SU‐E‐J‐101: Application of a Combination Method of Atlas‐Based and Model‐Based Auto‐Segmentation in Thorax‐And‐Abdomen Cancer

Purpose: Atlas‐based auto‐segmentation and model‐based auto‐segmentation were combined to improve accuracy of auto‐segmentation for thorax and abdomen cancer patient,herein referred to as AM 0.95 vs. 0.96 (p=0.42); 0.93 vs. 0.91 (p<0.01); 0.95 vs. 0.93 (p<0.001);and 0.93 vs. 0.88 (p<0.01);0.95 vs. 0.89 (p<0.01), 0.86 vs. 0.80 (p<0.001), respectively. Auto‐segmentation of all OARs using A&M method took about 12 minutes per patient in total, including ABAS 10 minutes and MBS 2 minutes. Conclusion: Our results demonstrated the AM method can significantly improve the results from ABAS only for most normal organs in thorax and abdomen cancer patients. And the AM method is more convention and accurate than MBS method only in clinical applications.