic segmentation software. Were assessed accuracy, consistency and speed of delineation regarding organs at risk (OARs) and prophylactic cervical lymph node levels (LNL) volumes in head and neck cancer patients. Two cases of head and neck cancers patients treated by exclusive intensity-modulated radiotherapy were delineated by an expert radiation oncologist and were considered as reference. Then, these two cases were delineated by residents divided into three groups: Group 1 (control group) composed of more experienced residents had never been exposed to automatic delineation system delineated manually, Group 2 (experimental group) composed of young residents on their first rotation trained with automatic delineation system during their learning period delineating manually prior (M-) and after using the automatic system (M+). The delineation accuracy was assessed using the Overlap Volume (OV) index for the OARs and LNL. The time needed to complete the delineation of the two cases was measured. Regarding the OARs, mean OV was 0.62 (SD = 0.05) for group 1, 0.56 (SD = 0.04) for group 2 M-, and 0.61 (SD = 0.03) for group 2 M+. Mean OV was higher in group 1 compared to group 2 M- (p = 0.01). There were no OV difference between group 1 and group 2 M+ (p = 0.67). The mean OV was higher in the group 2 M+ compared to the group 2 M- after paired analysis (p<0.003). Regarding LNL, mean OV was 0.53 (SD = 0.06) in group 1, 0.54 (SD = 0.03) in group 2 M-, and 0.58 (SD = 0.04) in group 2 M+. There was no significant difference between the groups despite a favorable trend in favor of the group 2 M+ (p = 0.079). Mean OV was higher in group 2 M+ for 11 of the 12 structures analyzed compared to group 2 M- (p = 0.016). Prior use of the automatic delineation software reduced the average contouring time per case by 34 to 40%. In this study of head and neck cancer delineation, prior use of atlas-based automatic segmentation by residents reduced the delineation duration, and led to reliable OARs and LNL delineations.