Program Information
An Automated Check of Cluster Distributions of Spot Maps as a Safety Procedure for IMPT Treatment
Y Zhang1*, B Jiang2,3 , X Wang2 , N Sahoo2 , M Gillin2 , X Zhu2 , X Zhang2 (1) University of Texas, Houston, Texas (2) The University of Texas MD Anderson Cancer Center, Houston, Texas (3) Tianjin Medical University Cancer Institute and Hospital,Tianjin,China
Presentations
SU-I-GPD-T-211 (Sunday, July 30, 2017) 3:00 PM - 6:00 PM Room: Exhibit Hall
Purpose: To develop an automated quality assurance procedure to detect the potential error in the proton spot scanning delivery process based on knowledge extraction of the previous delivered fields.
Methods: 1294 treatment fields from 2014-2016 in our center were analyzed for the knowledge extraction. Proton spot distributions are transferred to a binary matrix. A connected component labeling algorithm is developed to label all the spot clusters. An alert will be automatically sent out to prevent the potential accident or other mistakes if the field has more than one cluster by the filters developed in-house in our center.
Results: For the 1294 fields treated in our center, 1241 fields have one cluster. Among the 53 fields with more than one cluster, the 3mm/3% gamma passing rate for the dose distribution is below 70% for 26 fields at the distal end of the target. Three types of error will be automatically detected by the proposed algorithm: 1) the contouring error for the targets; 2) spots delivered outside of the patients; and 3) the unnecessary small cluster at the distal end causing the dose distribution not passing the QA criteria.
Conclusion: The automated quality assurance procedure based on data mining of the previous delivered fields can prevent the errors and improve the quality assurance process for the proton spot scanning delivery.
Contact Email: