Room: AAPM ePoster Library
Auto contouring is one of the fastest tools for delineating the normal structures in busy radiation oncology department. In this study we have compared variation between auto contoured and manually contoured structures and its dosimetric effect.
A total of 15 Head and neck and 15 pelvis patients were retrospectively analysed in this study. Normal structures were manually drawn in the computed tomography (CT) images in eclipse treatment planning system. Auto contouring module of the TPS called smart segmentation (version 13.5) was used to contour the the same normal again. For head and Neck patients, spinal cord, brainstem, and parotids were delineated and for pelvis patients, bladder and rectum were drawn with this feature. The original treatment plans were assigned to the auto segmented structures and dosimeric analysis was done
The mean Hausdroff distance between ASS and MCS of all cases was within 3 mm for spinal cord , Left Parotid Right parotid bladder Right Femur Head, left Femur Head and Bone marrow ,but greater for brainstem and rectum . Mean deviation of maximum doses received was 5.6 Gy for spinal cord and 11.8 Gy for Brain stem. The average variation of mean doses was within 13 Gy for Parotids & Bone marrow and 17.6 Gy for rectum. It was within 3.2 Gy for Bladder, Left Femur Head, and Right Femur Head. The dosimetric differences between ACS and MCS of Rectum, Spinal cord, and Brainstem is due to difference in number of slices in which contours are drawn .Bladder and femur heads have clear boundaries , so the auto segmentation does fairly well .
Our study shows that when auto segmentation is used ,manual intervention is required to correct the minor deviation particularly in structures like brainstem, Rectum and parotids.